Scolaro Elia 1197578

Università degli Studi di Padova
Dipartimento di Ingegneria Industriale (DII)

Corso di Laurea Magistrale in Ingegneria dell’Energia Elettrica
Tesi di Laurea Magistrale
Design of Electric Motors

for a Hybrid Farming Tractor
Studente Relatore
Elia Scolaro Prof. Luigi Alberti
Matricola 1197578
Correlatore
Dott. Diego Troncon
Anno Accademico 2019-2020

Contents
Riassunto esteso (summary in italian language) 4
Introduction 7
1 Powertrain electrification 9
1.1 Hybridization factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2 Original powertrain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3 Hybrid powertrain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4 Power management system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2 Case study: open field farming tractor 15

2.1 Data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1.1 Plowing tests measurements . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1.2 Power harrow test measurements . . . . . . . . . . . . . . . . . . . . . . . 16
2.1.3 Transportation test measurements . . . . . . . . . . . . . . . . . . . . . . 19
2.2 Mechanical circuit model and data pre-processing . . . . . . . . . . . . . . . . . . 19
2.3 Data post-processing: thermal equivalent torque . . . . . . . . . . . . . . . . . . 23
2.4 Engines selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.5 Hybridization tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.6 Hybridization results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3 Electrical machines sizing 31

3.1 Machine description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2 Electric drive and control system . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.3 Design specifications and volume constrains . . . . . . . . . . . . . . . . . . . . . 37
3.4 Choice of the permanent magnet . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.5 Preliminary sizing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.6 Winding design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.6.1 Voltage specifications and number of turns choice . . . . . . . . . . . . . . 49
3.6.2 Current specification and conductor sizing . . . . . . . . . . . . . . . . . . 51
3.7 Stator sizing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.8 Rotor and shaft sizing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.9 Volume limits fulfillment and materials cost computation . . . . . . . . . . . . . 57
4 Electromagnetic analysis 60
4.1 Performances analysis in unload condition . . . . . . . . . . . . . . . . . . . . . . 60
4.1.1 Cogging torque . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.2 MTPA search . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.3 Performances analysis at rated load and at maximum overload . . . . . . . . . . 69
4.4 Demagnetization test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.5 Inductance computation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.6 Winding resistance and Joule losses . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.7 Voltage specifications fulfillment . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.8 Iron and permanent magnets losses . . . . . . . . . . . . . . . . . . . . . . . . . . 78
2
4.9 Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5 Cooling system design and thermal analysis 86

5.1 Water jacket design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.2 Steady state thermal analysis at rated load . . . . . . . . . . . . . . . . . . . . . 90
5.3 Lumped parameter thermal network: transient thermal analysis . . . . . . . . . . 92
Conclusion 104
Appendix 106
Bibliography 110
3
Riassunto esteso
In questa tesi viene presentato uno studio di fattibilità sull’elettrificazione del powertrain di
un trattore agricolo da campo aperto. Il principale obiettivo è fornire utili risultati e validare
metodologie pratiche per la progettazione di motori elettrici per trattori ibridi. In particolare in
questo progetto sono stati dimensionati tre motori a magnete permanente superficiale a flusso
radiale raffreddati a liquido.
L’elettrificazione delle macchine agricole negli ultimi anni ha iniziato ad assumere maggior
rilievo nella letteratua tecnico-scientifica, spinta dal crescente interesse da parte delle aziende
operanti nel settore. Sebbene già da circa un decennio siano stati pubblicati studi di fattibilità
e analisi di scenario riguardo l’introduzione di trattori ibidi ed anche puramente elettrici, solo
recentemente si sta iniziando a valuatare seriamente la loro industrializzazione. La spinta
all’elettrificazione proviene in particolar modo dalle nuove legislazioni sulle emissioni. Infatti le
recenti normative emenate sia nell’Unione Europea (Stage V) sia negli Stati Uniti (Tier 4) im-
pongono stringenti limitazioni alle emissioni dei Non-Road Mobile Machinery (NRMM), catego-
ria nella quale rientrano tutte le macchine agricole. Tali normative constringono all’introduzione
di catalizzatori, filtri per il particolato e raccoglitori per gli oli esausti nei motori Diesel di
tutti i trattori agricoli di nuova immatricolazione. Questi componenti rendono i motori più
voluminosi e costosi. Quindi, parallelamente alla proposta di modelli di agricoltura a bassa
meccanizzazione, sta riscontrando un crescente interesse l’elettrificazione dei veicoli agricoli,
con il fine di seguire quanto si è già fatto e si sta facendo nell’industria automobilistica. Benchè
la linea di sviluppo delineta possa essere simile a quanto fatto in campo automotive, con una
prima introduzione di veicoli ibridi per poi passare progressivamente a powertrain totalmente
elettrici, è già stato messo in luce come l’elettrificazione dei trattori presenti una complessità
maggiore, a causa delle peculiarità delle macchine agricole che le distinguono largamente dai ve-
icoli stradali. Rispetto a questi ultimi infatti è stato evidenziato che gli azionamenti elettrici nei
trattori agricoli dovranno soddisfare requisiti diversi, poichè in questa applicazione la trazione
rappresenta spesso solo una parte della coppia di carico richiesta. Infatti molte operazioni in
agricoltura richiedono il traino di macchinari azionati dal motore del trattore tramite una presa
di forza meccanica esterna (PTO: Power Take Off ). Non sono poi trascurabili i carichi dovuti
alle pompe dell’olio dei circuiti idraulici del servosterzo, nonché di alcuni tipi di macchinari
che necessitano anche di una presa di forza idraulica. Inoltre nemmeno il carico di trazione è
paragonabile ai ben noti cicli di lavoro standardizzati usati in campo automobilistico, poiché
in agricoltura è frequente il traino di utensili (aratri, estirpatrici, ecc. . . ) che richiedono coppie
impulsive e molto oscillanti, come si potrà anche vedere nel corso di questa trattazione.
Durante questo lavoro di tesi è stata studiata l’elettrificazione del powertrain di un trattore
agricolo da campo aperto da 280 [Hp]. Lo scopo è dimensionare un motore elettrico per trasfor-
mare il powertrain tradizionale del trattore in un ibrido. Il progetto è focalizzato in particolar
modo sulla progettazione del motore elettrico, ma per la scelta delle specifiche di progetto è
stata necessaria un’analisi del veicolo ibrido ed una sua modelizzazione semplificata. Inoltre nel
corso del progetto si sono anche individuate alcune specifiche utili per la progettazione degli
altri componenti dell’azionamento, quali l’inverter ed il pacco batteria, nonstante questi aspetti
non siano stati approfonditi in dettaglio.
Durante il lavoro di tesi sono state utilizzate ed elaborate alcune misurazioni dei profili di
coppia e velocità all’albero motore effettuate sul trattore in esame durante quattro diverse op-
erazioni agricole. Le misurazioni sono state rese disponibli per questo lavoro di tesi nell’ambito
del progetto PRIN 2017 intitolato ‘Green SEED: Design of more-electric tractors for a more
4
sustainable agriculture’. Tali dati sono stati essenziali per ricavare verosimili specifiche di pro-
getto per la macchina elettrica e per dedurre o almeno ipotizzare le informazioni non disponibili
riguardo al veicolo. In particolare il motore Diesel originale è stato ipotizzato comparando la
caratteristica meccanica fornita nelle schede tecniche commerciali di alcuni motori ad uso agri-
colo con i punti di lavoro coppia-velocità e potenza-velocità estrapolati dalle misurazioni, per
vedere quale profilo si adattasse meglio ai cicli di lavoro. Una volta scelto il motore termico
del powertrain tradizionale, si sono selezionati anche tre motori di potenza via via inferiore,
per studiare l’elettrificazione a grado di ibridazione crescente, cioè con un motore Diesel sem-
pre più sottodimensionato ed una maggior richiesta di potenza alla macchina elettrica. Per il
powertrain ibrido si è scelta un’architettura parallelo, con il motore elettrico calettato diret-
tamente all’albero motore. Il motore elettrico e quello termico dunque lavorano sempre alla
stessa velocità, se si trascurano i transitori meccanici. In questo studio preliminare si è deciso
di adottare una semplice strategia di controllo (power management) a soglia, in cui è previsto
che il carico meccanico venga bilanciato interamente dal motore Diesel fino ad una certa soglia
prestabilita, oltre la quale interviene in aggiunta il motore elettrico per compensare i picchi di
potenza richiesta. In tale sistema di controllo non è contemplata la ricarica del pacco batteria
con veicolo acceso, in quanto non è previsto un eccesso di potenza da parte del motore termico
rispetto al carico.
La richiesta di coppia e potenza al motore elettrico è stata calcolata simulando il bilancio
di potenza all’albero motore attraverso un circuito meccanico. Le coppie di carico imposte nel
circuito sono state ottenute dalle misurazioni disponibili nei quattro cicli di lavoro, mentre la
coppia data dal motore termico è stata ricavata dalla caratteristica meccanica considerando
il power management adottato. I test di ibridazione sono stati effettuati considerando sia le
caratteristiche meccaniche di motori Diesel ad uso agricolo di potenza inferiore sia scalando la
caratteristica meccanica del motore termico originale. E’ stato provato che non c’è una dif-
ferenza significativa tra le due procedure, nonstante i motori scelti siano forniti da produttori
diversi. I profili di coppia e potenza richiesta alla macchina elettrica sono stati poi processati
con un filtro passa-basso provando diverse costanti di tempo. La costante di tempo del filtro
approssima la costante di tempo termica del motore. Le specifiche di coppia del motore elet-
trico infatti sono state scelte seguendo il metodo della coppia termica equivalente. Il concetto
alla base di tale approccio, ampiamente usato in campo automotive, consiste nello scegliere un
valore di coppia nominale inferiore alla massima coppia richiesta in modo che i picchi vengano
compensati in sovraccarico, senza però pregiudicare la tenuta termica della macchina, in par-
ticolare dell’isolante in cava e dei magneti. Infatti con coppie nominali più basse, a parità di
velocità, in genere si ottengono motori più compatti ed economici. Per identificare le specifiche
di progetto quindi la dinamica del motore deve essere inizialmente ipotizzata, in particolare si
è supposto un valore per la costante di tempo termica.
Tre motori elettrici a magnete permanente superficiale a flusso radiale sono stati dunque
dimensionati per uno dei tre gradi di ibridazione analizzati. I motori differiscono principalmente
nel numero di poli. Questa tipologia di motore è stata scelta per la sua elevata densità di potenza
e perchè è particolarmente adeguato in applicazioni a velocità relativamente basse, come quella
affrontata in questa tesi. La scelta del tipo di magnete come anche del numero di poli è stata
giustificata. Inoltre per ottimizzare lo spazio occupato e soddisfare i vincoli di volume, un
particolare dimensionamento preliminare è stato effettuato, tenendo conto sin da subito delle
testate e del totale ingombro assiale. Diversi tentativi di dimensinamento sono stati eseguiti,
variando il numero di cave, lo spessore ed il grado del magnete. La scelta delle dimensioni
principali delle tre macchine è stata effettuata favorendo la compattezza e l’economicità. Si è
proseguito dunque con il dimensionamento dell’avvolgimento, della lamiera di statore e di rotore.
Una stima del costo dei materiale è stata effettuata alla fine della sintesi elettromagnetica delle
tre macchine.
Le prestazioni dei motori sono state poi valuate con analisi agli elementi finiti e stime
analitiche. Si sono verificate le prestazioni a vuoto, calcolando la saturazione della macchina,
la forza elettromotrice e la coppia di impuntamento. Dunque si sono ricavate la coppia a carico
nominale ed in massimo sovraccarico con relativo ripple. Sono stati calcolati la resistenza
e l’induttanza sincrona e si è valutata di conseguenza la tensione richiesta dall’avvolgimento
a carico nominale ed in massimo sovraccarico, con il relativo fattore di potenza. Anche le
5
perdite e l’efficienza della macchina sono state stimate. In particolare nel calcolo delle perdite
nel ferro e nei magneti con gli elementi finiti si sono considerati anche i contributi dovuti
al contenuto armonico dell’induzione nonchè la presenza di una componente armonica nella
corrente di alimenatzione a frequenza di switching causata dalla PWM dell’inverter.
E’ stato poi dimensionato il sistema di raffreddamento. Si è scelta un camicia esterna in
acciaio inox con un condotto a spirale lungo la circonferenza esterna del pacco lamiere. Il liquido
refrigerante è una soluzione antigelo di acqua-glicole. La geometria della camicia ed il circuito
idraulico (il condotto di raffreddamento) sono stati progettati imponendo lo scambio termico
convettivo richiesto per smaltire le perdite precedentente calcolate ed attraverso delle stime
analitiche, avvalendosi dei numeri adimensionali che caraterizzano la fluidodinamica. Infine si
è analizzato il comportamento termico della macchina a regime sotto carico nominale ed in
transitorio durante i cicli di lavoro disponibili. La mappa termica a regime è stata ottenuta con
gli elementi finiti, mentre le simulazioni in transitorio sono state effettuate con una rete termica
equivalente a parametri concentrati.
Sebbene la costante di tempo termico effettiva risulti molto inferiore a quella supposta
inizialmente a causa dei volumi contenuti dei motori, le sovratemperature durante il ciclo di
lavoro più critico non superano il massimo valore ammesso dalla classe di isolamento scelta
e dalla classe termica del magnete. Ciò nonostante la soglia di sicurezza raccomandata dalla
normativa IEC è abbondantemente superata. Pertanto le prestazioni di coppia in sovraccarico
dei motori deveono essere limitate per evitare danni ai materiali isolanti e perdite permanenti
di isolamento elettrico. Tale calo delle prestazioni può essere accettato nei veicoli agricoli, nei
quali non sono fondamentali grandi accelerazioni. Per gli altri cicli di lavoro infatti non si
riscontrano problemi termici.
In questa tesi si è affrontato il dimensionamento e l’analisi di motori elettrici per un’applicazione
non ancora studiata approfonditamente in letteratura scientifica. Quindi, nonstante i risultati
raggiunti, il lavoro svolto si presta a molti possibili ulteriori sviluppi.
6
Introduction
In recent years the electrification of farming tractors has gained more relevance in the technical
research [1], thanks to an increasing interest from the companies involved in the market of
agricultural machinery [2]. Despite in the last decade some investigations have been proposed
in the technical literature about both hybrid electric tractors (HETs) [3] and pure electric
agricultural machinery (BETs: Battery Electric Tractors, FCETs: Fuel Cell Electric Tractors)
[4], only now the industrialization of such vehicles is beginning to be seriously considered.
This growing interest is mainly due to the more strictly emissions regulations issued recently
(2019-2020) both in the European Union (Stage V standard) [5] and in the United States
(Tier IV standard) [6], that prescribe very low particulate emissions for the so-called non-road
mobile machinery (NRMM), to which the farming tractors belong. The admissible threshold
of emissions for the new agricultural vehicles has been lowered a lot, thus to observe the new
standards the Diesel engines, which are installed on farming tractors almost worldwide, must
be equipped with particulate filters (SCR: Selective Catalytic Reducer, DPF: Diesel Particulate
Filter ) and tanks for the exhaust oils. Such components make the engines more complex, bulky
and expensive.
While the electric auxiliaries (fans, pumps, cooling systems,...) has been implemented in
NRMMs since long ago [7], the powertrain electrification is still in an initial stage if compared
to the on-road vehicles. Regarding this topic some progress have been done in case of wheel
loaders [8] and robotic tractors [9], but for farming vehicles only few proposals have been made
[3], [10], [11]. It has been underlined how the electric drives must fulfill different specifications
than the ones commonly adopted in on-road vehicles. Indeed in such application the traction is
only a limited part of the overall effort because the agricultural work often requires the pulling
of various tools that need to be powered from the engine through a mechanical power take-off
(PTO). Furthermore hydraulic loads such as power-steering pumps and lifting implements can-
not be neglected. Moreover neither the traction load is comparable to the standard duty cycles
commonly adopted in the automotive field, because operations such plowing and harrowing
require high-varying and pulse torque, as it will be also presented in this thesis.
Even though the research is at the beginnings, the development course that has just in-
terested electric cars and motorbikes could also be followed in this field [12]. At first the
development of hybrid powertrains can be planned: starting from mild-hybrid tractors, where
the electric motor will satisfy only the peak loads, full-hybrid powertrains will be gradually de-
veloped through a step-by-step downsizing of the Diesel engine. The electrical machine power
capability will increase until the load could be entirely fulfilled in pure electric mode at least
during light duty cycles. The electrification of auxiliaries and hydraulic circuits could be also
involved in this first stage of the development course. Drive-by-wire systems have been pro-
posed in place of the hydraulic power-steering [4]. The vehicle size and in particular the chassis
should remain unchanged as much as possible to make the industrialization easier. So the power
take off will be still mechanical.
For the hybrid powertrain the same architectures implemented in on-road vehicles have
been proposed: series, parallel and power-splitter [13]. Scenarios forecasting and economic
assessments have been presented in the technical literature [14] to evaluate which could be the
best choices for a hybrid electric tractor. Lately a project regarding the design and analysis of
a SPM motor (Surface Permanent Magnet) to electrify the powertrain of a specialized tractor
for orchards and vineyards has been proposed in [10], [15].
In a second development stage the electrification of the PTO and the related implements will
7
be carried out. Finally the feasibility of pure electric tractors will be assessed. Prototypes of
small-size electric agricultural machinery have been already manufactured for research purposes
[11] and investigations about the charging systems as well as the power electronics and the
battery packs have been just presented in literature [16].
The project presented in this thesis belongs to the context above described. The main goal is
the assessment of the thermal equivalent torque approach as a design method for SPM motors in
a hybrid electric farming tractor. The project follows what has been done by D. Troncon et al.
in [10], [15] for a specialized farming vehicle, but a 280[Hp] open field tractor is here considered.
A parallel architecture has been adopted for the hybrid powertrain: the electric motor has been
supposed directly installed on the engine shaft, as it is usually done in mild-hybrid vehicles [13].
The power-management is based on a simple threshold logic [13], so that the electrical machine
works only to boost the engine. An original Diesel engine have been chosen among commercial
datasheets looking at some available torque and speed measurements taken on an open field
farming tractor during four different agricultural operations. These data have been measured
in the frame of the PRIN 2017 project entitled ‘Green SEED: Design of more-electric tractors
®
for a more sustainable agriculture’. Then the available torque and speed trends have been
used as inputs in a Simulink mechanical circuit that model the power balance at the engine
shaft. Three downsized engines with decreasing rated power have been selected to simulate
growing hybridization factors. The torque required to the electric motor in each simulation has
been processed through a low pass filter with various time constants to estimate the equivalent
thermal torque. So the specifications needed to start the preliminary sizing of the electrical
machines have been chosen supposing a thermal time constant and looking at the resulting
torque trends. Then three SPM motors with a different number of poles have been designed.
Displacement constrains have been evaluated and a particular preliminary sizing method has
been implemented to cope with them. The machines main dimensions have been selected among
different sizing attempts preferring the compactness and the cost-effectiveness. To assess the
torque performances and evaluate the losses and efficiency a detailed electromagnetic analysis
®
have been carried out on the designed machines both with analytical computations and FEA
(Finite Element Analysis), using MATLAB and the open source software FEMM, developed
by David Meeker [17]. Finally the thermal behavior of the motors have been analyzed: the
steady state condition at rated load has been simulated with the thermal solver implemented
in FEMM (FEH: Finite Element Heating), while the step responses and the overtemperatures
®
trends during the four real duty cycles have been obtained by means of an equivalent lumped pa-
rameters thermal network implemented in Simulink . Hence the actual thermal time constant
has been computed from the step responses and the resulting slot and magnet overtemperture
trends have been compared to the expected ones.
The design of the other components of the electric drive, i.e. the inverter and the battery
pack, as well as a detailed sizing of the mechanical parts are not covered in this thesis. However
some information will be provided in order to give a starting point for further developments
on these topics. Indeed one of the aims of this work is to give valid results and suggestions
that could be useful in further electrification projects for farming tractors and agricultural
machinery.
8
Chapter 1
Powertrain electrification
The electrification of an open field farming tractor will be presented in this first chapter.
After a brief classification of electric hybrid vehicles examining the hybridization factor, the
original powertrain and a proposed hybrid architecture will be schematically described through
a systemic point of view. Thus the description will be focused on the tasks and the significant
features of the main components involved in power generation and transmission inside both
powertrains, so that the most relevant differences between the original architecture and the
electrified one are standing out. A very simple power management has been chosen for the
proposed hybrid tractor: its operating workflow will be explained in order to highlight how
it affects the performances of the vehicle and the technical requirements of the powertrain
components. Meanwhile, advantages and drawbacks of the previously mentioned choices will
be discussed.
1.1 Hybridization factor

An electric hybrid vehicle has two main mechanical power sources and two related energy
sources: an internal combustion engine (ICE) whose injection system is fed by a fuel tank and
an electrical machine (EM) driven by a power electronics unit and fed by a battery pack, or,
in rare cases, by a fuel cell with a hydrogen tank. The ICE and EM work together to satisfy
as best as possible all the mechanical loads of the vehicle. The cooperation between the ICE
and the EM is supervised by the power management system of the vehicle which prescribes
the working rules and dictates the input parameters for the driving controls of the engine and
the electric motor, i.e. speed or torque references. Battery charge is also managed through the
EM. A lot of different power management logics can be implemented, depending on the hybrid
powertrain architecture and the duty cycles of the vehicle [13, 18, 19].
An electric hybrid vehicle can be classified looking at the hybridization factor H (or DOH:
degree of hybridization [19, 20]) that is defined as the ratio between the rated power of the
electrical machine PEM and the total rated power of the powertrain, which is the sum of the
rated power of the ICE PICE with PEM :
PEM
H= (1.1)
PICE + PEM
The hybridization factor is an index expressing how much the powertrain is electrified. The
definition of H does not consider other electric motors that can be mounted on the vehicle
to power auxiliaries, such as pumps, air-cooling fans, electric servo-steering system, etc. H
concerns only the powertrain and takes into account only the electric motor that develops
torque for the main mechanical loads. It is relevant to notice that H = 0% means a traditional
non-electrified powertrain (PEM = 0) while H = 100% means a full electric vehicle (PICE = 0).
Relating to the value of H, electric hybrid vehicles can be classified as presented in Table 1.1.
Values of H in Table 1.1 stop at 60% as above this value an hybrid architecture has no longer
advantages but a full electric powertrain is much more worthwhile [13, 18]. The classification
9
H [%] Hybrid type
0−5 Light-hybrid
5 − 20 Mild-hybrid
20 − 60 Full-hybrid
Table 1.1: Electric hybrid vehicles classification
Auxiliaries
BAT
RECT
START GEN Hydrualic

P
implements
Traction
ICE C1 G
shaft
S
C2
PTO
R
shaft
FUEL
TANK
Figure 1.1: Original powertrain outline
in Table 1.1 is not overall accepted but the reported ranges of H vary slightly between different
literature papers.
1.2 Original powertrain

The powertrain of a common open-field farming tractor is outlined in Figure 1.1. It is as-
sumed to be the original powertrain under analysis as it is the most widespread architecture
for agricultural machinery.
The main power source is an internal combustion engine (ICE). Open-field tractors with
rated power higher than 100 − 120 [Hp] usually have a turbocharged Diesel engine with in-line
cylinders and electronic or hydraulic controlled common-rail injection. The ICE converts the
chemical energy stored in the fuel, i.e. Diesel oil, into rotational mechanical energy and makes
it available at the engine shaft.
The rotational energy can be directly converted into electric energy by a three-phase gen-
erator (GEN) that charges the battery pack (BAT) through a static conversion group (RECT)
when required by the vehicle’s ECU (Electronic Control Unit). The conversion group denoted
with RECT is usually composed by a three-phase diodes rectifier followed by a DC-DC voltage-
conditioning stage with a boost topology. In older tractors the generator could be even a
dynamo and in such case there is no need for the diodes rectifier. The ICE is unable to start by
itself: it cannot develop torque at null speed. So an electric motor, i.e. the starter (START),
has the task to develop a sufficient torque to put the crankshaft in rotation, so that the engine
pistons can do an adequate air compression inside the cylinders to ignite the first combustion
process. The starter is almost always a DC motor designed to develop an high starting torque
and being able to be fed safely by the battery. The generator and the starter are placed respec-
tively in the engine’s rear end and front end. Usually both of them are connected to the engine
shaft through speed reducers, i.e. torque multipliers, as generally electrical machines exploit
better high speed rates rather than high torque capabilities. In some new commercial tractors
the generator and the starter tasks are often performed by the same electrical machine named
ISA (Integrated Starter Alternator), resulting in a so-called light-hybrid powertrain architec-
ture. ISAs are driven by bidirectional power electronics units that are able to guarantee both
10
Charging system Auxiliaries
BAT L
CONV
E Traction
ICE D C3 C1 G
shaft
M
S
C2
PTO
R
shaft
FUEL
TANK
Figure 1.2: Hybrid powertrain outline
the battery discharging during the starting phases and the battery charging when required by
the ECU.
Commonly the battery is a lead-acid valve regulated accumulator with a high power density
in order to provide intensive discharges during the starts (currents peak values up to 1000 [A]),
but a lower energy density; so when the engine is working the accumulator can be continuously
recharged through the generator and a small capacity is sufficient to feed all the electrical
auxiliaries, i.e. fans, lights, control electronics, etc.
The mechanical load applied at the engine shaft are represented by the traction effort,
a mechanical power take off (PTO), some hydraulic implements and the already described
generator. Traction effort requires the transmission of mechanical energy to the driving wheels’
axles, that could be the rear wheels or all the four wheels in the 4x4 architectures. At the
same time mechanical energy should be available even for the PTO shaft when the agricultural
operations require the use of external implements such as pumps, power harrows, atomizers
etc. The ICEs’ are not suitable for traction purpose because they have a low torque capability,
especially at low speed rates. So, as it happens for on-road vehicles, the engine’s mechanical
characteristics must be adapted to traction’s power requirements with the help of a multi-gear
transmission system (G) connected to the engine shaft through a clutch (C1). Farming tractors’
transmission systems are much more complex than those commonly implemented in on-road
vehicles due to the great diversity of agricultural operations. Furthermore, the mechanical
PTO often is involved in the work together with traction. The analysis of the mechanical
transmissions of farming tractors is not in the aims of this thesis. A lot of solutions can be
implemented but, as far as this study concerns, only a double clutch (C2) with a speed reducer
(SR) on the PTO shaft’s side is considered as mechanical connection between the power source,
i.e. the ICE, and the mechanical loads, i.e. traction and PTO. Hydraulic-powered implements
such as rear or front lifts should be also taken into account: these loads require hydraulic circuits
where the fluid, i.e. oil, is elaborated by pumps (P) that are powered by the engine.
During the last decade a lot of new technologies have been implemented in the field of agri-
cultural machinery. More efficient and environmentally friendly Diesel engines with new control
and management systems have been developed [21, 22], following the increasingly limiting Eu-
ropean and American regulations on non-road vehicles’ emissions. Self-driving tractors along
with more safe and comfortable machinery are now available on the market [23]. Nevertheless
the powertrain architecture outlined in Figure 1.1 has not gone through substantial changes,
despite the variety of solutions proposed in recent years in the technical literature [4].
1.3 Hybrid powertrain

The hybrid powertrain proposed in this study is outlined in Figure 1.2.
11
The original ICE is replaced by a downsized Diesel engine (ICE D) of the same type, i.e.
turbocharged in-line cylinders with common-rail injection. To cover the power capability lost
due to the ICE downsizing an electric motor (EM) is included in the powetrain. The proper
design of this electric machine is the main purpose of this thesis. The EM shaft is connected to
the engine shaft through the clutch (C3). Thus the ICE and the EM are mechanically in parallel
because the torque developed by ICE τICE and the torque developed by the EM τEM sum up
to balance the load torque τL at the shaft. So it can be stated that this hybrid powertrain
has a parallel architecture. In the parallel architectures the ICE and the EM are mechanically
connected through a torque adder [13, 18], a mechanical transmission device that, neglecting
any dissipation due to frictions, does a weighted sum of the input torques τICE and τEM to
balance the load torque τL imposed by the vehicle duty cycle:
τL = gICE τICE + gEM τEM (1.2)
Where gICE and gEM are dimensionless weights that depends on how the torque adder is
physically made. The device working principle must fulfill the instantaneous mechanical power
balance:
pL = pICE + pEM
Where pL , pICE and pEM are respectively the load, engine and electric motor instantaneous
power. Remembering that the mechanical power p [W ] can be expressed as the product of
instantaneous torque τ [N m] by instantaneous rotational speed ω [rad/s], the power balance
can be written as follows:
τL ωL = τICE ωICE + τEM ωEM
Consequently, substituting τL of Equation 1.2 in the previous expression yields to the following
relations, where it is stated that both the rotational speed of the ICE ωICE and the EM ωEM
are binded to the speed required by the driver ωL :
(
ωICE = gICE ωL
(1.3)
ωEM = gEM ωL
In full-hybrid vehicles with parallel architectures, the torque adder is usually realized with a
set of three cogged gears as represented in the schematic drawing of Figure 1.3 [13]. In this
implementation the weights gICE and gEM depend on the ratio between the gears diameters.
So during the powertrain design the most suitable values can be chosen to fit as best as possible
the features of the engine and the EM. Generally this solution leads to an easier design of the
EM because the machine can be sized for a high rated speed and a low rated torque, with a
benefit in efficiency too. Nevertheless, an easier implementation of the torque adder has been
adopted in this project, which is quiet common in mild-hybrid vehicles. The device consists of
a single shaft shared between the ICE and the EM, except for the clutch C3. In this case the
coefficients gICE and gEM are both equal to 1 and Equations 1.2-1.3 becomes respectively:
τL = τICE + τEM
ωL = ωICE = ωEM
The EM and the ICE operate at the same speed, which is imposed by the driver, while the power
management system establishes how to split the torque contribution to balance the overall load
requirement. The clucth C3 is normally closed during ICE and EM cooperation (hybrid mode)
and also when only the ICE develops torque. C3 opens just in the case of full electric mode,
when only the EM develops torque. Thus, full electric mode is an optional in this architecture
and, if there are no interest in using it, C3 could be omitted.
The solution presented above has been chosen because it minimizes the transformation
costs of a traditional tractor into an hybrid one, due to the minimal changes required for the
mechanical structure of the vehicle. Indeed a commercial downsized Diesel engine could be
selected and, in the volume made available by the reduction of the engine displacement, a
proper designed EM and a clutch could be installed. In such a way, there is no need for further
12
τICEωICE
RICE
RL
ωL τL
τEM ωEM
REM
Figure 1.3: Torque adder three-gears implementation
bulky mechanical gears, that would also increase the friction losses. Besides the chassis does
not require any variation. In a tractor the chassis is the most expensive piece to change as it
is often manufactured by welding together cast iron bars. So any change in the chassis size
and shape should be avoided during a hybridization project in order to reduce manufacturing
costs. Nevertheless, this choice has some drawbacks, in particular regarding the EM design and
features. Indeed the EM must be sized for high rated torque and low rated speed, resulting in a
more complex design and a less efficient machine [24]. Furthermore, to fit in the volume cleared
by the engine displacement’s reduction, the electrical machine must have a low aspect ratio, i.e.
an high stator inner diameter compared to the axial stack length, a feature that in radial flux
machines highly compromises the efficiency. Besides the EM must have a great torque density
so that liquid cooling systems are required. Moreover, the EM should have a low inertia and
should produce low unload losses because, as explained before, it is mechanically connected to
the engine even when it is not developing torque, thus it represents quite always a rotating
mechanical load for the ICE.
The starter and the generator are no longer needed, as their tasks can now be performed
by the EM itself. One static conversion group (CONV) drives the EM. This power electronics
unit could be a unidirectional six-leg inverter if the EM behaves only as motor, or it could be a
bidirectional active rectifier with a DC-DC voltage conditioning stage to allow both the charging
and discharging of the battery pack (BAT L) when the electrical machine works respectively
as a generator or as a motor. In the former case the battery pack could be recharged only
through an external electric socket and a charging system, while in the latter case the charging
is possible even when the vehicle is running.
The battery pack is a lithium-ion accumulator. Such battery type has a power density
comparable to lead-acid accumulators but with a greater energy density. Indeed the battery in
an electric hybrid powertrain must be designed to feed the electric motors with high current
values (rms values up to 250 [A]) and to provide a proper energy storage to power all the other
electric loads. The design of the battery pack and the power electronics unit is not a goal
of this study, but some information and recommendations which could be useful to size such
components will be provided during the presentation.
The hydraulic implements of the tractor are assumed to be electrified with separate electric
drives. So the oil pumps are considerably powered by other electric motors through distinct
static conversion groups. It has been proved that in such a way the overall efficiency of the
vehicle is improved [7]. Thus only the mechanical loads, i.e. traction and PTO, are contemplated
in the following analysis.
13
1.4 Power management system
A large variety of power management systems have been proposed in the technical literature for
hybrid parallel architecture powertrains and many of them have been successfully implemented
in commercial on-road vehicles. But for farming tractors the research about this topic is still
at an initial stage and only few industrial solutions have been investigated [25]. Therefore, as
the focus of this thesis is on the electrical machine design, the most simple power management
has been chosen for the hybrid powertrain under analysis.
As the speed of both the ICE and the EM is imposed by the driver (and by the multi-gear
transmission), the power management system has only the task to establish instantaneously
how the torque contribution is split between the ICE and the EM to balance the mechanical
loads when the vehicle operates in hybrid mode. The chosen system is based on a threshold
logic. The load torque is fully provided by the ICE until a limit threshold τlim is reached. If
the load torque exceeds the value of τlim the EM is activated to develop the exceeding torque.
In case of negative load torque, i.e. by braking or driving downhill, the EM behaves as an
electromechanical brake, namely an electric generator, and it charges the battery recovering the
braking mechanical energy. The above described system workflow is mathematically represented
by the following expressions:
( 0 τL ≤ 0 ( τ
L τL ≤ 0
τICE = τL 0 < τL ≤ τlim τEM = 0 0 < τL ≤ τlim (1.4)
τlim τL > τlim τL − τlim τL > τlim
It is important to specify that τlim is not a fixed constant, but it depends on the speed
instantaneous value ωICE = ωL . The threshold τlim is actually a torque versus speed curve
obtained by a chosen downscaling of the ICE rated mechanical characteristic. This limiting
curve can be calculated searching the torque-speed working points with the highest efficiency
for every speed value in the range of the feasible values. Indeed in a parallel architecture
powertrain the ICE cannot always operate at its maximum efficiency working point, but the
power management system can exploit the highest efficiency regions for every speed value [13].
In agricultural Diesel engines the highest efficiency curve fit well the 75% of the rated mechanical
characteristic, as it will be proved in the next sections of this chapter. Thus a downscaling factor
of 0.75 was applied to the rated mechanical characteristic of the selected downsized ICEs for
the hybrid powertrain in order to obtain the τlim threshold curve. It is relevant to underline
that with such a power management logic the ICE’s highest efficiency working points are not
fully exploited because the engine does not always work near the τlim curve; yet in order to
follow lower load torque requirements, it can operate far below the threshold curve.
Looking at the electrical machine behavior, it has been demonstrated that for agricultural
duty cycles the braking energy recovering is negligible [26]. Then the electric motor works only
as torque boost for the engine.
Moreover the chosen power management does not allow the battery charging through an
ICE exceeding torque development. Thus the battery should have a very important capacity
to feed properly the electrical systems of the vehicle during a whole working day, as it can be
recharged only by an external socket. So the volume needed to place the battery pack becomes
a very relevant constrain in the design of this hybrid powertrain.
The above statements will find further explanations in the next sections of this chapter
with the help of graphs and charts of the chosen engines’ mechanical characteristics and the
processed data.
14
Chapter 2
Case study: open field farming

tractor
The case study that concerns this thesis will be presented in this chapter.
Some available data about four different duty cycles of an open field tractor have been
analyzed in detail to obtain all the needed information to carry on the hybridization process of
®
the original powertrain, or at least to make reliable assumptions. Then the available duty cycles
data have been processed with a Simulink one-dimensional mechanical model of the vehicle to
identify the torque and speed specifications of the electrical machine for three different choices
of downsized engines in the hybrid powertrain. Data analysis, the vehicle circuital model and
the method used to choose the electric motor torque and speed requirements are all described
in detail and the related results are discussed.
2.1 Data analysis

Some preliminary measurements on an open field farming tractor working in four different
conditions are available. They have been developed in the frame of the PRIN 2017 project
entitled ‘Green SEED: Design of more-electric tractors for a more sustainable agriculture’.
Information not yet available about the tractor powertrain and regarding the data have been
supposed.
The measurements have been performed by means of specifically-implemented transduc-
ers and digital signal processor units (DSPUs). Different quantities have been measured and
sampled over time during the following four operations in open field tests.
Plough 60: the tractor dragged a plow on the ground to make furrows with depth of
about 60 cm.
Plough 100: the tractor dragged a plow on the ground to make furrows with depth of
about 100 cm.
Power harrow: the tractor drove a power harrow through its power take off (PTO) shaft
and pulled it in the field.
Transportation: the tractor carried an unknown load in the field.
The available data for each test are here reported and analyzed. Relevant details about some
of them deserve a significant attention related to the next data elaboration in order to find out
the right inputs for the mechanical circuit model of the powertrain.
2.1.1 Plowing tests measurements

For the two plowing tests the following measures are available:
15
engine torque [N m]; torque developed by the ICE measured at engine shaft;
engine speed [rpm]; engine shaft rotational speed;
vehicle speed [km/h]; tractor linear velocity.
The plowing draft [N ]was measured too, probably with a system made up by a force trans-
ducer installed on the plough implement used for this specific purpose. The measured force is
the component of the ground resistance acting on the plow parallel to the tractor motion.
Engine parameters (torque and rotational speed at the shaft) have been sampled with the
same frequency, as the same number of samples are available for both quantities. The vehicle
speed and the draft force have been sampled with different time steps compared to the one used
for the engine torque and the engine speed, maybe because the vehicle speed was sampled by
the internal ECU of the tractor, while the draft force was sampled by a different DSPU than
the one used to process the engine quantities. As available data result from fewer vehicle speed
samples and more draft samples, it is feasible that the vehicle speed was sampled with a bigger
time step while the draft force was sampled by a smaller one compared to engine torque and
engine speed data.
Time samples are given in the same number of engine quantities samples, so vehicle speed
and draft force data are referred to engine measurements time step through a linear interpolation
in order to make comparable plots of all data. Therefore the time reference of the plots is based
on the given time samples.
All the available data about the two plowing tests are plotted in Figure 2.1-2.2. The in-
stantaneous power developed by the engine during the tests pICE has been computed as the
product of the measured torque τICE [N m] by the instantaneous engine shaft speed converted
to [rad/s] ωICE [rad/s]:
pICE = τICE · ωICE (2.1)
2.1.2 Power harrow test measurements

At this working stage both traction and PTO have been involved.
engine torque [N m]; torque developed by ICE measured at engine shaft;
PTO torque [N m]; torque measured at PTO shaft;
PTO speed [rpm]; PTO shaft rotational speed;
Engine’s quantities (torque and rotational speed at the shaft) and PTO quantities have been
sampled with the same frequency, as the same number of samples are available for these four
data sets. Instead the vehicle’s speed has been sampled with a smaller frequency (i.e. a
bigger time step), probably because it was processed by the internal ECU, while for the other
measurements an external DSPU has been used. Time samples are not given for this test, thus
their overall number is assumed equal to the one of the engine and PTO quantities and the
same time step from the plowing tests is supposed. Therefore the vehicle’s speed samples are
referred to the time step using a linear interpolation.
All the available data are plotted in Figure 2.3. The instantaneous power developed by the
engine is computed as in Equation 2.1 while the instantaneous power required by the PTO
pP T O is calculated as:
pP T O = τP T O · ωP T O
Where τP T O [N m] is the instantaneous torque required at the PTO shaft and ωP T O is the
instantaneous PTO shaft speed in [rad/s].
Looking at Figure 2.3, it is very important to notice that engine shaft and PTO shaft have
different rotational speeds. PTO shaft speed is about half of the engine shaft speed. This
16
Test: Plough 60 cm
Torque request
800
Engine torque
Torque [Nm]
600
400
200
0
0 20 40 60 80 100 120
Engine shaft speed

2000
Speed [rpm]
1800
1600
1400
1200
1000
0 20 40 60 80 100 120
Power request
150
Engine power
Power [kW]
100
50
0
0 20 40 60 80 100 120
Vehicle speed
1
Speed [km/h]
0.8
0.6
0.4
0.2
0
0 20 40 60 80 100 120
Draft
0
-1
Draft [kN]
-2
-3
-4
-5
-6
0 20 40 60 80 100 120
Time [s]
Figure 2.1: Available data from plough 60 test.
17
Test: Plough 100 cm
Torque request
1000
Engine torque
800
Torque [Nm]
600
400
200
0
0 20 40 60 80 100 120 0 20 40 60
Engine shaft speed

2000
Speed [rpm]
1800
1600
1400
1200
0 20 40 60 80 100 120 0 20 40 60
Power request
200
Engine power
Power [kW]
150
100
50
0
0 20 40 60 80 100 120 0 20 40 60
Vehicle speed
2
Speed [km/h]
1.5
0.5
0
0 20 40 60 80 100 120 0 20 40 60
Draft
2
0
Draft [kN]
-2
-4
-6
0 20 40 60 80 100 120 0 20 40 60
Time [s]
Figure 2.2: Available data from plough 100 test.
18
means that during the test the engine shaft and the PTO shaft have been connected through a
speed reducer with a fixed gear ratio, as previously outlined in Figure 1.1. This fixed gear
ratio G can be estimated as the ratio between the mean over time of the engine shaft speed
Ω̄ICE [rad/s] and the mean over time of the PTO shaft speed Ω̄P T O [rad/s]:
Ω̄EN G ∼
G= = 1.89 (2.2)
Ω̄P T O
Where Ω̄ICE and Ω̄P T O are evaluated in discrete form as the sum of the samples values Ωk
divided by the number of the considered samples, instead of the continuous integral average
over time: R t1 Pn1
ωICE dt k=n0 ΩICEk
Ω̄ICE = t0 ∼
=
t1 − t0 n1 − n0
R t1 Pn1
ωP T O dt k=n0 ΩP T Ok
Ω̄P T O = t0 ∼
=
t1 − t0 n1 − n0
In the computation of the mean values only halfway samples are considered, i.e. from n0 at
time t0 to n1 at time t1 , in order to select a time interval which in the speed trend does not
change too much.
The identification of this gear ratio G is very relevant to the next data processing: later it
will be used to find out the right inputs for the circuital simulator of the hybrid powertrain
outlined in Figure 1.2.
2.1.3 Transportation test measurements

All the available measures for this test have been taken from the same DSPU. They are the
following:
engine torque [N m]; torque developed by ICE measured at engine shaft;
All the three quantities have been sampled with the same frequency and the reference time
samples are given. All the available data are plotted in Figure 2.4. The instantaneous power
developed by the engine is computed as in Equation 2.1.
2.2 Mechanical circuit model and data pre-processing

Since the goal of a hybridization process is to identify the speed and torque requirements for
® ®
the electrical machine design, a mechanical equivalent circuital model of the hybrid powertrain
outlined in Figure 1.2 was developed in Simulink SimScape to achieve this purpose[10].
The mechanical circuit implemented in the simulator is shown in Figure 2.5.
The engine is represented as a torque generator which imposes a torque value on the shaft
that corresponds to the maximum one available at a given rotational speed, according to the
mechanical characteristic of the chosen engine. Thus the ICE can be seen as a speed-regulated
torque generator. The engine rotational speed and its rated mechanical characteristic are the
first inputs for the model. The torque developed by the ICE can be decreased proportionally
with a reduction factor R (0 ≤ R ≤ 1). This operation consists of a downscale of the mechanical
characteristic of the chosen engine and it was implemented for two purposes: first of all, setting
the reduction coefficient R is an easy method to pass a torque threshold curve τlim to the
power management system; secondly, R could be used to simulate a large range of increasing
hybridization grade, from the traditional powertrain (R = 1, H = 0%) to a full electric one
(R = 0, H = 100%), without using the rated mechanical characteristics of the downsized ICEs.
This topic will be deeply discussed in the next sections.
19
Test: Power Harrow
Torque request
2000
PTO torque
1500
Engine torque
Torque [Nm]
1000
500
-500
0 500 1000 1500 2000 2500 3000
Engine and PTO shafts speeds

2500
2000
Engine shaft speed
PTO shaft speed
Speed [rpm]
1500
1000
500
-500
0 500 1000 1500 2000 2500 3000
Power request
250
200
PTO power
Engine power
Power [kW]
150
100
50
-50
0 500 1000 1500 2000 2500 3000
Vehicle speed
1
0.8
Speed [km/h]
0.6
0.4
0.2
-0.2
0 500 1000 1500 2000 2500 3000
Time [s]
Figure 2.3: Available data from power harrow test.
20
Test: Transportation
Torque request
1200
1000
Engine torque
Torque [Nm]
800
600
400
200
0
0 500 1000 1500 2000
Engine shaft speed

2000
1800
1600
Speed [rpm]
1400
1200
1000
800
600
0 500 1000 1500 2000
Power request
250
Engine power
200
Power [kW]
150
100
50
0
0 500 1000 1500 2000
Vehicle speed
3
2.5
Speed [km/h]
1.5
0.5
0
0 500 1000 1500 2000
Time [s]
Figure 2.4: Available data from transportation test.
21
τT0 RAC
ωSHAF T
+
−
EM
+ +
ICE τICE τP0 T O
− −
Figure 2.5: Mechanical circuital model.
The load torque, i.e. traction and PTO, is also represented with torque sources, but they
are not regulated by the rotational speed. Indeed they impose on the engine shaft a given
torque regardless of the rotational speed of the shaft. So the traction torque τT0 RAC and the
PTO torque τP0 T O are inputs for the circuital model. It is important to underline that these
inputs must be referred to the engine shaft. In such a circuit all the quantities are referred to
the engine shaft. In this way the auxiliaries energy consumption as well as the friction losses
in the multi-gear transmission system are included in the model without the need of specific
data about this power contributions. The available data do not report the traction torque, and
the PTO torque is not referred to the engine shaft in power harrow test. So for each test the
traction torque used in the simulations is computed with the following torque balance, where
any friction in the shaft bearings and in the speed reducer is neglected:
τP T O
τT0 RAC = τICE −
G
The PTO torque referred to the engine shaft is obtained as:
τP T O
τP0 T O =
G
Where G is the gear ratio of the speed reducer computed in Equation 2.2. From the previous
equations it can be seen that the traction torque requested at the engine shaft τT0 RAC is equal
to the engine torque τEN G , except for the power harrow test, where the PTO is involved too.
The EM is simulated with a speed generator which imposes the engine speed ωSHAF T on the
shaft. The ICE and the EM are installed on the same single shaft (except for the clutch C3)
so they are bounded to the same rotational speed. All the components in the circuit have the
same rotational speed ωSHAF T , which is an input for the model, so the speed requirements for
the EM are automatically dictated by the duty cycles of the tractor.
The instantaneous torque on the EM is the only unknown in the circuit: it is computed by
the simulator starting from the following mechanical power balance on the shaft:
pICE + pEM + pT RAC + pP T O = 0
A mechanical power is calculated as the product of torque by rotational speed. So the power
balance can be rewritten as:
τICE · ωICE + τEM · ωEM + τT0 RAC · ωT0 RAC + τP0 T O · ωP0 T O = 0
Since all the rotational speed are equal as all the quantities are referred to the engine shaft, the
previous equation becomes:
τICE + τEM + τT0 RAC + τP0 T O = 0
22
And finally the EM torque is expressed as:
τEM = −τICE − τT0 RAC − τP0 T O
The load torques referred to the engine shaft τT0 RAC and τP0 T O are processed by the simulator as
negative quantities while the ICE torque τICE is processed as a positive quantity. Under these
statements a torque balance equivalent to the previous one can be obtained with the application
of Kirchhoff law at the only mesh in the circuit of Figure 2.5. Hence the Kirchhoff ’torque’ law
returns:
τEM = τT0 RAC + τP0 T O − τICE
Where both the load torque and the engine torque in this case are considered as positive
quantities.
The resulting EM torque is positive when the EM works as motor to boost the ICE and it
is negative when the engine torque is greater than the load torque, hence the EM behaves as
an additional load developing a braking torque and possibly working as an electric generator.
The latter case is not take into account in this study due to the chosen power management
system, whose logic, represented by Equation 1.4, does not allow the battery charging through
an exceeding engine power production. Thus only the positive torque values of the EM will be
considered to find out a sizing torque specification.
2.3 Data post-processing: thermal equivalent torque

The EM boosting torque signals have been post-processed according to the equivalent thermal
torque method in order to obtain the needed information that can be used to choose the torque
specifications for the electric motor design. This method is commonly used to size electrical
machines that have to withstand high changing duty cycles, for example it is applied to identify
the torque requirements of traction motors for electric vehicles. The base concept is to design
the EM not on the maximum torque value reached in the duty cycle, but to choose a smaller
rated performance and fulfill the peak values exploiting a proper overload capability. Indeed it
is not convenient to size the machine on the maximum torque because in a rapidly changing
duty cycle the peak values are reached only for small time intervals, if compared to the total
duration. So an electric motor designed in such a way will result greatly oversized and it will
not be suitable for applications where a small displacement is required, as the one that concerns
this project. Instead a machine sized on a smaller rated torque will be more worthwhile in terms
of volume and weight reduction.
The selected rated torque must be a halfway value that allows the EM to withstand the peaks
during the entire duty cycle avoiding excessive in-slot overtemperatures ϑ that can damage the
insulator materials. Moreover for permanent magnets motors a particular care must be taken
for the magnets overtemperatures too. To identify such a rated specification the equivalent
thermal torque can be computed. The computation is performed using a low pass filter applied
on the EM boosting torque signals, in order to approximate the average integral in the root-
mean-square definition of the thermal equivalent torque, as reported in [10, 15], so the thermal
dynamics of the machine must be supposed. Indeed the equivalent thermal torque can be defined
as the torque trend that produces the same total Joule losses PJtot of the actual duty cycle.
The thermal behavior of an electric motor can be summarized by the simplified RC equivalent
thermal circuit shown in Figure 2.6, characterized by its thermal time constant τth [s]:
τth = Rtheq Ctheq (2.3)
Rtheq [◦ K/W ] and Ctheq [J/◦ K] are respectively the thermal equivalent resistance and the ther-
mal equivalent capacity of the machine. They depend upon the thermal properties of the
material used to manufacture the machine and its volume and weight.
τth is the time constant of the low pass filter used to process the EM boosting torque.
Therefore some feasible values of τth must be supposed. The electrical machine type chosen for
this project is a radial flux liquid cooled surface permanent magnets (SPM) motor. For such a
23
+
PJtot Rth Cth ϑ
Figure 2.6: Simplified equivalent thermal circuit of an EM.
kind of machines the thermal time constant τth spans in the range 200 ÷ 1000 [s], regarding to
rated torque specifications in the range 100 ÷ 400 [N m]. A detailed mathematical explanation
of the equivalent thermal torque is reported by D. Troncon et al. in [15].
Choosing a suitable time constant is a thorny point, as the selected value should be verified
at the end of the machine design through a sufficiently detailed transient thermal analysis. The
choice is not trivial. With the same duty cycle, low values of τth lead to the choice of higher
rated specifications and lower overload capabilities, while high values of τth lead to lower rated
torque requirements and higher overload capabilities. But the thermal dynamic behavior of an
EM depends strongly on its thermal equivalent capacity, as expressed in Equation 2.3, rather
than on its equivalent thermal resistance, that should be minimized to limit the steady state
overtemperature. High values of Cth imply high weights. On the contrary, low τth are typical of
smaller machine with a lower rated torque, while high τth are characteristic of bigger machines
with higher rated specifications.
The above statements will be clarified in the next sections, when the hybridization tests’
results will be presented and the torque specifications for the EM design will be definitively
chosen. Further explanations on the thermal behavior of electric motors will be provided in
Chapter 5.
2.4 Engines selection

Since the model described in Section 2.2 requires at least one engine mechanical characteristic,
the ICE mounted on the original powertrain must be feasibly supposed from the available duty
cycles data .
Looking at the power and torque trends in all four working conditions reported in Figures
2.1-2.2-2.3-2.4 it can be observed that the transportation duty cycle is the most critical one,
when the engine worked at the top of its performance during the first starting and acceleration
phase. From Figure2.4 it could be deduced that the original ICE should be able to supply a
mechanical power a bit greater than 200 [kW ] and a torque of about 1100 [N m] at a speed a
bit lower than 2000 [rpm]. Moreover, always looking at Figure 2.4, after the first acceleration it
could be noticed that the tractor stopped for a long time, i.e. 2000 [s] ∼
= 33 [min]. During this
period the engine was kept on and it was working at idle speed, which is lower than 800 [rpm],
providing the torque only to power the auxiliaries, since no traction nor PTO torque was
required. The original ICE worked at its maximum speed during the power harrow operation.
Looking at Figure 2.3 the deduced maximum speed lies in the range 2200 ÷ 2300 [rpm].
Therefore different agricultural Diesel engines, whose declared technical specifications are
close to the values identified above, have been selected from commercial datasheets provided
by different companies [27, 28, 29, 30]. All the engines are manufactured under the Stage III
EU standard or under the corresponding Tier III US regulation. Their mechanical character-
istics (rated torque and rated power curves) are plotted together with the experimental points
in all four working conditions to see which one fits the working points better. The overload
capability of turbocharged Diesel engines has not been considered, so that the rated mechanical
characteristic corresponds to the maximum torque and maximum power versus speed curves.
These curves have been obtained by inspection from the datasheets of the selected engines.
24
Cummins QSM280 210kW
Mechanical characteristic
1800 300
1650
1500 250
1350
1200 200
Torque [Nm]
1050
Power [kW]
900 150
750
600 100
450
300 50
Max. torque curve
150
Max. power curve
0 0
800 1000 1200 1400 1600 1800 2000 2200
Speed [rpm]
Figure 2.7: Original ICE mechanical characteristic.
Instead the experimental points, since they are the working points of the original ICE during
the four different operations, have been extrapolated from the available duty cycles data (see
Section 2.1) through some plots of engine torque versus speed. The different mechanical char-
acteristics have been compared with the experimental points in order to determine a feasible
original ICE. All the experimental points in each working condition must be possible working
points for the engine, so they must be under the curves. Furthermore, there is no benefit in
choosing an extremely powerful engine because it will not be useful to the later downsizing in
the hybridization process and it is not feasible that the original engine were too much oversized.
Hence, under such considerations, the selected ICE that fits better the experimental points
is the Cummins QSM280 [27], a turbocharged in-line six-cylinder Diesel engine with a declared
rated power of 210 [kW ] (280 [Hp]), whose mechanical characteristic is reported in Figure 2.7.
Figure 2.8 shows the comparison between the candidate ICEs mechanical characteristics and
the working points during the transportation test, which is the most critical duty cycle in terms
of torque and power requirements and indeed it is the test that allowed the selection of the
original ICE.
Once a feasible original engine has been chosen, three downsized ICEs with a decreasing
rated power have been selected for the hybrid powertrain to perform the hybridization tests. The
downsized ICEs are all made by FPT Industrial under Stage IV European standard (Tier IV US
regulation) [30]. The goal is to achieve Stage V emission requirements without selective catalytic
reducer (SCR) and Diesel particulate filters (DPF), thanks to the lower fuel consumption of
the electric hybrid powertrain. A different manufacturer was chosen than the one which supply
the supposed original ICE, i.e. Cummins, but for simplicity all downsized engines belong to
the same series, namely FPT N67ENT.
2.5 Hybridization tests

Two hybridization processes have been carried out. The first one is performed through a pro-
gressive downscaling of the supposed original ICE mechanical characteristic (Cummins QSM280
210 [kW ]) by decreasing gradually the reduction coefficient R (see Section 2.2). Simulations
25
Test: Transportation
1400 250
1200
200
1000
Torque [Nm]
150
Power [kW]
800
600
100
400
FPT C87ENT 230kW FPT C87ENT 230kW
Cummins QSM280 210kW 50 Cummins QSM280 210kW
200 Perkins 1106D-E70TA 205kW Perkins 1106D-E70TA 205kW
Deutz TCD L62V 200kW Deutz TCD L62V 200kW
Sperimental points Sperimental points
0 0
800 1000 1200 1400 1600 1800 2000 2200 800 1000 1200 1400 1600 1800 2000 2200
Speed [rpm] Speed [rpm]
Figure 2.8: Original ICE research. Comparison against experimental points in transportation
working condition.
have been done with the following values of R.
R = 1: ICE develops the maximum torque available at any rotational speed; the engine
works on the original mechanical characteristic; the hybridization grade is H = 0 % and
the model represents the traditional vehicle powertrain.
R = 0.75: ICE develops the 75 % of the maximum torque available at any rotational
speed; the engine works on a mechanical characteristic reduced by 25 % .
speed; the engine works at half of its maximum performances.
speed; the engine works on a mechanical characteristic reduced by 75 %.
R = 0.00: ICE does not develop torque at any rotational speed; the hybridization grade
is H = 100 % and the vehicle works in pure electric mode; the required power is entirely
given by the EM.
Differently, the second hybridization process has been performed using the mechanical charac-
teristics of the three downsized ICEs previously selected. The mechanical characteristics of the
three different engines have been obtained by inspecting their commercial datasheets [30]. It is
relevant to highlight that the simulations with the downsized ICEs have been done by setting
the engines at the 75% of their maximum torque and power performances, i.e. with R = 0.75,
in order to simulate a threshold curve τlim that permits to exploit highest efficiency regions.
The above statement has been proved by comparison of the actual highest efficiency curve with
the mechanical characteristic of a selected engine reduced by 25%. The results are summarized
in Figure 2.9, where it can be seen that the curve with R = 0.75 stays almost always above the
real highest efficiency curve. The curves are plotted over the BSFC map (Brake Specific Fuel
Consumption) of the engine FPT N67ENT 175 [kW ], one among the three downsized ICEs.
The efficiency map has been computed with an interpolating function that fits the parameters
of a large variety of agricultural Diesel engines. Its values are reported in terms of specific fuel
consumption [g/kW h], thus the actual highest efficiency curve has been computed by searching
the working point with the minimum BSFC for every speed value.
The two hybridization processes are represented and compared in Figure 2.10. The graphs
show that there is not a great distance in terms of mechanical characteristics between the
downsized ICEs and the corresponding torque and power reductions of the original engine
through the R coefficient, even if the original ICE and the downsized ones are made by different
26
FPT Industrial N67 ENT 175 kW (238 Hp) BSFC map [g/kWh]
1000
0 Max. torque curve
0 5 21
22 21 5 0 R=0.75
900
20 20 19Highest
5 efficiency curve
800
210
190
700
200
Torque [Nm]
600
20
200
195
200 195
500 205
400 205
210
210
215 20
300 215 2
220
195
225
200 230
235 0
24 45
5 2 250
255
21 0
100 22 230 5
205
21
5
23 40
800 1000 1200 2 1400 1600 1800 2000 2200
Speed [rpm]
Figure 2.9: Comparison between the threshold curve with R = 0.75 and the actual highest
efficiency curve over the BSFC map of the engine FPT N67ENT 175 [kW ]
manufacturers. As an example, always looking at Figure 2.10, it can be seen that the curve
obtained with the mechanical characteristic of the FPT N67ENT 175 [kW ] fits sufficiently the
threshold curve obtained after the reduction of the mechanical characteristic of the original
Cummins QSM280 210 [kW ] with R = 175/210 ∼ = 0.83. Indeed the resulting EM boosting
torque profiles obtained from the hybridization tests performed using the reduction coefficients
R indicated in Figure 2.10 and multiplied by 0.75 are very similar to the results achieved with
the downsized ICEs mechanical characteristics limited with R = 0.75. For brevity purpose,
only the latter findings will be reported in the next section.
Such outcome proves that only the original ICE parameters and the rated torque of the
downsized engine need to be known in order to carry out an effective preliminary hybridiza-
tion process for any powertrain with an equivalent circuital model referred to the engine shaft.
Furthermore, a sweep among the values of the R coefficient could be used to perform opti-
mization analysis on the hybrid powertrain in order to establish which ones could be the most
advantageous hybridization degrees.
2.6 Hybridization results

For brevity purpose, only the outcomes achieved using the downsized ICEs characteristics
limited at the 75% are here reported and, among them, only the results that are relevant to
identify a rated torque specification for the electrical machine design are presented. Figures
2.11-2.12-2.13 show the EM boosting torque profiles filtered with time constants τ in the range
200 ÷ 1000 [s] for each downsized ICE during the transportation duty cycle, which is the most
critical one.
As expected, the torque requirements to the electric motor increase with the decrease of the
engine rated power, i.e. with the increase of the hybridization factor. Besides, in the same duty
27
Hybridization process
Cummins QSM280 210kW torque reduction Downsized ICEs torque

1400 1400
R=100%
1200 1200
1000 R=75% 1000

Torque [Nm]
800 800
R=50%
600 600
400 400
R=25% Cummins QSM280 210kW
R=83% FPT Industrial N67 ENT 175 kW (238 Hp)
200 R=77% 200 FPT Industrial N67 ENT 151 kW (205 Hp)
0 0
800 1000 1200 1400 1600 1800 2000 2200 800 1000 1200 1400 1600 1800 2000 2200
Cummins QSM280 210kW power reduction Downsized ICEs power

R=100%
200 200
R=75%
150 150
Power [kW]
R=50%
100 100
50 R=25% 50 Cummins QSM280 210kW

0 0
800 1000 1200 1400 1600 1800 2000 2200 800 1000 1200 1400 1600 1800 2000 2200
Speed [rpm] Speed [rpm]
Figure 2.10: Hybridization processes comparison.
28
Figure 2.11: Powertrain with FPT N67ENT 175[kW ] at 75% of max. performance. EM filtered
boosting torque during transportation duty cycle.
cycle with the same downsized ICE, the equivalent thermal torque has lower peak values for
higher thermal time constants τ , as previously disclosed in Section 2.3. Indeed machines with
higher τ heat up much slower when an overload current flows in the winding. But high values
of τ demand high overload capabilities to fulfill the peak torque requests.
Referring to Figures 2.11-2.12-2.13 the torque requirements to size the electric motor has
been chosen for each hybridization test. To do so the thermal time constant τ has been supposed
in the range 800 ÷ 1000 [s]. Synchronous SPM machines have been chosen for this project,
thus the rated speed has been set to the maximum value nR = 2300 [rpm] demanded in the
power harrow duty cycle. Thanks to this choice the EM should be controlled always in the
MTPA working region (Maximum Torque Per Ampere), so that the driving conversion group
components could be sized for lower current values and also the battery should withstand lower
discharging current [31]. Moreover, SPM machines have a very limited flux-weakened working
region, therefore there are few advantages in exploiting it. These concepts will be further
presented in the next chapter.
Once a rated torque TR [N m] and a rated speed nR [rpm] have been chosen, the rated power
PEM = PR [W ] has been computed for every EM as follows:
2π
PR = TR · nR ·
60
Consequently, the actual hybridization factor H has been obtained as expressed in Equation
1.1 and each hybrid powertrain has been classified according to Table 1.1. Besides the overload
torque capability TOL has been selected to make sure that the motor can fulfill the peak torque
requests. The chosen specifications are summarized in Table 2.1. It is relevant to point out
that such values are not mandatory. They are only needed to size the electric motor but they
could be changed or not fully observed during the machine design and analysis, providing that
in the end the motor is able to withstand all the duty cycles with sufficient performances.
29
Downsized ICE FPT 175 [kW ] FPT 151 [kW ] FPT 129 [kW ]
Equivalent R 0.83 · 0.75 0.77 · 0.75 0.65 · 0.75
TR [N m] 120 140 180
nR [rpm] 2300 2300 2300
PR [kW ] 29 34 43
H [%] 14 18 25
Type Mild-hybrid Mild-hybrid Full-hybrid
TOL [N m] 4 × TR 3.5 × TR 3.5 × TR
Table 2.1: EM rated and overload specifications for each hybrid powertrain.
30
Chapter 3
Electrical machines sizing
The electromagnetic sizing of a liquid cooled radial-flux SPM motor is presented in this chapter.
At the beginning, a description of the significant features of this type of machine and its
control system are provided, so that advantages and drawbacks are standing out. Other types
of EM that could be implemented in this application will be mentioned in order to give expla-
nations about the choice made for this project.
Three different SPM motors have been designed for one of the three hybrid powertrain under
analysis. The EMs differ mainly in the number of poles. The EMs sizing has been performed
starting from the performance specifications previously identified and under some realistic vol-
ume limits. The limited displacement available for the EM has been a very challenging constrain
and a particular sizing method have been implemented to cope with it. All the design choices
as well as the method used to find out the main sizes of the machines will be described in detail.
At the end, a material cost estimation and comparison are reported.
3.1 Machine description

The chosen type of electric motor is illustrated in Figure 3.1. It is an inner rotor radial-flux
surface permanent magnet (SPM) machine. Such kind of motor is defined as a synchronous
machine because its mechanical rotating speed is proportional, namely synchronized, to the
electrical frequency of the three-phase alternating current which flows in the armature winding
and it is independent of the current amplitude, i.e. of its electrical load [24]. This type
of machine has been chosen for its high power density and easy manufacturing process. If
compared to an induction motor (IM) with equal rated power, a SPM motor results to be
much more compact. Moreover, despite SPM machines are usually more suitable for high
speed applications than high torque ones, they have even a higher torque density than IMs,
i.e. with equal rated power and equal rated torque a SPM machine occupies less volume.
Furthermore, with equal rated power, permanent magnet machines are generally more efficient
than induction motors because they are much more less affected by rotor Joule losses. Indeed,
the IMs working principle requires that a relevant amount of current is induced in the rotor
cage or in the rotor winding, so that additional Joule losses are unavoidable. On the contrary,
in permanent magnet machines Joule losses are related to eddy current induced in the PMs;
such current is negligible at very low speed and can be severely limited with a proper rotor
design even at high speed because it is not involved in the torque generation. Instead, as far as
overload requirements are concerned, they could be satisfied in an easier and cheaper way with
an IM. Indeed, IMs withstand safer high peak torque values and overload working conditions,
while SPM motors could be very limited in terms of overload capability to avoid the permanent
magnets’ demagnetization, unless very expensive magnets are used. Besides, PMs represent
an important constrain also for the allowed working temperatures, while IMs are less sensitive
to environmental conditions. For all these reasons, choosing the PM and its sizing is a very
challenging key point in the design of a SPM machine. In automotive applications, interior
permanent magnet motors are also implemented. With a proper design and control system
31
Sez. A-A' Sez. B-B'
B-B' A-A'
Water jacket
Water jacket
Stator End-winding
αs Stator stack
Magnet
tm
Magnet
Rotor stack
αm
Flange
Rotor
De Dshaft Bearings Ds
Shaft
Lstk
Ltot
Figure 3.1: Water cooled radial flux SPM motor.
such machines can perform greatly in large speed ranges and the PMs are better shielded
against demagnetization. Yet the manufacturing process is more expensive and the design
more complex. Hence, these machines could not be worthwhile for the low speed application
that concerns this thesis.
The stator core is manufactured stacking a large number of electrical steel laminations
with thickness in the range of 0.25 ÷ 1.00 [mm] until the stack reaches the required length.
The electrical steel used in electric motors is a non-oriented grains silicon-iron alloy. A small
percentage of silicon (less than 5%) is needed to provide sufficient mechanical strength. Indeed,
pure iron is very brittle and it is not able to withstand the cutting process, which is needed
to make the slots geometries, without breaking. Whereas most of the alloy (above 95%) is
composed by iron in order to fulfill the function for which such material is adopted, i.e. to
provide a magnetic circuit path with very low reluctance R [Aturns/W b] [H −1 ]. To do so, the
alloy should behave as a so-called soft magnetic material. Generally the magnetic behavior of
a material is represented by the constitutive equation:
B̄ = µH̄ = µrel µ0 H̄ (3.1)
Where µ0 = 4π · 10− 7 [H/m] is the absolute permeability of the air, µ [H/m] and µrel are re-
spectively the absolute and relative permeability of the material, B̄ is the magnetic flux density
vector and H̄ is the magnetic field vector. Non-oriented grains electrical steels are isotropic ma-
terials. This means that µrel assumes the same value regardless of the magnetization direction
and Equation 3.1 is still valid in terms of vectors amplitudes B and H. Indeed, anisotropic soft
magnetic materials such as oriented grains steels are not suitable for making electrical machines
because the stator core is subjected to a rotating field without a preferential magnetization di-
rection. By definition, magnetic reluctance is inversely proportional to permeability:
1 1
R∝ ∝
µ µrel
32
Thus an electrical steel should have a high relative permeability µrel to allow high flux density
values B [T ] with low magnetizing efforts, i.e. low applied magnetic fields H [A/m]. Unfortu-
nately, such feature is always linked to a magnetic saturation phenomenon in iron alloys which
causes a non-linear magnetic behavior. Indeed, Figure 3.2(a) shows that the BH curve of a
soft magnetic material is not a straight line. It happens because the relative permeability is not
constant, but it depends on the magnetizing effort, so that Equation 3.1 becomes a non-linear
relation between B and H:
µrel = µrel (H)
The permeability reaches its maximum value near a particular point of the magnetic charac-
teristic called ’knee’. Beyond the ’knee’, the permeability rapidly decreases to a value near µ0 ,
so the gain in the flux density is smaller than the rise in the applied field and consequently it
is not worthwhile to apply such high magnetizing efforts. Therefore, a soft magnetic material
is exploited conveniently only near the ’knee’. Moreover, even the material response to an
alternating applied field is strongly non-linear, outcoming in the hysteresis loop illustrated in
Figure 3.2(a). Such phenomenon produces an energy dissipation into heat that is proportional
to the area delimited by the loop curve. Indeed the specific energy loss w [J/m3 ] is defined as
follows: I
w = BdH
So a soft magnetic material should have a very restricted hysteresis loop. The hysteresis losses
are added to eddy current Joule losses, which are also triggered by alternating applied fields.
It results then in overall iron losses. These losses increase with the frequency of the alternating
applied field. Such topic will be investigated later and deeper. The stator core is subjected to
an alternating magnetic field both in the time and space domain, i.e. a rotating field. So it is
deleterious in terms of efficiency and heat dissipation to build it in a single iron block. Hence the
use of a lamination stack. Moreover, laminations are electrically insulated in the axial direction,
commonly through very thin oxidation layers or insulating enamels. Such expedient avoids the
flow of eddy current induced by the alternating radial magnetic flux and consequently it limits
severely the related Joule losses.
The rotor yoke is subjected to a constant magnetic flux in steady state load conditions.
Hence it could be manufactured using a single iron block. Yet when a change in the motor load
occurs and the feeding current varies rapidly, a magnetic field variation affects the rotor yoke
too. Then the resulting losses contribute in worsening the overall efficiency and the reaction
field, because the induced eddy current slows down the electromechanical dynamic of the ma-
chine. Thus, for this project, a laminated rotor has been adopted. Moreover, thanks to this
choice, the inertia of the machine decreases with a further improvement in the electromechanical
responsiveness.
The excitation magnetic field is provided by some permanent magnets (PMs) that are pasted
on the rotor surface and wrapped with a carbon fiber or fiberglass bandage to prevent runaways
at high rotational speed. A permanent magnet is a so-called hard magnetic material. Its BH
characteristic is depicted in Figure 3.2(b). It can be observed that a hard magnetic material
has a very large hysteresis loop if compared to the one of soft magnetic materials. Therefore,
such materials are not suitable to provide low reluctance magnetic paths, but they are used
to produce a constant magnetic flux. In order to achieve it, they have to work in the second
quadrant of the BH plane, along the so-called recoil line, whose points can be represented by
the following equation (namely the recoil line equation):
B = Brem − µrec µ0 H
If the working point reaches the so-called ’knee’ the PM is irreversibly demagnetized. The
performance of a magnet is measured in terms of remaneance Brem , coercivity Hcb and intrinsic
field Hcj . These parameters will be described thoroughly later, when the magnet choice will be
discussed. The PMs are manufactured as arc-shaped blocks that cover the entire stack length.
Thus, the number of machine poles is equal to the number of magnets. Sometimes, in critical
high speed applications, i.e. above 6000 [rpm], with the objective to limit the Joule losses due
33
B [T ] Magnetization B [T ]
Bknee
curve
Brem
knee
Hysteresys Recoil
loop line
Hknee H [A/m] Hcb H [A/m]

Hcj
w [J/m3 ]
(a) Soft magnetic material (b) Hard magnetic material
Figure 3.2: Magnetic materials BH characteristics.
to eddy current flowing in the magnets, the machine poles could be composed by many magnet
blocks electrically insulated along the axial direction, emulating the expedient used to realize
the stator. In other applications, when the motor movement must be very accurate, the PMs are
often skewed to minimize the torque ripple. These solutions surely provide better performances
but they require more expensive manufacturing techniques. Neither the former case nor the
latter concern this project, therefore they have not been adopted. Nevertheless, the magnet
losses have been taken into account during the analysis of the motors and the related results
will be presented later.
Thanks to the use of permanent magnets, the motor is a brushless machine, as it does not
need an excitation winding on the rotor. Hence, brushed contacts are not necessary. A brushless
motor has a greater overall efficiency, a better heat dissipation and lower maintenance costs
than an equivalent brushed machine. Manufacturing costs are higher because the production of
the permanent magnets is much more expensive than a rotor field winding, but the advantages
given by the use of PMs justify the costs in an automotive application [32].
The stator laminations are cut to make the slots that house the armature winding coils sides.
The motor has a three phase double-layer distributed winding made of copper enameled wires.
Such layout has been preferred to a concentrated winding in order to improve the sinusoidal
waveform of the back emf (electromotive force) and reduce the armature reaction, despite a
concentrated layout could fit better a compact machine with a low number of slots. Double layer
coils has been chosen to allow the creation of fractional-slots winding and to provide an easy
implementation of short-pitches, which reduce the torque ripple and the harmonic content of
the back emf and the armature mmf (magnetomotive force). Wires have been adopted instead
of hair-pins only to reduce the manufacturing costs, even if it has been proven in [33] that
hair-pins windings are advantageous in terms of performances and volume reduction for the
application under analysis. Photos of the just mentioned windings are given in Figures 3.3-3.4-
3.5. In order to provide electrical insulation among different turns and between different coils,
the wires are enameled with a polyamide coating and the coils sides are wrapped with Nomex ®
tapes. Furthermore, the interior surface of the slots is covered with phenolic bakelite paper
sheets to guarantee electrical insulation between conductors and mass. Diamagnetic bakelite
wedges close the slots to avoid any unwanted movements of the coils sides. Finally, copper
conductors have been preferred rather than aluminum wires to allow higher current density.
The motor must have a very high power density to fulfill the displacement constrains. Such
purpose can be achieved through the increase of the current density in the armature winding.
Therefore, the specific Joule losses increase and liquid cooling systems are needed to avoid
excessive overtemperatures that can damage permanently the previous mentioned insulating
34
Figure 3.4: Distributed with Figure 3.5: Distributed with
Figure 3.3: Concentrated
wires hair-pins
ia +
ib
EM
ic
i∗
d = 0
[Sa Sb Sc ]
+
P Id
id
− u∗
d u∗
α
abc ⇒ dq dq ⇒ αβ SV M
− u∗
q u∗
β
iq
P Iq
+
i∗
q ∝ T
θe θe
Figure 3.6: Electric drive and control system outline.
materials inside the slots. If the machine would be self-ventilated or air-cooled, it would require
a larger outer surface to dissipate the same heat quantity, otherwise the current density would be
lower. For this project, an inox steel water-jacket with a circumferential duct has been adopted.
An external pump (not shown in Figure 3.1) moves the cooling fluid inside the channel. The
water-jacket has to be designed so that it encloses entirely the laminations stack and it provides
a proper heat dissipation. The cooling system design will be presented in detail in Chapter 5.
The rotor stack is built around a stainless steel shaft, through which the electromagnetic
torque developed by the machine is made available to the powertrain. The shaft is hinged to
two bearings inserted inside the front and rear plates that close the machine in axial direction.
3.2 Electric drive and control system

Two main configurations of the control system can be implemented to drive a SPM motor:
the machine can be controlled as a BLDC (brushless DC ) motor, injecting rectangular current
waveforms, or it can be fed with sinusoidal current as a BLAC (brushless AC ) motor. The
latter electric drive has been preferred because it provides better torque performance in terms
of ripple and harmonic content [31]. The outline of the system is shown in Figure 3.6.
The machine is driven through a torque loop control. The feedback signals are the phase
35
current, which are instantaneously measured by some current transducers. The three phase
signal (ia , ib , ic ) are converted into their related dq components, referring to an orthogonal
system that rotates at the same speed of the rotor and where the d-axis is aligned with the PM
magnetization direction. To perform a synchronous transformation the instantaneous electrical
angular position of the rotor θe [◦ ] is needed. So the mechanical angular position θ is measured by
an encoder installed on the shaft and consequently the electrical angle is obtained as θe = pθ.
The synchronous transformation of a balanced three phase quantity outcomes two constant
signals, as represented by the following relations:
ia = Icos(θe + θi ) id = Icos(θi )
ib = Icos(θe + θi − 120◦ ) =⇒ (3.2)
ic = Icos(θe + θi − 240◦ ) iq = Isin(θi )
To make sure that the phase feeding current have no zero sequence component, the motor
winding is star-connected without neutral wire, so that third harmonic current cannot flow.
Hence explained why the control system works with quantities defined about a rotating ref-
erence. Indeed such signals are constants or step-varying, thus they can be suitably tracked
by PI (Proportional-Integral) controllers. The feedback dq quantities are then compared to
the reference dq signals (i∗d , i∗q ). The resulting errors are processed by the PI controllers that
produce the dq components of the needed voltage. Such quantities are transformed back to the
orthogonal stationary αβ reference system and then they are given to the SVM (Space Vector
Modulation) algorithm that generates the gate signals for the three legs of the inverter.
The machine is always driven in MTPA condition (Maximum Torque Per Ampere). For
SPM motors it can be proven that, neglecting any non-linear iron saturation phenomena, the
MTPA condition is achieved by the injection of a three phase quadrature current, i.e. with only
q axis component in the rotating synchronous system. Thus the i∗d reference signals is always
set to zero so that the phase a starting angle results: θi = 90◦ . Instead the i∗q reference, which
is set by the power management system, defines the average torque value that the motor should
develop at any given time. Indeed the instantaneous electromagnetic torque can be expressed
in the dq rotating reference by the following well known relation [24, 31]:
3 ˇ ∂wI 3 ∂wI
τ= p= λ̄ ī + = λd iq − λq id + (3.3)
2 ∂θ 2 ∂θ
Where λ̄ˇ = λ − jλ is the space vector of the flux linkage, while ī = i + ji is the space
d q d q
vector of the feeding current. The variation coenergy ∂wI /∂θ is not involved in the mean
torque development, but it generates only the torque ripple component. SPM motors have
an isotropic magnetic circuit. This means that the magnetic reluctances along axis d and q
are equal. Indeed the differential permeability of the magnets is quite similar to the relative
permeability of air (µrec ∼
= 1). Therefore also the inductances of axis d and q are the same:
Ld = Lq . Under these statements and still neglecting any non-linearity, Equation 3.3 becomes:
3
T = pΛ0 iq
2
Where Λ0 is the peak value of the magnetic flux produced by the PM and linked with the
winding in unload condition. The previous expression states that in a SPM machine only the
quadrature component of the current generates an average torque. That is why the MTPA
condition requires only q component. So, when working in MTPA, Equation 3.3 can be further
expressed as follows:
3
T = pΛ0 I (3.4)
2
Thus the mean torque results proportional to the peak current I.
Figure 3.7 reports the steady-state working regions of a SPM machine in the id − iq plane
and the mechanical characteristic of this kind of motor. It can be seen that the rated torque,
which in the drawing corresponds to the maximum torque, is available in MTPA condition
until the so-called base speed ΩB . For speed values higher than ΩB the voltage limit forces the
36
ΩB
iq T
Ω0
TR
Ω
TR Ilim
T
flux
weakening
M MTPA
T
ΩM AX P
A
−Isc −Ilim id ΩB Ω0 ΩM AX Ω
(a) Working regions (a) Mechanical characteristic
Figure 3.7: Steady-state working regions (a) and mechanical characteristic (b) of a SPM motor.
electric drive to operate in flux-weakening mode, i.e. with an amount of demagnetizing feeding
current given by a negative d-axis component. The MTPA region is still available but the
torque outcomes lower than the flux-weakening curve at equal speed. However beyond the base
speed the maximum torque that can be developed is smaller, as the mechanical characteristic
bends down. SPM motors have a limited flux-weakening region because they present very low
inductances. So they are commonly designed to work always in MTPA condition, not further
of ΩB . Thus in this project the base speed corresponds to the rated speed. It can be stated
that the three motors have been designed according to the base point features. A detailed
dissertation about these topics is provided in [24, 31, 34].
3.3 Design specifications and volume constrains

Three SPM motors have been designed for the mild-hybrid powertrain with the downsized ICE
FPT N67ENT 175 [kW ] (238 [Hp]). The main difference between the three motors consists in
the number of poles 2p. The specifications identified after the hybridization tests and reported
in Table 2.1 require high rated and overload torque and a relatively low rated speed. Such
requirements lead to the design of so-called torque motors, which are characterized by low
machine ratios, i.e. larger stator inner diameters compared to the axial stack length, and high
numbers of poles. But, with equal rated speed nR [rpm], synchronous machines with high
numbers of pole pairs p require feeding current with high rated frequency fR [Hz], as stated by
the following well known relation [24]:
pnR
fR =
60
A three-phase voltage with the same fundamental frequency must be synthesized by the inverter,
which should work at a very high switching frequency to produce clean sinusoidal waveforms,
otherwise the harmonic content of the feeding current could affect the performances and the
efficiency of the entire electric drive. To do so more expensive electronic power switches are
needed and even with high-performance components the inverter losses increase with the square
of the switching frequency. Moreover the machine unload losses, which are due mainly to
iron hysteresis, increase more than linearly with the fundamental electrical frequency. So, for
the application that concerns this project is recommended not to overcome the fundamental
electrical frequency of 200 [Hz]. Thus machines with a number of poles from 6 to 10 have been
37
Original ICE Original
Cummins QSM powertrain
1280
210 kW (280 Hp) Engine
shaft
1306
1054 190
Downsized ICE
FPT N67 ENT
Hybrid
175 kW (238 Hp) powertrain
1005
E Engine
400
C3
shaft
M
62
Figure 3.8: Size constrains in the mild-hybrid powertrain.
Specification Symbol Value Unit

Rated torque TR 120 [N m]
Rated speed nR 2300 [rpm]
Rated power PR 29 [kW ]
Max. overload torque TOL 4 × TR [N m]
Number of poles 2p 6 8 10
Rated frequency fR 115 153 192 [Hz]
Size constrains
Max. total axial length LtotM AX 190 [mm]
Max. total outer diameter DtotM AX 400 [mm]
Table 3.1: Design specifications and size constrains.
designed.
Displacement constrains are also very important in this application. In order to realize a
cost-effective hybrid powertrain, the electrical machine must occupy no more than the volume
made available by the engine displacement’s reduction, otherwise very expensive modifications
of the tractor chassis could be required. Then the maximum sizes suitable for the EM have
been identified looking at the commercial datasheets of the chosen original engine and down-
sized ICE [27, 30]. The original engine has a declared length of 1306 [mm] while the downsized
ICE is 1054 [mm] long, clearing a space of 1306 − 1054 = 252 [mm] along the shaft. Consid-
ering to allocate 62 [mm] for the optional clutch between the ICE and the EM and for the
needed mechanical connections, the total axial length of the electric motor should not overcome
LtotM AX = 190 [mm]. The maximum total outer diameter of the machine has been supposed of
DtotM AX = 400 [mm], as the shaft is installed near the lower side of the engine. The reasoning
just presented above is illustrated in Figure 3.8. All the design specifications are summarized
in Table 3.1.
38
3.4 Choice of the permanent magnet
Before the preliminary sizing of the machines the permanent magnet type must be chosen. The
choice has been based on the behavior of hard magnetic materials and on the well-known sizing
torque relation of synchronous radial flux motors, which is derived by a force balance at the air
gap of an equivalent ideal machine [24].
π 2 (1)
TR = D Lstk B̂gR K̂sR (3.5)
4 s
Ds and Lstk are the main electromagnetic sizes, respectively the inner stator diameter and the
(1)
stack axial length. B̂gR is the fundamental peak value of the magnetic flux density distribution
in the air gap ([T ]) at rated load, i.e. the amplitude of the first harmonic of the distribution,
also called magnetic load. K̂sR is the rated electric load peak value, which is defined as the
amplitude of the linear current density ([A/m]) that would flow on the inner surface of the ideal
machine’s stator. The choice and computations of these parameters will be presented in the
next section, as the preliminary sizing consists exactly in finding out the main dimensions Ds
and Lstk . But, to do so, the magnet type must be previously selected.
All the machine parameters depends heavily on this choice, but primarily it affects the
flux density distribution in the air gap. The performances of a magnet are measured in terms
of remaneance Brem , coercivity Hcb and intrinsic field Hcj . These parameters are shown in
(1)
Figure 3.2(b). In particular, B̂gR depends heavily on Brem . PMs with higher remaneance
allow higher flux density in the air gap. Hence the machine could have a smaller size and
smaller electric load, resulting to be more compact, more efficient and easier to cool. Strong
magnets are characterized by high energy density. Usually they have high coercivities and also
high remaneances. Other significant key points that must be considered in the magnet choice
are the resistance against the demagnetization and the temperature stability, which strongly
depend on each other. A PM is all the more difficult to demagnetize as it has a high intrinsic
field value. But this feature could not be enough in those applications when the working
temperature varies in a large range. Indeed the intrinsic field value as well as the remaneance
of a permanent magnet tend to decrease with an increasing working temperature. Consequently
the PM becomes more sensitive to demagnetization and its performances drop down. Last but
not least, the cost of the magnet material and the manufacturing process must be taken into
account as the PMs represent about the 70% of the overall cost of the machine.
Four kinds of permanent magnet materials can be used in an electrical machine. Ferrite
magnets, also known as ceramic magnets, are the cheapest PMs available. They are electrical
insulating, corrosion free and very resilient against harsh environments [35]. On the other
hand they have the lowest performances in terms of remaneance and coercivity. Furthermore
their remaneance decreases heavily with a working temperature rise, but, on the contrary, their
intrinsic field increases. Thus, the temperature is not a problem in terms of demagnetization for
this kind of magnets, but it affects only the performances. AlNiCo magnets (Aluminum-Nickel-
Cobalt) are characterized by the highest remaneances (up to 1.4 [T ]) and the highest thermal
stability [36]. Unfortunately their coercivities and intrinsic fields are very low, then these kind
of magnets are the weakest against the demagnetization. So they are rarely adopted in electrical
machines and, even for those particular applications where they could be usefully exploited, the
design results very challenging. As best practice, their use should be avoided when high overload
current are involved. Finally rare-earth magnets are commonly used in electrical machines.
Two kind of these PMs can be adopted. NdFeB sintered alloys (Neodimium-Iron-Boron) are
the cheaper ones, but anyway they are several times more expensive than ferrite magnets. They
are the stronger PMs available [37], with coercivities up to 1000 [kA/m] and remaneances up
to 1.3 [T ], but they suffer a lot when a rise in the working temperature occurs, in terms of
both performances and weakening against demagnetization. High-temperature NdFeB PMs
withstand demagnetizing reaction fields quite better. Clearly they are more expensive as they
require small percentages of the so-called ultra-rare earths, like Disprosium and Gadolinum.
NdFeB magnets are always provided with a NiCuNi (Nickel-Copper-Nickel) coating to prevent
corrosion. However such kind of PMs are not suitable to work in especially harsh environments.
39
The other type of rare-earth PMs are the SmCo alloys (Samarium-Cobalt). They exhibits lower
remaneance and coercivity values than NdFeB alloys, but they are much more stable with the
temperature and stronger against demagnetization [38]. Moreover they are less affected by
corrosion and humidity. Unfortunately such magnets are the most expensive due to the presence
of cobalt.
In this project AlNiCo PMs are rejected firstly due to the great overload capability specifi-
cation and the high demagnetizing current involved. When possible, ferrite magnets should be
preferred to rare-earth PMs in order to obtain a cost-effective machine. However, for the chosen
type of EM, the machine specifications cannot be fulfilled using ferrite magnets. To prove this
statement a brief preliminary sizing has been performed adopting unfeasible but conservative
hypothesis.
The magnetic parameters of the most high-performance ferrite magnet, namely Y40 grade
[35], has been assumed at the working temperature of 100◦ C:
a remaneance Brem = 0.40 [T ];
an intrinsic magnetic field Hcj = 450 [kA/m];
a differential permeability of the recoil line µrec = 1, i.e. the slope of the line.
The intrinsic magnetic field can be approximate with the ’knee’ magnetic field value Hknee =
Hcj in a security-side way. Then the ’knee’ flux density value Bknee has been obtained from
the recoil line equation as follows:
Bknee = Brem − µrec µ0 Hknee = −0.17 [T ]
The working point of the magnet must not go under the ’knee’, otherwise the PM outcomes
demagnetized.
The thickness of the magnet has been set to tm = 12[mm] to provide a good strength against
demagnetization. Considering an air gap thickness of g = 1 [mm], the equivalent magnetic air
gap has been hypothesized as g II = 1.3 [mm]. Hence, the flux density value in the air gap
results from the computation of the equivalent magnetic circuit of the machine at rated load:
Brem
B̂g = µrec g II
= 0.36 [T ]
1+ tm
Supposing a flux density square-wave distribution in the air gap and that the magnet covers an
entire pole pitch, the amplitude of the fundamental flux density has been computed as follows:
(1) 4
B̂gR = B̂g = 0.46 [T ]
π
The flux density in the air gap can be considered the same value that define the working point
of the magnet Bm = B̂g = 0.36 [T ], because the PMs are placed in front of the air gap. Hence
the demagnetization margin in terms of flux density ∆Bdemg has also been derived:
∆Bdemg = Bm − Bknee = 0.53 [T ]
At rated load a large security coefficient ksecR = 8 has been adopted to avoid demagnetization
even when the maximum overload is applied. So the allowed flux density variation for the PM
working point has outcome:
∆Bdemg
∆BR = = 0.07 [T ]
ksecR
Considering the maximum total outer diameter of the machine DtotM AX = 400 [mm], a stator
maximum inner diameter Ds = 300 [mm] has been chosen, leaving a minimum radial length
of 100 [mm] for the laminations height and the water jacket thickness. Then the peak rated
electric load has been computed through the reaction field relation for the 10 poles machine,
40
N27 UH N35 UH N42 UH
20◦
Remaneance Brem [T ] 1.03 1.17 1.28
◦
20
Coercivity Hcb [kA/m] 764 875 875
20◦
Intrinsic field Hcj 1989 1989 1989
Energy density BHmax [kJ/m3 ] 199 263 318
Temp.coeff.induction αBrem [%/◦ C] -0.1 -0.1 -0.1
Temp.coeff.intrinsic αHcj [%/◦ C] -0.55 -0.55 -0.55
Max.temperature TmM AX [◦ C] 180 180 180
Table 3.2: Selected NdFeB sintered magnets for preliminary sizing.
which withstand better the demagnetization and consequently it allows higher current loads:
2p∆BR g I + µtrec

m
K̂sR = ∼ 24 [kA/m]
=
µ0 Ds
The air gap has been raised to g I = 1.1 [mm] to consider the magnetic reluctance due to slots
openings. Finally, remembering that TR = 120 [N m] and using the well-known sizing torque
equation for synchronous radial flux machines, the stack length has been obtained:
4TR
Lstk = (1)
= 154 [mm]
2
πDs B̂gR K̂sR
The resulting axial size does not fulfill the volume constrains because with such stack length
the EM would surely becomes longer than LtotM AX = 190 [mm], considering the space needed
for the end-windings and the front and rear plates. Moreover some suppositions made during
the sizing are unfeasible and too optimistic. Therefore it has been stated that ferrite magnets
are not suitable for this project, as they do not allow to fulfill the volume constrains even in
the best scenario.
So rare-earth magnets must be adopted. Among them sintered NdFeB magnets have been
preferred rather than SmCo alloys for cost-related reasons. Nevertheless in the application under
analysis temperature stability as well as strength against the demagnetization are also required.
Indeed the machine will be placed very close to a Diesel engine and power electronic components,
which are relevant heat sources. Furthermore the EM should withstand high overload peak
torque with consequently high specific Joule losses. For this reasons the high-temperature
UH class has been chosen. UH alloys, that are not the top-performance ones, guarantee a
20◦
declared maximum working temperature of 180◦ C, an intrinsic field Hcj = 1989 [kA/m] at
room temperature and variations of Brem and Hcj with the temperature respectively of αBrem =
−0.1 %/◦ C and αHcj = −0.55 %/◦ C [37]. As what concerns the grade, three options has been
tried and compared during the preliminary sizing. The parameters of the three magnets at
room temperature are reported in Table 3.2. The data have been taken from the minimum
values in [37], according to the Chinese standard. Extracts of the datasheet are reported in
Appendix 5.3.
3.5 Preliminary sizing

The commonly implemented preliminary sizing methods find out the machine main sizes Ds
and Lstk through Equation 3.5 and by imposing a security margin against the demagnetization
[24], likewise the procedure presented in the previous section that has been followed to exclude
ferrite magnets. The air gap thickness must be previously chosen and some feasible hypothesis
on the saturation at rated load must be done. Also the magnet parameters must be known and
its shape and thickness must be selected. To achieve their goal, these methods require to set
at least one size between Ds and Lstk or alternatively their ratio Lstk /Ds , namely the machine
aspect ratio. In order to identify the best choice, many trials are usually performed varying the
input parameters such as the air gap thickness, the PM size or the machine aspect ratio, in a
41
δ4
δ3
δ2 d
Ltot
αew
δ1
ps
τp0 Lstk
Figure 3.9: End-winding geometry.
range of feasible values. Nevertheless such methods are focused only on the stack length sizing,
so the machine total displacement cannot be known until the stator dimensions and the winding
features are completely defined. Therefore it is very complicated to fulfill size constrains. Then
an alternative procedure has been adopted to exploit wisely the available volume.
The main purpose is to take advantage of the total axial length limit as much as possible,
because it represents the major size constrain considering that a radial flux motor has been
chosen. Hence the total axial length has been imposed equal to the maximum axial dimension
allowed.
Ltot = LtotM AX = 190 [mm]
The geometry of the end-winding has been studied at the beginning. Figure 3.9 outlines
the end-turns of a distributed winding with all the geometrical dimensions used in the sizing
procedure. The layout shown in Figure 3.9 can be realized both with wires and hair-pins, so
the method can be also adopted to size hair-pins windings. The machine total length is related
to the stack length as follows:
Ltot = Lstk + 2(δ1 + δ2 + δ3 + δ4 ) (3.6)
The dimensions δ1 , δ3 and δ4 are independent parameters that must be set feasibly by the
designer. In this project the values chosen for all the EMs are: δ1 = δ3 = 10[mm], δ4 = 20[mm].
Instead δ2 depends on the motor geometry.
τp0
δ2 = tan(αew ) (3.7)
2
Where the end-winding slope angle αew can be further developed:
d
αew = arcsin
ps
d is a distance fixed by the designer: in this project d = 10 [mm] has been set to guarantee a
sufficient heat dissipation through natural convection in air. The slot pitch ps is related to Ds
and to the number of slots Qs :
πDs
ps =
Qs
The coil span τp0 cannot be known until the winding layout is completely defined, but it can be
approximated with the pole pitch τp in a conservative way. Indeed, τp0 = τp in case of integer
42
Table 3.3: Magnets parameters at working temperature.
N27 UH N35 UH N42 UH
Diff. permeability µrec [/] 1.07 1.06 1.16
Working temp. Tm [◦ C] 120 120 120
Remaneance Brem [T ] 0.927 1.053 1.152
Coercivity Hcb [kA/m] 688 787 788
’Knee’ field Hknee [kA/m] -895 -895 -895
’Knee’ induction Bknee [T ] -0.28 -0.14 -0.16
slot windings, while τp0 < τp when a chorded winding is adopted. So if a short-pitch will be
implemented the end-turns will outcome a bit shorter than expected, but the axial dimension
constrain will be fulfilled anyway. The pole pitch depends on Ds and on the number of poles:
πDs
τp =
2p
Then, Equation 3.7 can be rewritten substituting the previous expressions:
πDs dQ
s
δ2 = tan arcsin (3.8)
4p πDs
The stack length can be obtained from Equation 3.5.
4TR
Lstk = (1)
(3.9)
2
πDs B̂gR K̂sR
(1)
The rated magnetic load B̂gR is computed starting from the PM parameters in the same way
presented in the previous section. The remanenance Brem and the ’knee’ flux density value
Bknee are computed at the supposed working temperature of 120◦ C using the temperature
coefficients and the recoil line equation. The results are reported in Table 3.3 for each PM
under analysis. The flux density of the magnet working point at rated load is expressed as
follows:
Brem
Bm = B̂g = g II
(3.10)
1 + µrec
tm
As explained in the previous section, the flux density value in the magnet and in the air gap
can be supposed equal. The magnetic equivalent air gap g II is given by:
g II = g I ksatR = gkcarter ksatR = 1.32 [mm] (3.11)
The air gap thickness has been chosen of g = 1 [mm] for all the three machines, including
0.6 [mm] of air and 0.4 [mm] of diamagnetic fiberglass bandage for the magnets. The Carter
coefficient, which takes into account the reluctance due to slot openings, has been feasibly
supposed kcarter = 1.1, while the saturation coefficient, which considers the mmf needed to
magnetize the iron paths at rated load, has been set to ksatR = 1.2. The thickness of the PM
tm must be chosen by the designer; more than one size has been tried.
A PM covering of 2/3 of the pole pitch has been selected for all the motors, as it can be
proven that such choice gives the best ratio between the resulting amplitude of the flux density
(1)
fundamental distribution B̂gR and the PM volume. Hence, the electrical covering angle of the
magnet is 2αme
= 120◦ . So, supposing a square-wave distribution of the flux density in the air
gap, the expression of the rated magnetic load becomes:
√
(1) 4 e 2 3
B̂gR = B̂g sin(αm ) = B̂g (3.12)
π π
43
And the substitution of Equation 3.10 yields to:
√
(1) 2 3 Brem
B̂gR = (3.13)
π 1 + µrec gII
tm
Considering a ×4 overload capability the security factor against demagnetization at rated load
has been set to ksecR = 8, so that the margin at maximum overload should be near ksecOL ∼ = 2.
Then the allowed variation of flux density is expressed as follows:
Brem
|Bknee − g II
|
|Bknee − Bm | 1+ µrec
tm
∆BR = = (3.14)
ksecR ksecR
So the rated electric load can be derived as:
tm

2p∆BR g I + µrec
K̂sR = (3.15)
µ0 Ds
Now, Equation 3.6 can be written by substituting the Equations 3.8-3.9 and after some manip-
ulations:
π dQ
s 4TR
Ds3 + 2(δ1 + δ3 + δ4 ) − Ltot Ds2 +

tan arcsin (1)
=0 (3.16)
2p πDs π B̂gR K̂sR (Ds )
The just found Equation 3.16 appears in the form:
f (Ds ) = 0
Indeed the left hand side is a non-linear function f (Ds ) in the variable Ds . It is relevant to
notice that in Equation 3.16 is underlined that inside f also the electric load is a function of Ds
(K̂sR (Ds )), according to what is expressed in Equation 3.15. However, all the other quantities
inside f are constant or they depend only on the parameters Qs , tm , Brem and Bknee . To
obtain suitable results, some constrains on the solution of Equation 3.16 have been applied.
First of all, looking at the argument of the trigonometric part of f , the function exists in a real
domain only if:
dQs
Ds >
π
Moreover a maximum size of 300 [mm] has been imposed for Ds in order to keep a minimum
margin of 100 [mm] for the stator laminations height and for the water-jacket thickness. Thus
the solution must lie in the range:
dQs
< Ds ≤ 300
π
Otherwise it is excluded.
Equation 3.16 can be solved with a Newton based root-finding algorithm. If applied to the
function f under analysis, these algorithms converges always, provided that proper starting
points are given. For simplicity reasons, in this project the secant method has been used, as it
®
does not require the analytical computation of the derivative of f (Ds ), although it needs two
starting points. The algorithm has been implemented in a Matlab script. Many trials are
computed changing the parameters from which the function f depends on: Brem and Bknee
are related to the PM grade; values of tm have been accepted until 12 [mm], above which it
has been supposed that the manufacturing of the PM becomes more complex; the number of
slots has been selected so that the slots per pole and per phase qs outcomes integer or integer
±0.5, i.e. 1 − 1.5 − 2 − 2.5− etc..., because it is known that such winding layouts reduce the
harmonic content of the mmf and the torque ripple. Furthermore machines with rated electric
loads out of the range 20 ÷ 50 [kA/m] are excluded, as for K̂sR > 50 [kA/m] the cooling could
become complicated even with a water-jacket, while for K̂sR < 20 [kA/m] the machine is too
44
1.15
1.1
1.05
0.95
0.9
0.85
8 8.5 9 9.5 10 10.5 11 11.5 12
Figure 3.10: Rated magnetic load as a function of magnet thickness on varying the PM grade.
Figure 3.11: Rated electric load as a function of magnet thickness and inner stator diameter on
varying the number of poles.
little electrically exploited and consequently not convenient.

According to Equation 3.13 the PM grade, i.e. Brem , affects heavily the magnetic load, which
instead depends less on the magnet thickness tm , as shown in Figure 3.10. On the contrary the
PM grade has no effects on the electric load, which depends greatly on tm , as represented by the
parametric surfaces in Figure 3.11. Indeed the chosen NdFeB magnets belong to the same class,
so their ’knee’ points are very close, resulting in an equal behavior against demagnetization.
Yet an increase in the magnet thickness allows to withstand higher demagnetizing current,
outcoming in more compact machines with greater electric loads. Finally, Figure 3.12 shows
the behavior of the sizing function f for 8 and 10 poles motors on varying the above mentioned
parameters. It can be seen that f does not present a root in the selected range of Ds for all
the selections of the parameters. In particular there are never satisfactory solutions if qs > 1.5,
for all the three numbers of poles. Besides it is interesting to notice that f has the dimension
of a volume ([dm3 ]). This happens maybe because it is related to displacement constrains.
The admissible outcomes from the secant algorithm within the related electric and magnetic
rated loads are reported in Table 2.4. The resulting main sizes have been rounded to the unit of
millimeters. All the accepted attempts present low machine ratios (Lstk /Ds = 0.2÷0.3), typical
feature of torque motors but a disadvantage for the efficiency. It can be seen that machines with
a higher number of poles withstand better the demagnetization, thus they allow higher electric
loads at equal PM thickness. Then, the use of high grade magnets makes the motors more
compact. Yet the parameter which affects more the inner stator diameter is the PM thickness.
45
10
-2
10
-2
50 100 150 200 250 300 50 100 150 200 250 300 50 100 150 200 250 300
Figure 3.12: Sizing function on varying all the parameters.
Indeed increasing tm higher electric loads are allowed, so the inner stator diameter decreases.
Table 2.4: Sizing trials.

(1)
PM grade Qs tm Ds Lstk K̂sR B̂gR V olm
[mm] [mm] [mm] [kA/m] [T ] [dm3 ]
(UH class) ≤ 12 ≤ 300 20 ÷ 50
2p = 6
11 285 79 26.2 0.91 0.497
N27 18
12 260 79 31.3 0.91 0.489
10 283 79 23.6 1.02 0.450
18 11 255 79 28.9 1.03 0.441
N35
12 232 79 34.6 1.04 0.432
27 12 292 63 27.6 1.04 0.438
10 262 79 25.5 1.10 0.415
18 11 236 79 31.2 1.11 0.406
N42 12 215 79 37.4 1.13 0.397
11 296 63 24.8 1.11 0.411
27
12 271 63 29.7 1.13 0.403
2p = 8
9 270 79 30.3 0.88 0.383
24 10 241 78 37.6 0.90 0.375
N27 11 218 78 45.7 0.91 0.367
11 280 61 35.6 0.91 0.373
36
46
12 258 60 42.0 0.91 0.366
8 276 79 25.9 0.99 0.350
24 9 242 78 33.2 1.00 0.341
10 217 78 41.3 1.02 0.333
N35
10 278 61 32.1 1.02 0.337
36 11 254 60 38.6 1.03 0.331
12 235 58 45.5 1.04 0.325
8 256 79 27.8 1.07 0.324
24 9 225 78 35.7 1.09 0.315
10 201 78 44.4 1.10 0.307
N42
9 289 61 27.8 1.09 0.318
10 261 60 34.3 1.10 0.311
36
11 239 59 41.2 1.11 0.304
12 222 57 48.4 1.13 0.299
2p = 10
7 289 78 27.6 0.85 0.320
30 8 250 78 36.4 0.87 0.312
N27 9 222 77 46.1 0.88 0.304
9 291 58 35.1 0.88 0.308
45
10 267 57 42.5 0.90 0.302
7 261 78 29.9 0.97 0.287
30
8 226 77 39.5 0.99 0.278
N35
8 297 59 30.1 0.99 0.281
45 9 268 57 37.5 1.00 0.274
10 247 55 45.2 1.02 0.268
6 289 78 23.1 1.01 0.276
30 7 244 77 32.1 1.04 0.266
8 211 76 42.3 1.07 0.257
N42
8 280 58 31.9 1.07 0.260
45 9 254 55 39.6 1.09 0.253
10 235 53 47.5 1.10 0.247
For each motor, the final choice between the different attempts has been based on cost-
effectiveness reasons. In a PM motor the magnets represent about the 70% of the overall
cost, so the volume of the arc-shaped PMs V olm [dm3 ] has been computed as follows for every
admissible machine:
Ds − 2g − tm e

V olm = 2αm[rad] tm Lstk 2p = 2αm Ds − 2g − tm tm Lstk (3.17)
2 [rad]
e
Where αm = αm /p is the mechanical covering angle of the PMs. Table 2.4 reports also these
quantities.
The volume of magnets depends heavily on the PM thickness and on the number of poles.
Machines with more poles and thicker magnets require less PM material. It can be see that
make the magnets thicker has advantages in terms of size, electric load and magnet usage.
Therefore, even if it may seem odd, increasing the PM thickness helps a lot in reducing the
machine cost. A threshold of 12 [mm] has been set to tm because it has been supposed that
over this value the production of the PM could become more expensive.
However, to assess properly which is the most cost-effective choice for every motor, the
specific cost (¿/dm3 ) of each magnet grade should be available, or at least the percentage
increment between the specific costs of the three NdFeB grades under analysis must be known.
47
3
4
7
8
Figure 3.13: Star of slot
Unfortunately NdFeB magnets price changes a lot between different manufacturers and valid
cost-related information have not been found for this project. Some reasonings on cost incre-
ments in relation to the magnet volume variations between the different attempts have been
made. Nevertheless, without reliable price specifications, the final choices have been based on
the supposition that the N42 grade leads to the most affordable machines. The selected values
after the preliminary sizing are underlined in Table 2.4. It is important to specify that such
sizes are not the final ones, but they could be modified a bit during the next steps of the design
procedure.
3.6 Winding design

All of the three motors selected after the preliminary sizing have a double-layer fractional-slot
winding. Their number of slots per pole and per phase results: qs = Qs /(3 · 2p) = 1.5. So the
star of slot is the same for the three chosen windings, that have only a different periodicity.
The star of slot is shown in Figure 3.13.
The same coil span of 4 slots has been chosen too, so that from an electrical point of view
the three windings outcome equal, i.e. they have the same fundamental winding coefficient kw
and they generate the same harmonic content in the emf and in the mmf. Indeed they have
a different slot mechanical angle αsm but the same slot electrical angle αse , according to the
following definitions:
360 ◦
αsm = [ ]
Qs
360
αse = p = 40 [◦ ]
Qs
Then the distribution and pitch factors have been obtained as:
αes
sin qs 2
kd = αes[rad] = 0.955
qs 2
βe
p
kp = cos = 0.985
2
48
Figure 3.14: MMF harmonic spectrum.
Figure 3.15: Windings layout under a pole pair.
Where βpe is the shortening electrical angle, that is half of the slot electrical angle due to the
chosen coil span:
αe
βpe = s = 20 [◦ ]
2
Fractional-slot windings are always short-pitched with a consequent improvement of the si-
nusoidal waveforms of the back emf. Finally the fundamental winding coefficient is defined
as:
kw = kd kp = 0.94
Fractional-slot windings has the advantages to yield to lower cogging torque and lower on-load
torque ripple, as compared to machines with integer-slot windings at equal rated specifications.
On the other hand, a balanced three-phase current flowing in an integer-slot winding produces
a mmf with only odd harmonics not multiple of three, while the spectrum of the three-phase
mmf generated by a fractional-slot winding could exhibits also unwanted even harmonics when
some particular combinations of Qs and p are made, as it happens for all the three selected
machines. The harmonic content of the three-phase mmf has been computed with the open
source software DOLOMITES and it is reported in Figure 3.14, while the actual layout of the
windings under a pole pair is outlined in Figure 3.15. The complete schema for each motor can
be obtained by repeating the sequence of Figure 3.15 a number of times equal to the periodicity
of the winding, i.e. 5 for the 10 pole machine, 4 for the 8 pole winding, 3 for the 6 pole one.
3.6.1 Voltage specifications and number of turns choice

In each winding the maximum feasible number of parallel connections npp has been adopted.
The parallel connections are made between coils under subsequent pole pairs, so that the max-
imum feasible npp basically corresponds to the periodicity of the winding. The implementation
of parallel connections allows to use conductors with a smaller section at equal current spec-
ifications. Despite the copper weight does not change, as the conductor size lowers but the
number of actual turns in slots increases proportionally, the coils become easier to wind up and
the slot openings needed to insert the conductors can be reduced, with a consequent benefit in
49
the decrease of the magnetic reluctance of the air gap.
However, in the design and analysis of electrical machines it is more handy to refer to the
equivalent series winding, which is the winding that generates the same mmf and the same emf
without parallel connections, i.e. with all the coils in series. The number of series equivalent
turns per phase Ns depends on the speed that has to be achieved, i.e. on the feeding frequency,
and on the wanted back emf amplitude at such speed. SPM machines are designed looking at
the base point specifications. The reference speed is the base speed, which should correspond
to the rated speed nR , because these motors commonly are always driven in MTPA region.
Thus the rated frequency fR is taken as reference frequency and the back emf should be the
one produced at rated load ÊR . But, if the machine is not so saturated at rated load, as it
should be expected, the back emf is quite similar to the emf generated in unload condition Ê0 .
Hence it can be assumed that Êr ∼ = Ê0 . Then, supposing a sinusoidal waveform for the emf,
the number of equivalent series turns per phase can be computed as follows:
√
2E0
Ns = (1)
(3.18)
πkw ΦpR fR
(1)
Where E0 [V ] is the rms value of the emf and ΦpR [W b] is the fundamental flux per pole at
rated load, which has been obtained with the following relation:
(1) (1) Ds Lstk

ΦpR = B̂gR (3.19)
p
The emf rms value is related to the machine voltage. Thus some voltage specifications must
be chosen. A DC bus voltage of Vdc = 700 [V ] has been assumed. Such value can be achieved
by arranging the battery pack with strings of 200 series-connected 3.5 V lithium-ion cells.
Assuming that the three-phase inverter synthesizes the sinusoidal waveforms through the SVM
(Space Vector Modulation) technique, the DC bus voltage is the peak value of the maximum
sinusoidal line-to-line voltage obtainable. The motor phases are star-connected without neutral
wire to avoid the flow of zero-sequence current. So, the rms value of maximum sinusoidal
winding voltage VwindM AX has been derived as follows:
Vdc Vdc
VwindM AX = √ √ = √ ∼ = 285 [V ]
2 3 6
A margin of 5 [V ] has been taken in favor of the control system to consider variation in the DC
bus voltage: VwindM AX = 280 [V ]. Then, to obtain the emf rms value a 5% voltage drop has
been supposed due to the winding impedance:
E0 = Vwind (1 − 0.05)
For the design purposes, the rated winding voltage Vwind has been set quite below the maximum
value to prevent the inverter to work in overmodulation in overload condition, when a higher
voltage drop is expected due to the higher feeding current: Vwind = 204 [V ] for the 10 and 8
pole motors and Vwind = 214 [V ] for the 6 pole machine.
After that Ns was computed with Equation 3.18, the number of in-slot series equivalent
conductors ncs and the number of actual in-slot conductors nc has been obtained as follows:
3Ns
ncs =
Qs
3Ns
nc = ncs npp = npp
Qs
nc must be an even integer to allow the realization of a symmetric double-layer winding. If
this not happens, nc must be approximate to a close even integer. As a consequence Ns and
ncs need to be recomputed through the previous relations as well as the fundamental flux per
pole, by inverting Equation 3.18. Then, if needed, some quantities among the stack length Lstk ,
50
(1)
the magnetic rated load B̂gR and inner stator diameter Ds must be also adjusted according to
Equation 3.19. In this project slight modifications of the stack length has been chosen to gain
on torque capability.
The quantities presented in this section are summarized in Table 3.5 for each motor with
all the other winding parameters and relevant sizes. It can be seen that the stack lengths have
been slightly increase if compared to the ones in Table 2.4.
3.6.2 Current specification and conductor sizing

To choose a suitable conductor size the current specifications must be imposed. Indeed the
conductor section Sc [mm2 ] is related to the rated current IR [A] and to the rated current
density JR [A/mm2 ] by the following equation.
Sceq IR
Sc = = (3.20)
npp JR npp
Where Sceq is the section of the equivalent series conductor.

The current density JR [A/mm2 ] is imposed by the designer. Its choice is made by looking
to similar machines and by experience. In particular it depends on the cooling system of the
motor and on the thermal class of the adopted insulating materials. According to Equation
3.20, higher current density allow smaller conductor sizes at equal rated current. Thus the
slots need less cross section area and the machine results more compact. But to achieve higher
current density without compromising the in-slot insulation more complex cooling systems are
needed because the specific losses increase a lot. In this project a H thermal class has been
chosen for the insulation in order to withstand slot temperatures up to 180◦ C while the cooling
system consists in a water-jacket with a circumferential duct. So, following the advice given in
[33], the conductors have been sized for a current density JR = 9 [A/mm2 ].
The rated current instead depends on the chosen winding voltage. Indeed the motor behaves
actually as an electric load powered by a current controlled voltage source, i.e. the battery pack
and the inverter, although sometimes it could be useful to think the electric drive as a current
source. So, assuming three-phase balanced sinusoidal feeding current and symmetric voltages,
the power balance applied to the machine relates the rated current to the winding voltage as
follows:
PR
IR = (3.21)
3 cos ϕR ηR Vwind
Where the fundamental power factor and the efficiency at rated load has been supposed as
cos ϕR = 0.95 and ηR = 0.90.
Finally, an effective commercial section close to the one computed with Equation 3.20 has
been selected from commercial enameled copper wires. The actual current density has been
recomputed inverting Equation 3.20 in the end.
The quantities presented in this section are summarized in Table 3.5 for each motor with all
the other winding parameters and relevant sizes. In the table the maximum overload current
is IOL = 4 × IR .
3.7 Stator sizing

The next step in the electromagnetic design consists in deciding how to assembly and cut the
stator laminations. Firstly the soft magnetic material type must be chosen. A good compromise
between the specific losses (p [W/kg]) due to the hysteresis and the commercial availability is a
M350 lamination. Then 0.5 [mm] gauge electrical steel sheets are more than sufficient to limit
the eddy current at the relatively low speed of the application under analysis. So a M350-50A
lamination has been adopted for all of the three motors. The DC magnetization curve and the
specific losses characteristic at 50 [Hz] are reported respectively in Figures 3.16-3.17.
51
Parameter Symbol Unit Values
Number of poles 2p / 6 8 10
Number of slots Qs / 27 36 45
Slot mechanical angle αsm [◦ ] 13.3 10 8
Slot electrical angle αse [◦ ] 40 40 40
Winding factor kw / 0.94 0.94 0.94
Slot pitch ps [mm] 31.5 19.4 16.4
Pole pitch τp [mm] 141.9 87.2 73.8
Coil pitch τp0 [mm] 126 77.6 65.6
Parallel connections npp / 3 4 5
Rated frequency fR [Hz] 115 153 192
Winding voltage (rms) Vwind [V ] 214 204 204
Rated current (rms) IR [A] 53 55 55
Overload current (rms) IOL [A] 212 220 220
Inner stator diameter Ds [mm] 271 222 235
Stack length Lstk [mm] 64 58 54
Series turns per phase Ns / 126 168 174
In-slot series conductors ncs / 14 14 11.6
In-slot conductors nc / 42 56 58
Conductor section Sc [mm2 ] 2.011 1.539 1.227
Wire diameter dc [mm] 1.60 1.40 1.25
Current density (rms) JR [A/mm2 ] 8.8 8.9 9.0
Table 3.5: Winding parameters and relevant quantities.
52
M350-50A
DC magnetization curve 10 3
3 5
2.95
2.9
2.85
2.8
2.75
2.7 4.5
2.65
2.6
2.55
2.5
2.45
2.4 4
2.35
2.3
2.25
2.2
2.15
2.1 3.5
2.05
2
1.95
1.9
1.85
1.8 3
1.75
1.7
1.65
1.6
1.55
1.5 2.5
1.45
1.4
1.35
1.3
1.25
1.2 2
1.15
1.1
1.05
1
0.95
0.9 1.5
0.85
0.8
0.75
0.7
0.65
0.6 1
0.55
0.5
0.45
0.4
0.35
0.3 0.5
0.25
0.2
0.15
0.1
0.05
0 0
10 2 10 3 10 4 10 5
Figure 3.16: Lamination DC magnetization curve.
M350-50A
Specific losses
5
4.9
4.8
4.7
4.6
4.5
4.4
4.3
4.2
4.1
4
3.9
3.8
3.7
3.6
3.5
3.4
3.3
3.2
3.1
3
2.9
2.8
2.7
2.6
2.5
2.4
2.3
2.2
2.1
2
1.9
1.8
1.7
1.6
1.5
1.4
1.3
1.2
1.1
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8
Figure 3.17: Lamination specific losses.
53
hbi
wse
wt De
I
hsi Sslot
hs
ws
hwed II
Sslot
hso wso Ds
Figure 3.18: Slot geometry outline.
The actual axial length of the magnetic material Lf e [mm] results a bit lower than the stack
length due to the electrical insulation between the laminations:
Lf e = kpack Lstk
The packing coefficient has been set to kpack = 0.96 considering that the insulating purpose
is obtained through the oxidation of the electrical steel sheets surfaces. Hence the number
of laminations needed to assemble the stack can be easily found out as Nf e = Lf e /0.5. The
laminations must be cut to realize the slots, the teeth and the back iron outline. The chosen
geometry is drawn in Figure 3.18 with all the relevant dimensions, some of which have been
linearized: it considers trapezoidal-shape slots and rectangular-shape teeth.
The first size to set is the tooth width wt . To do so the wanted flux density in the tooth
B̂tR at rated load must be chosen. It is expected that the teeth will be the iron parts where the
flux density will reach the higher value. So, the choice made for this project is B̂tR = 1.6 [T ]
for all of the three motors, in order to have a margin of 0.4 [T ] to the saturation value, which
is about 2 [T ], considering that a large overload capability has to be fulfilled. Hence, supposing
no flux leakages, the tooth width has been computed using the flux balance between the tooth
and the air gap, namely the Gauss law for the magnetic field:
B̂gM AX ps
wt =
B̂tR kpack
The maximum flux density value at the air gap B̂gM AX has been obtained as:
B̂gM AX = B̂g + ∆BR
Where ∆BR is the flux density variation due to the rated electric load calculated through
Equation 3.14 or Equation 3.15. The sizing procedure continues with the definition of the slot
geometry dimensions. The slot opening height hso and the wedge height hwed must be set by
the designer. The choices made during this project are: hso = 0.6 [mm] and hwed = 1 [mm] for
all of the three machines. The slot opening width wso [mm] is a designer choice too. A wise
decision is a size double of the wire diameter wso = 2dc . In this way the insertion of the wires
inside the slots becomes easier and consequently the winding manufacturing is simplified and
speed up. At the same time the magnetic reluctance increase of the air gap is not so dramatic.
At this point the width ws [mm] has been obtained as follows:
π(Ds + 2hso + 2hwed )

ws = − wt
Qs
54
Then the slot cross section area Sslot [mm2 ] has been computed from the conductor effective
section Sc considering a fill factor kf ill = 0.4:
I II Sc nc
Sslot = Sslot + Sslot =
kf ill
II
Sslot is made up by the section occupied by the wedge Sslot and the area earmarked to the
I
conductors Sslot :
I II
Sslot = Sslot + Sslot
II
The section Sslot [mm2 ] has been calculated as:
II ws + wso
Sslot = hwed
2
I
So also the area Sslot [mm2 ] has been computed:
I II
Sslot = Sslot − Sslot
Yet such area can be expressed as:
I ws + wse
Sslot = hsi (3.22)
2
Where the width wse can be further written as:
π(Ds + 2hso + 2hwed + 2hsi )

wse = − wt (3.23)
Qs
After some manipulations, the substitution of the previous 3.23 into the 3.22 yields to the
following quadratic equation in the only unknown hsi :
π 2 hw
se π(Ds + 2hso + 2hwed ) wt i I
hsi + + − hsi − Sslot =0 (3.24)
Qs 2 2Qs 2
A feasible height hsi [mm] has been selected solving the Equation 3.24. Finally the width
wse [mm] has been computed through Equation 3.23 whereas the total height of the slot hs [mm]
has been calculated as:
hs = hso + hwed + hsi
hs has been oversized a bit to consider the loss of slot area due to corners rounding. The last
dimension that has to be find out is the back iron height hbi [mm]. It is computed using the
Gauss law in the same way of the tooth sizing. In this case the flux balance considers the flux
in the air gap on the PM surface Φp [W b] and the maximum flux in the back iron Φbi [W b]:
Φp 1 πDs
Φbi = = B̂g Lstk
2 3 2p
The flux density in the air gap B̂g has been computed by inverting Equation 3.12, according to
the hypothesis of a square-wave distribution. Then hbi has been calculated as follows:
Φbi
hbi =
B̂bi Lf e
Where the flux density maximum value in the back iron has been set to: B̂bi = 1.3 [T ].
In the end the laminations outer diameter De [mm] has been obtained as:
De = Ds + 2hs + 2hbi
The relevant geometric size and materials data are reported in Table 3.6 for all the three
55
Parameter Symbol Unit Values
Inner stator diameter Ds [mm] 271 222 235
Stack length Lstk [mm] 64 58 54
Air gap thickness g [mm] 1 1 1
Permanent magnet N42 UH
m ◦
PM covering angle 2αm [ ] 20 30 24
PM thickness tm [mm] 12 12 10
Laminations M350-50A
Laminations number Nf e / 123 112 104
Outer diamter De [mm] 394 330 332
Teeth width wt [mm] 24.3 15.0 12.4
Slot height hs [mm] 22.2 30.8 29.6
Slot ext. width wse [mm] 12.4 9.8 8.2
Slot area Sslot [mm2 ] 212 215 178
Wedge height hwed [mm] 1 1 1
Opening height hso [mm] 0.6 0.6 0.6
Opening width wso [mm] 3.2 2.8 2.5
Table 3.6: Relevant geometric size and materials data.
machines.
3.8 Rotor and shaft sizing

For the purposes of the electromagnetic analysis the shaft could be neglected supposing that
it should have a small diameter as compared to the rotor yoke or that it could be made by a
ferromagnetic material. However, to present a more complete view of the project, a simplified
but effective sizing of the shaft has been done. The shaft is very short (LtotM AX = 190 [mm])
and the mechanical load are concentrated near the supports, i.e. the bearings in the front
and rear plates. So any flexural strains can be neglected and only the torsional stress can be
considered. The maximum torque that must be withstood is equal to the peak value of the
mechanical characteristic of the original ICE, which is declared as TM AX = 1350 [N m]. Indeed
the hybrid powertrain should exhibit the same performances of the traditional vehicle. So the
shaft diameter results the same for all the three motors and it has been computed with the
following relation. s
5TM AX[N mm]
DSHAF T =
3
∼
= 60 [mm]
τm
Where the admissible mechanical strain has been set to τm = 35 [N/mm2 ] considering a solid
C40 stainless steel shaft connected to a 6-cylinder 4-stroke Diesel engine.
The rotor yoke is laminated in the same way of the stator in order to limit the eddy current
during the electromagnetic transients. The chosen electrical steel is also the same used for
the stator: M350-50A. Yet in the rotor yoke it has been decided to shear off some triangular-
shaped holes, centered below each magnet. The holes shape has been designed to not obstruct
the magnetic flux generated by the PMs. Such choice has three advantages.
1. The weight and the inertia of the rotor decrease with a further improvement of the elec-
tromechanical responsiveness of the machine and a lightening of the mechanical load on
the bearings. Moreover it is important to remind that the EM behaves as a mechanical
load for the ICE even when it does not work. So a reduced inertia improve the kinematics
56
Jr [kg m2 ]
2p 6 8 10
Without holes 0.230 0.096 0.111
With holes 0.182 0.072 0.085
Table 3.7: Rotor inertia.
of the entire powertrain too.

The inertia Jr [kg m2 ] can be expressed as a mass integral or with a corresponding volume
integral: Z Z Z
Jr = r2 dm = (x2 + y 2 )dm = (x2 + y 2 )γdv (3.25)
M M V ol
p
Where γ [kg/m3 ] is the density of the material and r = x2 + y 2 is the element distance
from the rotation axis. In the machines rotor three materials contributes to the inertia:
the iron (γf e = 7800 kg/m3 ), the PMs (γpm = 7500 kg/m3 ) and the magnets fiberglass
bandage (γband = 1800 kg/m3 ). So Equation 3.25 becomes:
Z Z Z
2 2 2 2
J r = γf e (xf e + yf e )dv + γpm (xpm + ypm )dv + γband (x2band + yband
2
)dv
V olf e V olpm V olband
The volume integrals have been automatically computed with FEMM. Table 3.7 reports
the rotor inertia of the three motors with and without holes. It can be seen that the
introduction of holes decreases the inertia of about 20 ÷ 25%
2. The particular triangular shape produces high flux density regions near the magnets, thus
it opposes variation of flux in the PMs and consequently it helps in reducing the eddy
current losses, as explained in [39].
3. Despite the machines are not equipped with a fan and they have a horizontal shaft, the
holes provide additional cooling inside the rotor yoke through free convection because the
laminations stacks are very short and the vortexes generated near the rotating surfaces
can expand inside the air ducts.
The technical drawings of the three sized motors are reported in Appendix 5.3.
3.9 Volume limits fulfillment and materials cost compu-

tation
At this point the electromechanical structure of the three motors is completely defined. So it
can be verified how well the volume constrains are fulfilled, according to the statements made
during the preliminary sizing (Section 3.5). As far as the radial dimension limit concerns,
looking at Table 3.6, the 8 and 10 pole machines have an external stator diameter quite below
400[mm], with a margin of about 70[mm] that is enough for the water-jacket thickness, whereas
the 6 pole motor leaves only 6 [mm] for the cooling duct and surely the total outer size will be
a bit greater than the limit. Instead the total axial length Ltot can be estimated according to
Figure 3.9. The dimensions δ1 , δ3 , d and the thickness of the front and rear plates δ4 remain
the same that have been chosen during the preliminary sizing. Thus only the length δ2 has to
be computed through Equation 3.7, where the coil span τp0 [mm] of all the three machines is
expressed as:
τp0 = 4 · ps
Then Ltot can be finally calculated with Equation 3.6. Table 3.8 reports the results and the
chosen end-winding sizes for the three motors. It can be seen that the total axial length is
lower than 190 [mm]. The available volume is wisely exploited and at the same time all the size
57
2p / 6 8 10
De [mm] 394 330 332
Lstk [mm] 64 58 54
δ1 [mm] 10 10 10
δ3 [mm] 10 10 10
δ4 [mm] 20 20 20
d [mm] 10 10 10
ps [mm] 31.5 19.4 16.4
δ2 [mm] 21 23.5 25.2
Ltot [mm] 186 185 184
Table 3.8: Radial and axial sizes.
Material γ [Kg/m3 ] c [¿/kg]

Iron 7800 2
Copper 7900 7
NbFeB 7500 70
Table 3.9: Materials density and specific costs.
constrains are fulfilled. Thus the preliminary sizing method presented in Section 3.5 has been
validated.
Furthermore a comparison between the material costs of the three motors has been done. To
do so the volumes and weights of the various machines’ parts have been estimated with analytical
expressions or computed through the software FEMM. The materials density γ [kg/m3 ] and
specific costs [¿/kg] used for the computations are reported in Table 3.9, where ’iron’ means
electrical steel laminations. Slots and winding insulation as well as the PMs fiberglass bandage
are neglected.
The iron total volume V olf e , and so the iron total weight Gf e , has been divided into three
components:
V olf e = V olf et + V olf ebi + V olf erot
Gf e = Gf et + Gf ebi + Gf erot
The three contributions have been obtained as follows:
Stator teeth:
Gf et = γf e V olf et = γf e Qs hs wt Lf e
Stator back iron:

Gf ebi = γf e V olf ebi = γf e π(De − hbi )hbi Lf e
Where the back iron height can be calculated as
De − (Ds + 2hs )
hbi =
2
Rotor yoke Gf erot = γf e V olf erot : the volume has been computed with the FEMM because
of the triangular-shape holes. The results have been adjusted with the stacking coefficient
kpack = 0.96 to consider the insulation between the iron sheets.
The magnets weight Gpm = γpm V olpm has been obtained using the volumes calculated through
Equation 3.17 whereas the copper weight has been estimated only analytically with the following
expression:
GCu = γCu V olCu = γCu Qs nc Sc lc
58
Table 3.10: Materials volumes, weights and costs.
2p / 6 8 10
3
V olf et [dm ] 0.89 0.93 0.86
V olf ebi [dm3 ] 2.70 1.20 0.96
V olf erot [dm3 ] 1.70 0.96 1.00
V olf e [dm3 ] 5.30 3.10 2.80
Gf et [kg] 7 7.2 6.7
Gf ebi [kg] 21 9.7 7.5
Gf erot [kg] 13 7.5 7.8
Gf e [kg] 41 24.4 22
Iron cost ¿ 82 48.8 44
V olpm [dm3 ] 0.40 0.30 0.25
Gpm [kg] 3.0 2.3 1.9
PMs cost ¿ 210 161 133
3
V olCu [dm ] 0.50 0.54 0.52
GCu [kg] 4.0 4.3 4.1
Copper cost ¿ 28 30 28.7
Overral cost ¿ 320 240 206
The half turn length lc has been computed as follows:
lc = Lstk + lew
Where lew is the length of the end-turns, which can be calculated according to Figure 3.9:
2δ2
lew = 2δ1 +
sin αew
The outcomes are reported in Table 3.10. The 10 pole machine is the cheapest one and, at a first
glance, it may seem the most worthwhile too. But, considering the costs and the performances
of the entire electric drive, the things could change. Indeed the 10 and 8 pole motors require
higher current and feeding frequency, so that the inverter power switches and the battery pack
must withstand more trying working conditions. Efficiency is also a key point for the application
under analysis. Therefore a proper comparison between the three designed machines can be
done only after a detailed performances and losses analysis. Moreover the cooling system and
the thermal behavior should be taken into account too.
59
Chapter 4
Electromagnetic analysis
In this chapter the electromagnetic analysis of the three designed machines will be presented.
The three EMs have been studied using both analytical relations and finite element analysis
(FEA) in order to assess the machine performances and validate the hypothesis made during
the sizing process. The machines electromagnetic behavior in unload condition, with the rated
current value and with the maximum overload current value, has been analyzed and simulated
®
through magnetostatic FEA. The significant parameters of the machines have been computed
too. FEMM and MATLAB softwares have been used for this purpose. The results of such
analysis will be presented and exhaustively discussed so that some relevant comparisons between
the computations and the differences with the initial suppositions will arise. The performances
between the three designed EMs will also be compared. Finally the machine losses are mapped
over the admissible working points in order to provide the necessary information to do a proper
thermal design and to analyze the powertrain efficiency.
4.1 Performances analysis in unload condition

The flux linkage and the emf generated by the machines while they are rotating with open
circuit, i.e. with no feeding current, can be estimated considering the magnetic circuit shown
in Figure 4.1 [24].
It is supposed that the mmf ([Aturns]) produced by the magnet pushes the magnetic flux
through three reluctance: the air gap, the tooth and the back iron. The path in the rotor yoke
is neglected supposing that its reluctance is much lower if compared to the tooth or the back
iron. A further analysis with finite elements will shown that such hypothesis could not be done
for these machines, because of the particular geometry of their magnetic circuits. The PM is
represented as flux source with a parallel reluctance, which depends on the magnet thickness
and differential permeability. So the mmf drop on the magnetic reluctances must be balanced
entirely by the magnet (there are no feeding current).
The flux density in the air gap B̂g0 , which corresponds to the flux density value of the PM’s
working point Bm , has been computed using Equation 3.10. The magnetic equivalent air gap
g II is defined as:
g II = ksat0 g I = ksat0 kcarter g
The saturation coefficient in unload condition ksat0 must be supposed. At the end of the
magnetic circuit computations the value of ksat0 will be verified: if it will be too far from the
®
supposed one the circuit must be solved again trying a different coefficient. Thus, such analytical
method consists in an iterative procedure, which has been implemented in a MATLAB script.
The Carter coefficient kcarter , that takes into account the reluctance increase due to slots
openings, has been estimated with the following simplified expression:
ps
kcarter =
ps + g − 0.75 · wso
Then the magnetic field value of the PM’s working point in unload condition can be also
60
Rbi ∝ lbi /µrel
Rt ∝ hs /µrel
Rg ∝ g/µ0
Φp0 Rm ∝ tm /µrec
Figure 4.1: Magnetic circuit.
calculated as:
Bm − Brem
Hm0 =
µrec µ0
While the magnetic flux under a pole pitch can be estimated as:
2 πDs
Φp0 = B̂g0 Lstk
3 2p
Considering that the PM cover 2/3 of the pole pitch.

From the flux density value B̂g0 the mmf drop in the air gap has been computed as follows:
B̂g0
Ĥg0 g I = gkcarter
µ0
Then, from the magnetic flux balance between the air gap and the tooth (Gauss law), the flux
density in the tooth has been obtained:
B̂g0 ps
B̂t0 =
wt kpack
So, the corresponding value of magnetic field Ĥt0 has been found out with the help of the
lamination DC magnetization curve in Figure 3.16. As a consequence the mmf drop along the
tooth can be computed: Ĥt0 hs .
It can be supposed that the magnetic flux in the air gap slits in half when it reaches the back
iron, so that the maximum flux in the back iron is half the flux in the air gap: Φbi = Φp0 /2.
Thus the maximum flux density in the back iron has been obtained as:
Φbi
B̂bi =
hbi Lf e
61
The flux density is not constant along the back iron because each tooth adds up its flux contri-
bution. So a detailed computation would be done by dividing the path in segments. However
a fast estimation based on similar machines can be done. The mean flux density in the back
iron has been estimate from the maximum value:
∗
B̂bi = 0.90 · B̂bi
∗
Then the corresponding magnetic field value Ĥbi has been found out using the lamination BH
∗
curve and the mmf drop on the back iron ha been calculated as: Ĥbi lbi , where the back iron
length under a pole has been computed as follows:
π(De − hbi )
lbi =
2 · 2p
In the end the actual saturation coefficient has been obtained using its definition:
∗
Ĥg0 g I + Ĥt hs + Ĥbi
P
Ĥl lbi
ksat0 = =
Ĥg0 g I Ĥg0 g I
After that ksat0 outcomes different no more than 2 ÷ 5% of the value supposed at the beginning
of the procedure, the flux linkage and the emf generated by the PM can be computed. At first
the fundamental flux density at the air gap, namely the peak of the first harmonic, have been
obtained as in Equation 3.12, supposing a square-wave distribution and considering that the
PM covers 120 electrical degrees. Then the corresponding fundamental flux has been computed
as follows:
(1) (1) Ds Lstk
Φp0 = B̂g0
p
Hence the peak value of the flux linkage Λ̂0 and the peak value of the emf Ê0 have been derived
with the following expressions:
Ns (1)
Λ̂0 = kw Φ
2 p0
Ê0 = ω Λ̂0 = 2πf Λ̂
The analytical computations just presented have been compared with an unload magnetostatic
FEA performed using the software FEMM for each motor. Only a pole pair has been analyzed
for every machine, taking advantage of the periodic boundary conditions (BCs) provided by
the software. Furthermore a sliding band has been set inside the air gap in order to simulate
rotations without the need of remeshing for every rotor position. Indeed steady-state rotations
have been also simulated with a series of magnetostatic FEA, changing the rotor position at
each shot. The initial position θe = 0◦ is achieved when the magnetization direction, namely
the d-axis, is aligned with the magnetic axis of the phase a [40].
The preprocessing and mesh layout of the 6 pole machine is reported in Figure 4.2 as
example. Relatively coarse meshes have been adopted in order to reduce the computational cost
without affect too much the precision. The 6 pole motor, which is the one that has required
more mesh elements, has been discretized using a mesh with 7489 nodes. The permanent
magnets properties has been set at the supposed working temperature of 120◦ C, according to
the parameters in Table 3.3, with a parallel magnetization.
The unload flux maps of the three machines are shown in Figures 4.3-4.4-4.5, while, as
example, Figure 4.6 reports the 6 pole machine flux density distribution in the air gap with
its harmonic content until the 11th order. It can be seen that the square waveform of the flux
density distribution is affected by the concentrated reluctances of the slot openings, thus the
peak value of the first harmonic, i.e. the magnetic load, results actually a bit smaller than the
expected one, which is the quantity obtained through Equation 3.12. Moreover from the flux
maps it can be stated that all the three machines present flux density values in the iron paths
lower than the ones analytically calculated with the magnetic circuit, but this not necessary
means that the they are less saturated. Indeed the contribution of the rotor yoke cannot be
62
Cond_7 Cond_6
Cond_8
Cond_5
Cond_9 NdFeB N42 UH
M350-50A
Cond_4
Cond_3
NdFeB N42 UH
air Cond_2
M350-50A
air
air Cond_1
air
Figure 4.2: 6 poles machine: preprocessing settings and mesh.
1.197e+000 : >1.260e+000
1.134e+000 : 1.197e+000
1.071e+000 : 1.134e+000
1.008e+000 : 1.071e+000
9.450e-001 : 1.008e+000
8.820e-001 : 9.450e-001
8.190e-001 : 8.820e-001
7.560e-001 : 8.190e-001
6.930e-001 : 7.560e-001
6.300e-001 : 6.930e-001
5.670e-001 : 6.300e-001
5.040e-001 : 5.670e-001
4.410e-001 : 5.040e-001
3.780e-001 : 4.410e-001
3.150e-001 : 3.780e-001
2.520e-001 : 3.150e-001
1.890e-001 : 2.520e-001
1.260e-001 : 1.890e-001
6.300e-002 : 1.260e-001
<0.000e+000 : 6.300e-002
Density Plot: |B|, Tesla
Figure 4.3: 6 pole machine: flux map in unload condition.
63
1.235e+000 : >1.300e+000
1.170e+000 : 1.235e+000
1.105e+000 : 1.170e+000
1.040e+000 : 1.105e+000
9.750e-001 : 1.040e+000
9.100e-001 : 9.750e-001
8.450e-001 : 9.100e-001
7.800e-001 : 8.450e-001
7.150e-001 : 7.800e-001
6.500e-001 : 7.150e-001
5.850e-001 : 6.500e-001
5.200e-001 : 5.850e-001
4.550e-001 : 5.200e-001
3.900e-001 : 4.550e-001
3.250e-001 : 3.900e-001
2.600e-001 : 3.250e-001
1.950e-001 : 2.600e-001
1.300e-001 : 1.950e-001
6.500e-002 : 1.300e-001
<0.000e+000 : 6.500e-002
8.835e-001 : >9.300e-001
8.370e-001 : 8.835e-001
7.905e-001 : 8.370e-001
7.440e-001 : 7.905e-001
6.975e-001 : 7.440e-001
6.510e-001 : 6.975e-001
6.045e-001 : 6.510e-001
5.580e-001 : 6.045e-001
5.115e-001 : 5.580e-001
4.650e-001 : 5.115e-001
4.185e-001 : 4.650e-001
3.720e-001 : 4.185e-001
3.255e-001 : 3.720e-001
2.790e-001 : 3.255e-001
2.325e-001 : 2.790e-001
1.860e-001 : 2.325e-001
1.395e-001 : 1.860e-001
9.300e-002 : 1.395e-001
4.650e-002 : 9.300e-002
<0.000e+000 : 4.650e-002
1.5 1
0.9
1
0.8
0.7
0.5
0.6
0 0.5
0.4
-0.5
0.3
0.2
-1
0.1
-1.5 0
0 90 180 270 360 1 2 3 4 5 6 7 8 9 10 11
Figure 4.6: 6 pole machine: air gap flux density distribution in unload condition.
64
neglected because the triangular-shaped holes causes flux density values comparable with the
ones in the tooth and in the back iron.
The mmf drops in each part of the magnetic circuit, rotor yoke included, have been com-
puted with FEMM along the path of a magnetic flux line using the general definition of mmf
drop: Z
H̄ · t̄ dl
l
Also the magnetic flux under a pole pitch has been obtained from the post-processing interface
of FEMM with the general definition of magnetic flux through a surface:
Z
Φ= B̄ · n̄ dS
S
To derive the flux linkage and the emf an unload rotation has been simulated. At each rotor
position the integral of the vector potential Ā over every slot cross section area has been
collected. Then the flux linkage with the generic phase x (x = a, b, c) has been computed with
the following relation:
Ā
λx = pncs kx Lstk
Sslot
The expression is multiplied by p because only a pole pair has been analyzed. kx is the row of
the slot matrix corresponding to phase x. The slot matrix, which is the same for all the three
motors, is reported below. It represents the winding layout shown in Figure 3.15.
Qs 1 2 3 4 5 6 7 8 9
ka 1 0 0 0 −1 −0.5 0 0 +0.5
kb 0 0 +0.5 1 0 0 0 −1 −0.5
kc 0 −1 −0.5 0 0 +0.5 1 0 0
The emf on phase the generic phase x is defined according to the Faraday-Neumann induction
law:
∂λx ∂λx
ex = − = −ω
∂t ∂θe
The continuum derivative could be easily discretize with its incremental ratio. But, from a
computational point of view, this could be a risky operation. Indeed rapidly changing signals
with a small incremental step can lead to wrong and noisy outcomes. So, the emf signals have
®
been derived from the Fourier series of the flux linkage, whose coefficients have been computed
through the FFT function (Fast Fourier Transform) already implemented in MATLAB . As
example, Figures 4.7-4.8 show the 6 pole machine flux linkage and emfs, with harmonic spectrum
until the 11th order. The results for all the three motors both from the analytical computations
and the FEA are reported in Table 4.1.
It can be seen that the fractional slot winding layout produces even harmonics into the emfs.
Nevertheless the sinusoidal waveforms of both the flux linkage and the emfs are satisfying.
Furthermore there is a good match between the peak values computed with the analytical
method and the FEA. On the contrary the mmf drops obtained with FEA outcomes differently
distributed along the magnetic flux lines than what is expected from the analytical solution
of the magnetic circuit. This is due to the non-negligible rotor yoke contribution. Moreover
the flux per pole are also very far from the expected values, maybe because of the actual flux
density distribution in the air gap. However the saturation coefficients results very similar, so,
from the overall performances point of view, the analytic method can be considered satisfactory.
4.1.1 Cogging torque

In an unload rotation the machines do not develop any mean torque because no current flows
into the winding. But variations of magnetic coenergy occur due to the reluctance changes close
to the slot openings. Physically, the magnets are subjected to cogging. The resulting cogging
torque has zero-mean because it is generated only by a coenergy variation [40]. Nevertheless it
65
0.4
0.3
0.2
0.1
-0.1
-0.2
-0.3
-0.4
0 90 180 270 360
Figure 4.7: 6 poles machine: flux linkage in unload condition.
300 1
0.9
200
0.8
0.7
100
0.6
0 0.5
0.4
-100
0.3
0.2
-200
0.1
-300 0
0 90 180 270 360 1 2 3 4 5 6 7 8 9 10 11
Figure 4.8: 6 poles machine: emfs in unload condition.
Table 4.1: Magnetic circuit analysis in unload condition

6 poles 8 poles 10 poles
Quantity Symbol Unit
Analytic FEA Analytic FEA Analytic FEA
I
Air gap mmf drop Ĥg0 g [Aturns] 869 820 877 792 860 806
Tooth mmf drop Ĥt0 hs [Aturns] 16 9 21 12 22 13
∗
Back iron mmf drop Ĥbi lbi [Aturns] 22 15 17 10 17 9
Rotor yoke mmf drop Ĥr0 lr [Aturns] / 18 / 8 / 6
Saturation coeff. ksat0 [/] 1.04 1.05 1.04 1.04 1.04 1.03
Flux per pole Φp0 [mW b] 6.3 5.9 3.5 3.3 2.7 2.2
Flux linkage peak Λ̂0 [V s] 0.39 0.37 0.29 0.27 0.23 0.22
Emf peak Ê0 [V ] 281 267 279 260 277 267
66
4
-1
-2
-3
-4
0 2 4 6 8 10 12 14 16 18 20
170.64
170.62
170.6
170.58
170.56
170.54
170.52
0 2 4 6 8 10 12 14 16 18 20
Figure 4.9: 6 pole motor: cogging torque.
is important that its amplitude remains contained, especially in the application under analysis,
where the EM is always a mechanical load for an ICE and additional mechanical strains should
be avoided to preserve the working life of all the powertrain components.
The trend of the cogging torque has been obtained for all the three designed motors with
the software FEMM using three different methods.
1. Gap integral: the software computes the discretized integral formulation of Maxwell stress
tensor considering the radial and tangential components of the magnetic fields and flux
density vectors along a contour inside the sliding band in the air gap.
2. Weighted stress tensor: the software discretizes the Maxwell stress tensor as a weighted
sum over the rotor volume considering the magnetic fields and the flux density vectors
along some contours chosen by a particular algorithm implemented inside FEMM.
3. Coenergy variation: according to Equation 3.3, with no feeding current the electromag-
netic torque can be computed by discretizing the derivative of the magnetic coenergy with
its incremental ratio:
∂wI ∆W I
τcog = =
∂θ ∆θ
The simulation of an entire electrical rotation is not needed because the period of the cogging
torque Θcog is much smaller than 360 electrical degrees:
360
Θcog = p = 20◦
lcm(Qs , 2p)
The results for the three machines are reported in Figures 4.9-4.10-4.11. It can be seen that the
outcomes from method (1) and (2) are the same. So in the next torque computation only the
Gap integral has been used. The amplitude of the cogging torque decreases with the increase
of the number of poles. The 10 pole machine indeed presents the lower peak value. However
for all the three motors the cogging remains under the 3% of the rated torque.
67
3
-1
-2
-3
0 2 4 6 8 10 12 14 16 18 20
123.49
123.48
123.47
123.46
123.45
123.44
123.43
123.42
123.41
0 2 4 6 8 10 12 14 16 18 20
2.5
2
1.5
1
0.5
0
-0.5
-1
-1.5
-2
-2.5
0 2 4 6 8 10 12 14 16 18 20
101.52
101.51
101.5
101.49
101.48
101.47
101.46
0 2 4 6 8 10 12 14 16 18 20
68
6 poles machine: MTPA search
500
450
400
350
300
250
200
150
100
50
0
90 95 100 105 110 115 120
Figure 4.12: 6 pole motor: MTPA search.
4.2 MTPA search

A MTPA search has been done in order to asses properly the on-load torque performances and
to provide a lookup table that can be useful to map the reference signals of the control system
according to the required average torque. The analysis has been performed varying the current
phase angle θi in the range 90◦ ÷ 120 for many current amplitudes I [A] until the maximum
overload. For every trial the average electromagnetic torque has been computed as the mean
of the Maxwell stress tensor, simulating a half electrical rotation. Indeed the torque has a
®
periodicity of 180 electrical degrees. Such analysis is very time consuming and to do it the
MATLAB parallel computation toolbox has been used. The results are very similar between
the three motors. As example, Figure 4.12 reports the outcomes of the 6 pole machine MTPA
search. It can be seen that the maximum torque is obtained with a quadrature current for
every rms value. Thus the statements presented in Section 3.2 have been verified. The MTPA
condition is actually achieved by imposing three phase current with θi = 90◦ even at maximum
overload. This means that the motors are never enough saturated to shift the MTPA region in
the second quadrant of the id − iq plane.
4.3 Performances analysis at rated load and at maximum

overload
The motors are always driven in MTPA condition, so √ the on-load performances have √ been
assessed by imposing quadrature current: id = 0, iq = 2IR for the rated load and iq = 2IOL
for the maximum overload. The flux maps of the three motors at rated load are shown in Figures
4.13-4.15-4.17, while the maximum overload conditions are reported in Figures 4.14-4.16-4.18.
It can be seen the impact of the electric load, that slant the flux lines from the d axis towards
the quadrature axis. Such phenomenon can be identified more clearly in the air gap flux density
distribution, reported in Figures 4.19-4.20 for the 6 pole machine. The magnetic circuit has
been analyzed in the same way adopted for the unload condition, except for the teeth, that now
are magnetized differently due to the flux density distorsion. For brevity purpose the results
are not reported. However the comments made for the unload condition are still valid.
69
1.805e+000 : >1.900e+000
1.254e+000 : >1.320e+000 1.710e+000 : 1.805e+000
1.188e+000 : 1.254e+000 1.615e+000 : 1.710e+000
1.122e+000 : 1.188e+000 1.520e+000 : 1.615e+000
1.056e+000 : 1.122e+000 1.425e+000 : 1.520e+000
9.900e-001 : 1.056e+000 1.330e+000 : 1.425e+000
9.240e-001 : 9.900e-001 1.235e+000 : 1.330e+000
8.580e-001 : 9.240e-001 1.140e+000 : 1.235e+000
7.920e-001 : 8.580e-001 1.045e+000 : 1.140e+000
7.260e-001 : 7.920e-001 9.500e-001 : 1.045e+000
6.600e-001 : 7.260e-001 8.550e-001 : 9.500e-001
5.940e-001 : 6.600e-001 7.600e-001 : 8.550e-001
5.280e-001 : 5.940e-001 6.650e-001 : 7.600e-001
4.620e-001 : 5.280e-001 5.700e-001 : 6.650e-001
3.960e-001 : 4.620e-001 4.750e-001 : 5.700e-001
3.300e-001 : 3.960e-001 3.800e-001 : 4.750e-001
2.640e-001 : 3.300e-001 2.850e-001 : 3.800e-001
1.980e-001 : 2.640e-001 1.900e-001 : 2.850e-001
1.320e-001 : 1.980e-001 9.500e-002 : 1.900e-001
6.600e-002 : 1.320e-001 <0.000e+000 : 9.500e-002
<0.000e+000 : 6.600e-002
Figure 4.14: 6 pole motor: flux map on maximum over-

Figure 4.13: 6 pole motor: flux map on rated load.
load.
1.463e+000 : >1.540e+000
1.264e+000 : >1.330e+000 1.386e+000 : 1.463e+000
1.197e+000 : 1.264e+000 1.309e+000 : 1.386e+000
1.131e+000 : 1.197e+000 1.232e+000 : 1.309e+000
1.064e+000 : 1.131e+000 1.155e+000 : 1.232e+000
9.975e-001 : 1.064e+000 1.078e+000 : 1.155e+000
9.310e-001 : 9.975e-001 1.001e+000 : 1.078e+000
8.645e-001 : 9.310e-001 9.240e-001 : 1.001e+000
7.980e-001 : 8.645e-001 8.470e-001 : 9.240e-001
7.315e-001 : 7.980e-001 7.700e-001 : 8.470e-001
6.650e-001 : 7.315e-001 6.930e-001 : 7.700e-001
5.985e-001 : 6.650e-001 6.160e-001 : 6.930e-001
5.320e-001 : 5.985e-001 5.390e-001 : 6.160e-001
4.655e-001 : 5.320e-001 4.620e-001 : 5.390e-001
3.990e-001 : 4.655e-001 3.850e-001 : 4.620e-001
3.325e-001 : 3.990e-001 3.080e-001 : 3.850e-001
2.660e-001 : 3.325e-001 2.310e-001 : 3.080e-001
1.995e-001 : 2.660e-001 1.540e-001 : 2.310e-001
1.330e-001 : 1.995e-001 7.700e-002 : 1.540e-001
6.650e-002 : 1.330e-001 <0.000e+000 : 7.700e-002
<0.000e+000 : 6.650e-002

load.
1.254e+000 : >1.320e+000
8.550e-001 : >9.000e-001 1.188e+000 : 1.254e+000
8.100e-001 : 8.550e-001 1.122e+000 : 1.188e+000
7.650e-001 : 8.100e-001 1.056e+000 : 1.122e+000
7.200e-001 : 7.650e-001 9.900e-001 : 1.056e+000
6.750e-001 : 7.200e-001 9.240e-001 : 9.900e-001
6.300e-001 : 6.750e-001 8.580e-001 : 9.240e-001
5.850e-001 : 6.300e-001 7.920e-001 : 8.580e-001
5.400e-001 : 5.850e-001 7.260e-001 : 7.920e-001
4.950e-001 : 5.400e-001 6.600e-001 : 7.260e-001
4.500e-001 : 4.950e-001 5.940e-001 : 6.600e-001
4.050e-001 : 4.500e-001 5.280e-001 : 5.940e-001
3.600e-001 : 4.050e-001 4.620e-001 : 5.280e-001
3.150e-001 : 3.600e-001 3.960e-001 : 4.620e-001
2.700e-001 : 3.150e-001 3.300e-001 : 3.960e-001
2.250e-001 : 2.700e-001 2.640e-001 : 3.300e-001
1.800e-001 : 2.250e-001 1.980e-001 : 2.640e-001
1.350e-001 : 1.800e-001 1.320e-001 : 1.980e-001
9.000e-002 : 1.350e-001 6.600e-002 : 1.320e-001
4.500e-002 : 9.000e-002 <0.000e+000 : 6.600e-002
<0.000e+000 : 4.500e-002

load.
70
1.5 2
1.5
1
1
0.5
0.5
0 0
-0.5
-0.5
-1
-1
-1.5
-1.5 -2
0 90 180 270 360 0 90 180 270 360
Figure 4.19: 6 pole motor: air gap flux den- Figure 4.20: 6 pole motor: air gap flux den-
sity distribution in rated load. sity distribution in maximum overload.
Table 4.2: Rated load and maximum overload torque performances.

2p 6 8 10
TR [N m] 126 127 130
Rated load
ripple [%] 12 8.0 7.5
TOL [N m] 492 481 480
Max. overload
ripple [%] 10 6.2 5.2
In the electromagnetic torque computation the ripple component due to the coenergy varia-
ˇ
tion has been distinguished from the average value generated by = λ̄ī]. The results are reported
in Figures 4.21-4.22-4.23 for the rated load and in Figures 4.24-4.25-4.26 for maximum over-
load. The performances of the three motors are summarized in Table 4.2. The specifications
are entirely fulfilled and the ripples are admissible.
4.4 Demagnetization test

The behavior of the PMs against demagnetization has been assessed through a magnetostatic
FEA in the most critical√working condition, i.e. with a maximum overload direct-axis demag-
netizing current: id = − 2IOL , iq = 0. Such situation should never happen during the normal
operations, but it can occurs in the event of an inverter fault or a control system failure.
The magnetic fields and the flux density vectors in the PMs elements, which are referred to
x − y cartesian coordinates in the FEMM interface, have been projected onto the magnetization
direction. Then the minimum values Bmin , Hmin have been compared with the ’knee’ point,
whose parameters are taken from Table 3.3 for the N42 UH magnet at the supposed working
temperature of 120◦ . The results are reported in Table 4.3. The PM does not demagnetize
even in this extreme condition as its working point stays always above the ’knee’.
Table 4.3: PM working point in the most critical demagnetizing condition.

2p / 6 8 10
Bknee [T ] -0.16 -0.16 -0.16
Hknee [kA/m] -895 -895 -895
Bmin [T ] -0.04 -0.04 +0.05
Hmin [kA/m] -816 -814 -751
71
132
130
128
126
124
122
120
118
116
114
0 45 90 135 180
-2
-4
-6
-8
-10
0 45 90 135 180
Figure 4.21: 6 pole motor: rated load torque.
135
130
125
120
0 45 90 135 180
6
-2
-4
-6
-8
-10
0 45 90 135 180
Figure 4.22: 8 pols motor: rated load torque.
72
138
136
134
132
130
128
126
124
122
0 45 90 135 180
6
-2
-4
-6
-8
0 45 90 135 180
Figure 4.23: 10 pole motor: rated load torque.
520
510
500
490
480
470
460
450
0 45 90 135 180
Figure 4.24: 6 pole motor: maximum overload torque.
495
490
485
480
475
470
465
460
0 45 90 135 180
73
495
490
485
480
475
470
465
460
0 45 90 135 180
4.5 Inductance computation

The designed machines have an isotropic magnetic structure, so the inductance on the d axis and
the inductance on the q axis are equal and correspond to the so-called synchronous inductance:
Ld = Lq = L. L is composed by three main contributions: the magnetizing inductance Lm , due
to the magnetizing electric load; the slot flux leakage Lσslot and the end-winding flux leakage
Lσew . Leakages due to magnets or teeth skewing do not concern the designed motors, as no
skewing has been planned. √ To compute L through a magnetostatic FEA, a direct-axis rated
current is imposed (id = 2IR , iq = 0) with the rotor in the θe = 0 position and the magnets
have been ’turned off’, i.e. their coercivities have been set to zero. Then the inductance has
been calculated with three different methods to have a comparison on the results.
1. Magnetic energy:
4W
L0 =
3id
2. Flux linkage:
L0 = Λd /id
3. Fundamental flux density distribution in the air gap:

1
Bgd kw Ns Ds Lf e
Lm =
2pid
1
Where Bgd is the peak value of the first harmonic of the flux density distribution. As
an example, the distribution for the 6 pole machine is shown in Figure 4.27, with its
harmonic spectrum.
Methods (1) and (2) takes into account both the magnetizing contribution and the slot
leakage (L0 = Lm + Lσslot ) but they cannot consider the end-winding leakage, because the
simulation is bi-dimensional. So the end-turns leakage induction Lσew has been estimated with
Someda formula [24]:
Lσew = 0.5µ0 τp 2pqs2 ncs
Method (3) instead considers only the magnetizing inductance, so both leakage contributions
must be summed up.
Combining the outcomes from these three computations the synchronous inductance has
been separated into its three main components for all the three motors. The results are reported
in Table 4.4. The inductances are very small: this is a characteristic feature of SPM motors,
that provides a high electromagnetic responsiveness but at the same time limits the speed
range of the machine and makes worse the effects of fault in the electric drive. Moreover the
filtering action on the high frequency harmonics generated by the inverter PWM technique
74
0.4 1
0.9
0.3
0.8
0.2
0.7
0.1
0.6
0 0.5
0.4
-0.1
0.3
-0.2
0.2
-0.3
0.1
-0.4 0
0 90 180 270 360 1 2 3 4 5 6 7 8 9 10 11
Figure 4.27: 6 pole motor: flux density distribution due only to rated d-axis current.
Table 4.4: Machines inductances in rated load.

2p / 6 8 10
Lm [mH] 0.70 0.57 0.45
Lσslot [mH] 0.20 0.38 0.34
Lσew [mH] 0.20 0.19 0.14
L [mH] 1.10 1.14 0.93
is very poor. Making a comparison between the three motors, the 8 pole machine has the
highest slot leakage because it has the longest slots, whereas the 10 pole machine has the lowest
end-winding leakage as it has the shortest end-turns. Then 6 poles motor has the highest
magnetizing inductance because it is the less saturated machine and it has the lowest electric
load. Finally it is important to underline that the inductance is not completely independent
of the load. Indeed the magnetizing contribution depends slightly on the applied current. In
this thesis for brevity purpose only the results at rated load are presented and the synchronous
inductance is considered independent of the load.
4.6 Winding resistance and Joule losses

The winding resistance and the Joule losses cannot be computed properly with FEMM. So they
have been estimated analytically.
The rated in-slot working temperature at steady state has been supposed at 140◦ C. An H
class has been chosen for the insulating materials, so the maximum in-slot allowable temperature
is 180◦ C. Considering a safety margin of 15◦ C recommended by the IEC standards, a temper-
ature rise of 25◦ C can be withstand safely during the overloads. The conductor resistivity at
the supposed steady-state working temperature has been obtained as follows:
ρ = ρ20◦ C (1 + α(140 − 20)) = 0.0266 [Ω mm2 /m]
Where the parameters of the copper conductor at environmental temperature are: ρ20◦ C =
0.018 [Ω mm2 /m], α = 0.004 [◦ C −1 ]. Hence the winding phase resistance has been calculated
with the following expression:
Ns (Lstk + lew )
R=ρ
Sc npp
Where the end-turns length lew has been computed as presented in Section 3.9.
Then the Joule losses have been mapped in each working point using the following relation:
PJ = 3RI 2
75
Table 4.5: Winding resistance and rated load Joule losses.
Winding resistance R [Ω] 0.121 0.126 0.125
In-slot losses PJslot [W ] 300 382 369
End-turns losses PJew [W ] 721 757 739
Total losses PJ [W ] 1021 1139 1108
Figure 4.28: 6 pole motor Joule losses map.
That is equivalent to:

PJ = ρJ 2 V olCu
Where I and J are respectively the rms values of the applied current and current density, while
ρJ 2 [W/m3 ] are the specific losses. The Joule losses depend on the square of the feeding current,
thus they are proportional on the square of the mean torque, according to Equation 3.4. Any
losses contribution related to the applied frequency, i.e. the machine speed, due to skin or
proximity effects has been neglected because all the chosen wires diameters are much smaller
than the copper skin depth at 200 [Hz].
The outcomes of the above presented computations are reported in Figures 4.28-4.29-4.30
and in Table 4.5, where the contributions of the in-slot conductors and end-turns to the rated
total Joule losses are distinguished.
4.7 Voltage specifications fulfillment

Once that the winding resistance and the machine inductance are known, the three phase
sinusoidal voltage needed to inject a specified current can be found out analyzing the steady-
state electrical behavior of the motor when working in MTPA condition, which is graphically
represented by the three-phase circuit of Figure 4.31(a) and by the related space vectors diagram
in Figure 4.31(b). Thus the needed voltages rms values on rated load and maximum overload
have been calculated considering the space vector representation of the three phase quantities
or, in an equal way, the phasors of the equivalent one-phase circuit:
q 2 2
Vwind = E0 + RI + 2πfR LI
76
77
Table 4.6: Needed winding voltage on rated load and on maximum overload.
Poles number 2p / 6 8 10
Vwind [V ] 195 191 196
Rated load
cos φ / 0.99 0.99 0.99
Vwind [V ] 215 212 217
Max. overload
cos φ / 0.99 0.99 0.98
q
e0a ωLI¯
L R ia
− + RI¯
+ V̄wind
vwinda Ē0
e0b
L R ib
− +
+ φ I¯
vwindb
e0c
L R ic
− +
+ Λ̄0 d
vwindc
(a) (b)
Figure 4.31: SPM motor equivalent circuit (a) and space vectors diagram in MTPA working
condition (b).
E0 + RI
cos φ =
Vwind
The results are reported in Table 4.6. It can be seen that the inverter works always with a very
high power factor and the needed winding voltage are far below the maximum sinusoidal value.
So it can be stated that the power electronic components are well exploited in terms of current
but not in terms of voltage. Hence a DC bus voltage lower than 700 [V ] could be chosen with
a benefit in the battery pack sizing.
4.8 Iron and permanent magnets losses

Iron and permanent magnets losses have been mapped through a series of finite elements analy-
®
sis. The mapping has required several hundreds of FEAs, so the FEMM simulations have been
managed using the MATLAB parallel computation toolbox. To compute the rotating losses
the method presented in [41, 42] has been implemented.
Stator iron losses have been obtained considering also the harmonic content due to the
non-linear polarization of the flux density vectors. Indeed, looking at the fields pointwise, the
vectors do not follow a straight line direction, as it is pointed out in Figures 4.32-4.33, where
the ideal linear polarization is compared to the actual behavior of the flux density vector in the
middle of the tooth and in the middle of the back iron during a complete electrical rotation of
the 6 pole machine. So the iron loss density pf e [W/m3 ] has been computed in each stator mesh
element by applying the well-known Steinmatz formula to every flux density harmonic until the
12th order at the given fundamental frequency f [Hz], i.e. at a specified rotating speed n [rpm]
in the range 0 ÷ nn = 2300 [rpm]:
12 h
X i
pf e = Chyst fh Bh2 + Ceddy fh2 Bh2
h=1
78
By
0.8 1.5
0.6
1
0.4
0.5
0.2
0 0
Bx -0.2
-0.5
-0.4
-1
-0.6
-0.8 -1.5
-1.5 -1 -0.5 0 0.5 1 1.5 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8
Figure 4.32: Idel linear Figure 4.33: 6 poles motor: flux density vector polarization at rated
polarization. load.
Where Bh [T ] and fh = hf [Hz] are respectively the flux density magnitude peak value and
the frequency of the hth harmonic. The Fourier series coefficients of the flux density magnitude
®
in each mesh element centroid, namely the Bh values, have been computed through the FFT
algorithm implemented in MATLAB . Harmonic orders higher than 12th have not been taken
into account because of the relatively coarse mesh used in the simulations, which causes a lot of
noise at high frequency. The Steinmatz equation considers both the two contributions to iron
losses: the hysteresis effect is taken into account with the coefficient Chyst and changes linearly
with the frequency while the eddy current losses are included with the coefficient Ceddy and
vary with the square of the frequency. The coefficients have been evaluated as follows:
γf e pf eref
Chyst = 0.75 2 = [W/(Hz T 2 m3 )]
kpack fref Bref
γf e pf eref
Ceddy = 0.25 2 B2 = [W/(Hz 2 T 2 m3 )]
kpack fref ref
Where the iron density has been set to γf e = 7800 [Kg/m3 ] while the reference frequency
and the reference flux density peak value have been taken respectively as fref = 50 [Hz] and
Bref = 1.0 [T ], so that the related reference specific loss has been found out from the loss
characteristic in Figure 3.17: pref = 1.4 [W/kg]. As it can be seen the stacking coefficient
kpack has been also taken into account. It can be noted that the entire loss density scales by
B 2 , so it would be expected that the concentration of flux would increase the loss inside the
2
lamination by 1/kpack . Nevertheless, the iron per unit volume also goes down, thus there is
less iron per unit volume to generate losses, which gives back a factor of kpack . Overall, the
stacking factor can be incorporated in the Steinmatz formulation by dividing the coefficients
by the stacking factor. It can be also noted that the main loss contribution is given by the
hysteresis effect, whereas eddy current become relevant only at very high frequency thanks to
the small laminations thickness and the electrical insulation in axial direction.
The total iron losses have been computed by summing up the loss density elementwise
multiplied by the related mesh element volume. The contributions of the teeth and the back
iron have been separated to underline their weight in the overall losses and to account for the
laminations weakening due to the cutting process. The back iron part is less subjected to cut
damages ,then the back iron losses have been increased by 50%, whereas the teeth are affected
heavily by the slots shearing off, so the teeth losses have been raised by 100%.
Rotor yoke losses are negligible in steady-state working condition and, thanks to the lamina-
tions, they are limited also during the electromechanical transients. Instead permanent magnets
are not laminated and their electrical resistivity is not high enough to avoid the flow of eddy
current. As a consequence they produce non negligible Joule losses. PMs ’see’ a changing flux
even during steady-state rotations, thus they are always affected by eddy current. Indeed the
slots openings and the mmf generated by the electric load causes a magnetic reluctance varia-
tion in the air gap and a distortion of the flux density distribution, which lead to a varying flux
on the PMs. The low frequency magnets losses due to slotting and mmf have been calculated
79
as presented in [42]. The low frequency loss density in each PM mesh element ppmlf [W/m3 ]
can be expressed by the following relation.
1
ppmlf = ρpm Jˆpm
2
(4.1)
2
The NdFeB resistivity at the supposed working temperature of 120◦ C has been set to ρP M =
1.5[Ωmm2 /m] while the peak value of the induced current density Jˆpm can be further expressed
as the time variation of the magnetic vector potential, ignoring the reaction field produced by
the eddy current:
∂ Âpm 2πf ∂ Âpm
Jˆpm = −σpm =−
∂t ρpm ∂θe
The magnetic vector potential taken at the centroid of each PM mesh element has been de-
composed in Fourier series in the same way adopted for the stator flux density vectors. Each
harmonic is a sine wave, hence it can be represented in the phasors domain. So the hth harmonic
of the induced current density can be written as:
2πf
Jˆh = −j Âh
ρpm
And consequently Equation 4.1 becomes:

12
1 4π 2 X 2 2
ppmlf = fh Âh
2 ρpm
h=1
Then the magnets low frequency losses have been obtained by summing up the loss density
elementwise multiplied by the related mesh element volume.
Another non-negligible contribution to permanent magnets losses is given by the inverter
PWM (Pulse Width Modulation). Indeed the synthesized feeding voltage is only approximately
a sine wave because a high frequency harmonic content due to switching is always present.
Therefore also the feeding current waveform is affected by high order harmonics close to the
switching frequency. The amplitude of such noisy high frequency current is heavily related to the
filtering action provided by the series inductances in the electric circuit. SPM motors have very
low inductances, so the harmonic filtering action is very poor. Thus a noisy signal amplitude
of 10% of the load fundamental current has been supposed. For what concerns the inverter,
the current specifications of the power switches are relatively trying (rated current above 50 [A]
with overloads up to 220 [A]), therefore the working frequency should not be high considering
standard components. Indeed power switches with high current specifications usually allow
relatively low working frequency. So a switching frequency of 10 [kHz] has been hypothesized.
High frequency PMs losses cannot be computed neither with magnetostatic FEA, as they do
not consider any eddy current, nor with pure harmonic analysis, as they do not account for
the constant field offset given by the magnets. Therefore the computation has been performed
thanks to the incremental permeability procedure available in FEMM software, which allows
the steady-state solution of a harmonic problem at a given frequency starting from a DC offset.
The harmonic solution can be calculated steady-state because the switching frequency is always
several hundreds of times higher than the fundamental frequency, which can be included in the
DC offset together with the flux generated by the PMs. The high frequency loss density in each
PM element ppmhf [W/m3 ] has been computed as in Equation 4.1, where the current density
is the one induced by the noisy feeding current at switching frequency. Then again the total
losses have been obtained summing up the loss density elementwise multiplied by the related
mesh element volume.
The iron and permanent magnets losses maps are reported in Figures 4.34-4.36-4.38 for all
the three designed motors. It can be noted that the PM losses cannot be neglected as they are
about the 50% of the overall iron losses. All components depend more on the feeding frequency,
i.e. the speed, than on the applied electric load, except for the high frequency PM losses,
which vary with the feeding current and they are independent on the speed. Such contribution
80
Table 4.7: Iron and PM losses on rated working condition.
Poles number 2p / 6 8 10
Back iron Pf ebi [W ] 270 214 254
Teeth Pf et [W ] 215 284 374
Overall iron Pf e [W ] 485 498 628
PM low freq. Ppmlf [W ] 70 191 217
PM high freq. Ppmhf [W ] 136 210 189
Overall PM Ppm [W ] 206 401 406
Figure 4.34: 6 pole machine: iron an PM losses maps.
becomes very relevant at high electric loads. Figures 4.35-4.37-4.39 highlight the weights of all
the rotating losses components as well as the contribution of the polarization harmonic content
in the overall iron losses.
The results on rated working condition, i.e. at the base point, are summarized in Table 4.7
for all the three machines.
4.9 Efficiency
The efficiency map of each machine has been obtained from the losses characterizations pre-
sented in the previous sections.
The electrical machines are also affected by mechanical losses caused by frictions in bearings
and ventilation in the air gap and around the shaft. This losses contribution has been estimated
through the following empirical formulation, that is based on statistical investigations on similar
motors in the same speed range [24]:
√
Pmec = 0.7 · P[kW ] n[rpm]
In every working point, P[kW ] is the useful mechanical power generated at the shaft while n[rpm]
is the rotating speed. Hence the efficiency has been computed with a 10% increase of the overall
losses to consider also additional non evaluable contributions:
P
η=
P + 1.1(PJ + Pf e + Ppm + Pmec )
The resulting efficiency maps are shown in Figures 4.40-4.41-4.42, where the rated torque (red
line) and the maximum overload capability (blue line) are underlined. The losses and efficiency
at rated load are summarized in Table 4.8. It can be seen that the efficiency is quite similar
81
300 500
Teeth 450 Overall

Back-iron First harmonic
250
PM low freq. 400
PM high freq.
350
200
300
150 250
200
100
150
100
50
50
0 0
0 23 46 69 92 115 0 23 46 69 92 115
Figure 4.35: 6 pole machine: iron an PM losses contributions at rated load and varying fre-
quency.
300 500
Teeth 450 Overall

Back-iron First harmonic
250
PM low freq. 400
PM high freq.
350
200
300
150 250
200
100
150
100
50
50
0 0
0 31 61 92 123 153 0 31 61 92 123 153
Figure 4.37: 8 pole machine: iron an PM losses contributions at rated load and varying fre-
quency.
82
400 700
Teeth Overall
350 Back-iron 600 First harmonic
PM low freq.
300 PM high freq.
500
250
400
200
300
150
200
100
50 100
0 0
0 38 77 115 153 192 0 38 77 115 153 192
Figure 4.39: 10 pole machine: iron an PM losses contributions at rated load and varying
frequency.
83
Table 4.8: Losses and efficiency at rated load.
Joule losses PJ [W ] 1021 1139 1108
Iron losses Pf e [W ] 485 498 628
PM losses Ppm [W ] 206 401 406
Mechanical losses Pmec [W ] 1000 1019 1038
Efficiency η [/] 0.909 0.900 0.898
500
0.45
5
0.3
0.6
0.7
0.6
450
0.4
0.55 .5
75
0
400
0.
350
0.6
0.8
5
300
0.6
0.7
250
75
5
0.8
0.
200
8
0.
150
5
0.8
100 0.9
50
0.9
0
0 500 1000 1500 2000
Figure 4.40: 6 pole machine: efficiency map.
between the three motors. The 10 pole machine suffers a bit more the iron and permanent
magnet losses due to its higher feeding frequency.
84
500
5
0.3
0.6
0.6
450
0.4
0.45
0.7
0.55
0.5
400 75
0.
350 0.6
300
5 0.8
0.6
7
0.
250
75
200 0. 0.85
150 0.8
5
100 0.8 0.9
50
0.9
0
0 500 1000 1500 2000
500
0.45
0.6
0.7
0.6
0.3
450
0.4
0.5
75
0.
0.55
400
350 0.8
5
0.6
0.6
300
0.7
250
75
0.
200
0.85
0.8
150
5
0.8
100 0.9
50
0.9
0
0 500 1000 1500 2000
85
Chapter 5
Cooling system design and

thermal analysis
In this chapter the design of the cooling system and consequent thermal analysis on the machines
will be presented.
At the beginning the choice of the cooling device will be briefly discussed and justified. The
chosen system has been designed following an analytical method based on a semi-empirical fluid-
dynamics model of the convective heat exchange with a coolant. The model makes extensive use
of the dimensionless numbers that characterize fluids motion and the convection heat transfer.
After that the cooling system has been completely defined, some significant thermal analysis
on the three designed motors have been done. The temperature maps of the three machines
have been obtained firstly at steady state on rated load, i.e. when the machine works at the
base point, with finite element analysis accomplished thanks to the thermal solver (FEH: Finite
Element Heating) of the FEMM software. Then the thermal step response to rated load has
been analyzed by solving a linear lumped-parameters equivalent network (LPN) in Simulink ®
in order to compute the actual thermal time constants of the motors and make a comparison
with the values supposed before the machines design (Section 2.6). A comparison between
the thermal steady state obtained with FEMM and with the LPN has also been done, looking
especially at the slot and magnet temperature, which are the most thermal-sensitive parts of
the motors. In the end the transient thermal behavior of the three designed machines has been
analyzed using the LPN during the most critical duty cycle available, i.e. the transportation
test, in order to assess how much the motors are able to withstand a real working condition
without suffer damaging overtemperatures.
5.1 Water jacket design

For this project the cooling system of the three designed motors consists in an inox steel water
jacket with a spiral channel which in the coolant is kept in motion by a pump.
The denomination ’water jacket’ could mislead. Indeed in the application under analysis
pure water cannot be use as in cold seasons it can easily freeze with consequent damages to the
entire system. So the coolant is a solution of propylene glycol: a concentration of 20 ÷ 30% is
enough to lower the freezing point under −10◦ C. Propylene glycol has quite the same thermo-
physics properties of ethylene glycol, another fluid commonly used in cooling systems, but it is
far less toxic and easier to synthesized. Therefore, when possible, propylene glycol is preferable.
The water jacket is made of inox steel. Such material has been preferred to aluminum
because it is more suitable for a hydraulic system thanks to its greater resistance against
oxidation and corrosion, despite its weight and production cost are higher. A layout with a
single spiral duct has been chosen instead of a single axial channel because the aspect ratio of
all the designed motors is very low. As a consequence, in axial direction there is less available
surface to heat exchange and an axial duct should be manufactured with a lot of tight pipe-
bends, so that the friction losses increase a lot and more power from the pump is needed to
86
FLOW IN FLOW OUT
0
hD
w
twj wD
Lwj
De
Figure 5.1: Layout of the water jacket.
overcome the pressure drop and keep the coolant in motion at the wanted velocity.
The water jacket designed has been carried out referring to the geometry drawn in Figure
5.1. As it can be seen the spiral pitch is developed along an entire turn, i.e. along all the
external circumference. In this way the heat dissipation is more uniformly distributed along
the stack length avoiding localized hot spots. Moreover the inflow and the outflow sections of
the cooling channel are situated in the same side of the machine. Such trick is worthwhile to
shorten the pipe length that does not take part into the heat exchange with the machine in
order to reduce unwanted hydraulic losses and require less power from the pump. The total
axial length of the water jacket has been set to Lwj = 70[mm], a bit more than the stack lengths
of the three machines in order to cover properly all the stacks and at the same time to have
enough space to try different number of turns for the spiral duct NDt . The halfway widths w0
as well as the inox steel thickness twj have been set to 4 [mm] to allow an easy manufacturing
without subtract too much volume to the cooling channel. The width of the duct wD has been
computed for each attempt of spiral turns as follows:
Lwj − (NDt + 2)w0

wD =
NDt + 1
To obtained the height hD , the half-perimeter of the duct rectangular cross section pD =
hD + wD has been compute for each machine using the following analytical procedure [33].
The heat dissipation capacity of the coolant can be expressed as follows:
qloss = ṁcpf ∆Tf = γf Q̇cpf ∆Tf = γf vf SD cpf ∆Tf (5.1)
The coolant temperature raise has been set to ∆Tf = Tout − Tin = 4◦ C, a difference that
has been considered small enough to provide a heat exchange uniformly distributed along the
cooling channel without requiring an excessive effort to pump the needed flow rate. So supposing
an inflow temperature Tin = 40◦ C, as suggested by the IEC standards for the thermal study
of the electrical machine [24], the fluid leaves the circumferential duct at the temperature of
Tout = 44◦ C. Thus, to evaluate the convective heat transfer, the thermo-physics properties of
the coolant has been taken at the average temperature of Tf = 42◦ C. The data are reported
in Table 5.1.
From Equation 5.1 it can be noted that setting the temperature raise ∆Tf means to define
also the coolant flow rate at equal heat power qloss [W ], both in terms of mass ṁ [kg/s] and
volume Q̇ = [l/s]. The heat power qloss considered in the water jacket design is the sum of
the total Joule losses, the stator iron losses and the permanent magnet losses at rated torque
87
Table 5.1: Coolant thermo physics properties at average working temperature.
Coolant 30% solution of propylene glycol
Parameter Symbol Unit Value
Mean working temperature Tf [◦ C] 42
3
Density γf [kg/m ] 1026
Dynamic viscosity µf [kg/m s] 1.213 · 10−3
Kynematic vicosity νf [m2 /s] 1.183 · 10−6
Specific heat capacity cpf [J/kg ◦ K] 3782
Thermal conductivity λf [W/m◦ K] 0.496
Prandtl number Pr [/] 9.246
and rated speed, namely when the motor works at the base point. The design has been done
on the rated load. Indeed the overload losses cannot be dissipated through the coolant because
the thermal time constant of the heat exchange between duct wall and fluid, which determine
how quickly the temperature difference increase at every applied heat flux, is much greater
than the thermal time constant of the heat exchange between the slots and the stator iron
[33]. Mechanical losses have not been taken into account as it has been supposed that they are
entirely dissipated near the bearings.
The fluid average velocity is related to the dimensionless Reynolds number Re, which is
defined as:
γf vf dD v f dD
Re = = (5.2)
µf νf
Where the hydraulic diameter of the duct dD is defined as the ratio between 4 times the cross
section area SD and the perimeter touched by the fluid:
4SD 2SD
dD = = (5.3)
2(hD + wD ) pD
So, making explicit the fluid velocity, Equation 5.2 becomes:

Reµf pD
vf = (5.4)
2γf SD
And the substitution of the previous expression in Equation 5.1 yields to:
Reµf pD cpf ∆Tf
qloss = (5.5)
2
The Reynolds number is very relevant in fluid-dynamics studies as it defines the type ot flow:
Re < 2300 is assumed as pure laminar flow; Re > 10000 is considered fully developed turbulent
flow; values in the range 2300 ÷ 10000 means the presence at the same time of both laminar
and turbulent flow. In electrical machines cooling devices a fully developed turbulent flow is
preferred because it improves the convective heat exchange [33]. But too high Reynolds numbers
leads to very high fluid speed that could require too much power from the pumping system (or
the fans system in case of air cooled machines). So in this project Reynolds number has been set
to Re = 10000, as suggested also in [33] for the same type of motors. Thus, making explicit pD
in Equation 5.5, the duct half-perimeter has been computed as follows for each sizing attempts:
2qloss
pD = cp ∆Tf
Reµf f
And the height hD has been consequently obtained as:
hD = pD − wD
88
At this point the coolant average speed vf can be also computed with Equation 5.4.
Then the convective coefficient αwj has been calculated. αwj represents the heat exchange
between the water jacket walls and the coolant so it is fundamental for further thermal analysis
of the machines, as it will be described later on. The convective coefficient can be expressed as
follows:
λf N u
αwj = (5.6)
dD
Where N u is the Nusselt dimensionless number. A lot of experimental correlations have been
proposed to evaluate N u. Among them the Gnielinsky equation has been adopted in this
project, because it is one of the latest proposal and it is affected by the lowest errors in a wide
range of different flows, as reported by the validity constrains below:

ξ
8 (Re − 1000) · P r
Nu = 0.5
1 + 12.7 · 8ξ

· P r2/3 − 1

0.5 ≤ P r ≤ 2000
3 × 103 ≤ Re ≤ 50 × 106
P r is the dimensionless Prandtl number. It depends only on the fluid properties, according to
its definition: µf cpf
Pr =
λf
The Prandtl number of the coolant at the average working temperature is reported in Table
5.1 among the other thermo-physics parameters [43].
ξ is the Darcy-Weisbach friction factor [44], which takes into account the improvement in
the convective heat exchange due to the asperities in the duct walls surfaces. Also to evaluate
this dimensionless coefficient a lot of empirical correlations have been proposed in the technical
literature. Furthermore ξ can be quickly estimated using the well-known Moody diagram. In
this project the friction factor has been evaluated by solving the Colebrook-White equation
with a Newton-based root-finding method [44]. The secant algorithm has been implemented
again, as it was done during the preliminary sizing of the motors.
1 2e 18.7
√ = 1.74 − 2 log + √
ξ dD Re ξ
The inox steel sharpness has been set to e = 0.1 [mm]. Colebrook-White correlation is not
one of the most recent proposal but it guarantees satisfying results in a wide range of different
flows.
Several attempts have been made to size the water jacket. The final choice for each motor
has been based on the power performances required to the pump. The power needed to keep
the coolant flowing has been computed as follows:
Ppump = γf Q̇gH
Where g = 9.81 [m/s2 ] is the gravitational acceleration while H [m] is the head of the pump.
The friction losses causes the main pressure drop along the channel, so H has been estimated
with the Darcy-Weisbach equation, supposing that the duct curvature can be neglected because
of the high external diameter of the machines as well as the localized losses in the inflow and
outflow joints.
vf2 LD
H=ξ
2dD g
The total lenght of the duct has been evaluated as:
LD = NDt π(De + 2twj + hD )
The outcomes from the sizing attempts are reported in Table 5.2, where the definitive choices
89
Table 5.2: Water jacket design attempts.
2p qloss NDt wD hD vf αwj Q̇ H Ppump
[/] [W ] [/] [mm] [mm] [m/s] [W/m2 ◦ C] [l/s] [m] [W ]
3 12.5 6.5 1.38 3112 0.11 0.65 0.73
6 1712 4 9.2 9.8 1.25 2754 0.11 0.62 0.70
5 7.0 12 1.34 2992 0.11 0.98 1.10
2 18.0 5.0 1.51 3457 0.14 0.48 0.66
3 12.5 10.5 1.04 2223 0.14 0.22 0.30
8 2038
4 9.2 13.8 1.07 2310 0.14 0.32 0.44
5 7.0 16 1.21 2672 0.14 0.61 0.84
2 18.0 6.0 1.31 2930 0.14 0.31 0.44
3 12.5 11.5 0.99 2103 0.14 0.19 0.27
10 2142
4 9.2 14.8 1.04 2238 0.14 0.30 0.45
5 7.0 17 1.19 2616 0.14 0.58 0.83
Figure 5.2: Insulating thermal class.
for each machine are underlined.
5.2 Steady state thermal analysis at rated load

Some relevant thermal analysis have been done to assess the actual working temperatures of
the most critical parts of the machines, i.e. the slot and the permanent magnets. Both the
in-slot insulating materials and the permanent magnets belong to the H thermal class, so they
can withstand a maximum temperature of 180◦ C without damaging, as reported in Figure 5.2.
However the IEC standard for the electrical machines design suggests a safety margin of 15◦ C
and fix the reference environmental temperature to 40◦ C. Thus, an overtemperature below
the limit of 125◦ C is recommended. The motors thermal behavior has been studied both at
steady-state and during a duty cycle transient. The steady state temperature maps at rated
load have been obtained through FEA using the thermal solver implemented in FEMM soft-
ware. The outcomes are reported in Figures 5.3-5.45.5. The great quantity of points scattered
in the stator lamination and in the PMs has been needed in the pre-processing phase to repre-
sent concentrated heat generation spots. Indeed, while the total Joule losses produced by the
rated current have been uniformly allocated in the slots cross sections, the iron and PM losses
have been imposed according to the loss density distribution computed in Section 4.8 at rated
90
speed and rated torque. It is important to specify that the Joule loss density set in the slots
takes into account also the end-winding contribution, as it has been likely supposed that the
heat dissipation coefficient, which represents the free convection in the end-turns, is too low
if compared to the copper thermal conductance, so the heat flux can be considered originated
entirely from the in-slot conductors.
®
Inside the slots are placed enameled copper wires electrically insulated with Nomex tapes
and impregnated in epoxy resin. To properly model the thermal behavior of such cluttered
assembly, an equivalent thermal conductivity λslot has been evaluated with Milton formula
[33, 45].
(1 + kf ill )λCu + (1 − kf ill )λres
λslot = λres (5.7)
(1 − kf ill )λCu + (1 + kf ill )λres
Where kf ill = 0.40 is the slot fill factor, λCu is the copper thermal conductivity and λres =
0.20 [W/m◦ C] is the thermal conductivity assumed for all the insulating materials (enamel,
®
resin and Nomex ). The thermal resistance of the insulating layer between slot and iron as
well as the wedge are also considered, but using a conductivity of λins = 0.18 [W/m◦ C].
The water jacket cooling has been represented with an external convective boundary whose
coefficient is the one reported in Table 5.2 for each motor. The conductive heat transfer provided
by the inox steel thickness twj has also been considered as well as an additional layer of 2 [mm]
with a thermal conductivity of 2.47 [W/m◦ C], which represents the thermal resistance due to
the non-ideal contact between the stator laminations and the water jacket [33]. To include
the permanent magnets losses into the simulation the heat exchange in the air gap must be
modeled. To consider the impact of the rotational flow on the convection phenomenon in an air
gap, various experimental correlations have been proposed in the technical literature. In this
project the traditional empirical formulation based on the dimensionless Taylor number T a has
been adopted [45]. The Taylor number is defined as:
s
vrot (g − tband )3
Ta = (5.8)
µair Rrot
Where:
g − tband = 0.6 [mm] is the actual air gap radial thickness; it is useful to remember that
the dimension g = 1 [mm] includes 0.6 [mm] of air and the permanent magnets bandage
thickness tband = 0.4 [mm], which can be neglected in electromagnetic studies but it must
be distinguished in thermal analysis;
Rrot = (Ds − 2(g − tband ))/2 is the outer rotor radius;
vrot = Rrot ω = Rrot πn/30 [m/s] is the rotor peripheral velocity; in this case the rotor
angular speed is the rated speed n = nR = 2300 [rpm];
µair is the dynamic viscosity of air at the environmental temperature of 40◦ C and at
atmospheric pressure.
The Nusselt number has been computed with one of the following correlations, depending on
the Taylor number, which defines the type of rotational flow:
( 2 T a < 41 laminar
Nu = 0.212T a0.63 P r0.27 41 ≤ T a ≤ 100 vortex (5.9)
0.386T a0.5 P r0.27 T a > 100 turbulent
Where P rair is the Prandtl number of air at 40◦ C and atmospheric pressure. The thermo-
physics properties of air are reported in Table 5.3, among the other materials parameter used
to set up the thermal simulations [44, 45]. It is important to underline that all the materials
have been considered isotropic, which means that their thermal conductivities are equal along
every direction. Finally the convective coefficient has been calculated as:
λair N u
αgap = (5.10)
(g − tband )
91
The rotor holes behave as additional cooling ducts. The motors have horizontal shafts and
are not equipped with fans, so neither a free convection nor an air forced convection can take
place. Nevertheless, the machines’ aspect ratios are very low, thus the vortex flow generated on
the external rotating surfaces can easily expand inside the short stack length through the rotor
holes. Again, to model this phenomenon a wide variety of empirical correlations have been
proposed. In this study the traditional Morris formulation has been used [45]. Such theory
considers the case of a forced flow inside rotating ducts about a parallel axis, where a parallel
straight flow combines with a vortex flow. Indeed the Nusselt number is derived as follows:
N u = 0.012Re0.78 0.18
axi Rerot (5.11)
Where Reaxi and Rerot are the Reynolds numbers correlated respectively to the axial flow and
to the rotating flow. Raxi has been evaluated with Equation 5.2 supposing a very low axial
average velocity (0.2 [m/s]), as the motors under analysis are not equipped with fans. Rerot
instead is defined by the following expression:
ωd2rot
Rerot = (5.12)
8νair
For steady-state FEA the speed has been set to rated speed ω = ωR = πnR /30 [rad/s] while the
hydraulic diameter of the rotor holes drot has been obtained as in Equation 5.3, where the cross
section areas and the perimeters have been calculated with the tools implemented in FEMM
post-processing interface. Then the convective coefficient has been computed with Equation
5.6. In the end a zero heat flux boundary condition has been imposed on the shaft diameter.
It is important o underline that the reference temperature in the FEA simulations has been
set to 0◦ C, so that the results must be red in terms of overtemperatures. Looking at Figures
5.3-5.45.5 it can be seen that the slot overtemperatures remain far below the limit of 125◦ C in
each motor. This is due to the prudent design of the water jacket, when also the magnets losses
have been included into the amount of heat that should be dissipated by the coolant. But the
temperatures maps shows that the major part of the PMs losses are actually dissipated in the air
gap and in the rotor holes. Thus, while the slot insulating material are thermally safeguarded,
the PMs are a bit hotter (130◦ C) than the threshold prescribed by the IEC standard in the 8
and 10 pole machines. Indeed these motors requires higher feeding frequency, so they present
higher PM losses than the 6 pole machine, as reported in Table 4.7 (Section 4.8). Moreover
the PM loss density is higher because of the smaller magnets and even the rotor holes have a
tighter cross section causing a reduced cooling effect. However the thermal analysis have been
done under severe hypothesis: the Joule losses have been concentrated inside the slots and
any cooling end effect has been neglected. Therefore all the three machines can be considered
thermally safeguarded in steady state operation.
5.3 Lumped parameter thermal network: transient ther-

mal analysis
A lumped parameter equivalent thermal network (LPTN) [45] has been implemented in Simulink ®
to analyze the step response to rated load and the transient thermal behavior during the trans-
portation duty cycle. The equivalent thermal circuit is shown in Figure 5.6. All the lumped
parameters have been evaluated with analytical expressions. A detailed description is listed
below. The material thermo-physics properties used in the computations are reported in Table
5.3. It is important to underline that the model has been referred only to a slot pitch, thus
some parameters have been divided or multiplied by the number of slots QS .
Slot resistance
hsi
Rslot =
2λslot (hsi + wse )Lstk
The equivalent thermal conductivity λslot is the same computed with Equation 5.7.
92
Table 5.3: Thermo-physics properties of the materials used in the simualtions.
Parameter Symbol Unit Value
◦
Air at 40 C and atm. pressure
Density γair [kg/m3 ] 1.128
Dynamic viscosity µair [kg/m s] 1.91 · 10−5
Kinematic viscosity νair [m2 /s] 1.70 · 10−5
Specific heat capacity cpair [J/kg ◦ C] 1007
Thermal conductivity λair [W/m◦ C] 0.027
Prandtl number P rair [/] 0.712
Copper
Density γCu [kg/m3 ] 8900
Specific heat capacity cCu [J/kg ◦ C] 382
Thermal conductivity λCu [W/m◦ C] 386
Enamel and epoxy resin
Density γres [kg/m3 ] 1100
Specific heat capacity cres [J/kg ◦ C] 1000
Thermal conductivity λres [W/m◦ C] 0.20
Slot insulation
Thermal conductivity λins [W/m◦ C] 0.18
Iron
Density γf e [kg/m3 ] 7900
Specific heat capacity cf e [J/kg ◦ C] 460
Thermal conductivity λf e [W/m◦ C] 58
Inox steel
Thermal conductivity λinox [W/m◦ C] 17
Fiberglass bandage
Thermal conductivity λband [W/m◦ C] 0.045
Permanent magnet
Density γpm [kg/m3 ] 7500
Specific heat capacity cpm [J/kg ◦ C] 502
Thermal conductivity λpm [W/m◦ C] 8.96
93
1.047e+002 : >1.100e+002
9.940e+001 : 1.047e+002
9.410e+001 : 9.940e+001
8.880e+001 : 9.410e+001
8.350e+001 : 8.880e+001
7.820e+001 : 8.350e+001
7.290e+001 : 7.820e+001
6.760e+001 : 7.290e+001
6.230e+001 : 6.760e+001
5.700e+001 : 6.230e+001
5.170e+001 : 5.700e+001
4.640e+001 : 5.170e+001
4.110e+001 : 4.640e+001
3.580e+001 : 4.110e+001
3.050e+001 : 3.580e+001
2.520e+001 : 3.050e+001
1.990e+001 : 2.520e+001
1.460e+001 : 1.990e+001
9.300e+000 : 1.460e+001
<4.000e+000 : 9.300e+000
Density Plot: Temperature (K)
Figure 5.3: 6 pole machine: temperature map on rated load.
1.144e+002 : >1.200e+002
1.088e+002 : 1.144e+002
1.032e+002 : 1.088e+002
9.760e+001 : 1.032e+002
9.200e+001 : 9.760e+001
8.640e+001 : 9.200e+001
8.080e+001 : 8.640e+001
7.520e+001 : 8.080e+001
6.960e+001 : 7.520e+001
6.400e+001 : 6.960e+001
5.840e+001 : 6.400e+001
5.280e+001 : 5.840e+001
4.720e+001 : 5.280e+001
4.160e+001 : 4.720e+001
3.600e+001 : 4.160e+001
3.040e+001 : 3.600e+001
2.480e+001 : 3.040e+001
1.920e+001 : 2.480e+001
1.360e+001 : 1.920e+001
<8.000e+000 : 1.360e+001
1.145e+002 : >1.200e+002
1.090e+002 : 1.145e+002
1.035e+002 : 1.090e+002
9.800e+001 : 1.035e+002
9.250e+001 : 9.800e+001
8.700e+001 : 9.250e+001
8.150e+001 : 8.700e+001
7.600e+001 : 8.150e+001
7.050e+001 : 7.600e+001
6.500e+001 : 7.050e+001
5.950e+001 : 6.500e+001
5.400e+001 : 5.950e+001
4.850e+001 : 5.400e+001
4.300e+001 : 4.850e+001
3.750e+001 : 4.300e+001
3.200e+001 : 3.750e+001
2.650e+001 : 3.200e+001
2.100e+001 : 2.650e+001
1.550e+001 : 2.100e+001
<1.000e+001 : 1.550e+001
94
Slot capacity
cCu GCu
Cslot = + γres cres Sslot Lstk (1 − kf ill )
Qs
It can be seen that the capacity of the end-winding, which is very relevant, has been
included too. Indeed GCu is the overall copper weight (Table 3.10, Section 3.9).
Slot insulating layers

tins
Rinslat =
λins hsi Lstk
tins
Rinsup =
λins wse Lstk
hwed
Rwed =
λins ws Lstk
Tooth resistance
hs
Rt =
λf e wt Lf e
Tooth capacity
cf e Gf et
Cf et =
Qs
Back iron resistance

Qs hbi
Rbi =
λf e π(De − hbi )Lf e
Back iron capacity

cbi Gf ebi
Cf ebi =
Qs
Contact resistance
Qs tcont
Rcont =
λcont πDe Lstk
Where tcont = 2 [mm] and λcont = 2.47 [W/m◦ C].
Water jacket resistance

Qs twj
Rwj =
λinox π(De + twj )Lwj
Where the inox steel thickness has been chosen twj = 4 [mm] and the water jacket length
is Lwj = 70 [mm].
Coolant equivalent convective resistance

Qs
Rcool =
αwj π(De + 2twj )Lwj
Where the coolant convection coefficient αwj is reported in Table 5.2 for each motor.
Air gap equivalent resistance

Qs
Rgap =
αgap π(Ds − g + tband )Lstk
Where the convective coefficient αgap has been computed using Equation 5.10. This
resistance varies with the machine speed because the heat exchange depends on the rotor
peripheral velocity, as stated by Equations 5.8-5.9.
95
Fiberglass bandage resistance
Qs tband
Rband =
λband π(Ds − 2g + tband )Lstk
Permanent magnet resistance

Qs tm
Rpm =
λpm π(Ds − 2g − tm )Lstk
Permanent magnet capacity

cpm Gpm
Cpm =
Qs
Rotor yoke resistance
Qs (Ds − 2g − 2tm )
Rrot =
4λf e π(De − 2g − 2tm )Lf e
At the nominator a mean thickness has been assumed, so the previous expression is only
an estimation of the actual rotor yoke resistance.
Rotor yoke capacity
cf e Gf erot
Crot =
Qs
Rotor hole equivalent convective resistance
1
Rduct =
αrot drot Lstk
Also this resistance varies with the rotor speed, as expressed by Equations 5.11-5.12.
The heat sources depend on the machine working point, i.e. on the motor torque and speed,
according to the losses maps presented in Sections 4.6-4.8. The circuit is linear: any relation
between temperature and Joule losses and any thermal effect on the PMs performances has
been neglected.
The step responses of the slot and PM overtemperatures have been obtained by imposing
the rated current and the rated speed. The results are reported in Figures 5.7-5.8-5.9 for each
motor. It can be seen that the temperature maps computed with FEMM match well the steady
state values calculated with the LPN for the slot overtemperatures, but they differ a lot in the
PM outputs, especially for the 8 and 10 pole machines. Such difference could be caused by the
rough estimation of the rotor yoke resistance and rotor hole cooling phenomenon. Furthermore
the PM losses are considered concentrated in the middle of the magnet in the LPN, while the
actual distribution of the loss density may be far from being uniform. From the step responses
the thermal time constant τth of the three machines has been derived by looking at the time
instant when the slot overtemperatue reach the 63.2% of the steady state value. The time
constant results about 300 [s] for all the motors (a bit higher for the 6 pole machine). This
value is much lower than the range chosen to define the torque specification for the machines
preliminary sizing, i.e. 800 ÷ 1000 [s]. This is due mainly to the compactness of the designed
motors, which have a volume and a weight that lead to reduced heat capacity and do not
allow high thermal time constants. It can be also noted that the step responses of the PM
overtemperatures have a higher time constant.
The slot and PM overtemperatures during the transportation duty cycle have been obtained
by imposing the torque and speed trends of this operation. The outcomes are reported in
Figures 5.10-5.11-5.12. Despite the thermal time constant is much lower than the expected
one, the slot overtemperature reach a peak value below 140◦ C, which can be considered the
maximum limit that the H class insulating materials can withstand. Nevertheless the safety
96
margin suggested by The IEC standard is overcome. So, adopting a conservative way, the
machine overload capability should be limited or insulating materials that belong to a superior
thermal class should be used. On the other hand, in the former case the vehicle performances
in terms of power and acceleration will be reduced, while in the latter case the motor will be
more expensive. The first solution may be the best compromise because in farming tractors
high torque values are needed more often in steady state operations (e.g. to drive some high
power tools through the PTO or to drag plowing implements) than to achieve high acceleration.
In the end, after the presented results from the thermal analysis and the related considera-
tion, it can be stated that the designed motors are not able to provide the same performances
of the traditional powertrain without the risk of overtemperature damages. Nevertheless, if a
small reduction in the torque performances can be accepted, all the three machines can safely
work in the hybrid powertrain. The temperature trends during the other duty cycles are re-
ported in Figures from 5.13 to 5.21. Plowing duty cycles and power harrowing do not present
any thermal problem.
97
Rcool
Rwj
Rcontact
Pbi Rbi /2
Rbi /2
Cf ebi
Rinsup
Rt /2
Cf et
Rslot PJ
Rinslat
Pf et
Rt /2
Rwed Cslot
Rgap
Rband
Ppm
Rpm /2
Rpm /2
Cpm
Rrot Cf erot
Rduct
Figure 5.6: Thermal equivalent network.
98
120 45
40
100
35
80 30
25
60
20
40 15
10
20
5
0 0
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
Figure 5.7: 6 pole machine: thermal step response.
90 60
80
50
70
60 40
50
30
40
30 20
20
10
10
0 0
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
90 70
80
60
70
50
60
50 40
40 30
30
20
20
10
10
0 0
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
99
140 14
120 12
100 10
80 8
60 6
40 4
20 2
0 0
0 10 20 30 40 0 10 20 30 40
Figure 5.10: 6 pole machine: temperature trend during the transportation duty cycle.
140 18
16
120
14
100
12
80 10
60 8
6
40
4
20
2
0 0
0 10 20 30 40 0 10 20 30 40
140 18
16
120
14
100
12
80 10
60 8
6
40
4
20
2
0 0
0 10 20 30 40 0 10 20 30 40
100
10 10
9 9
8 8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
0 0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Figure 5.13: 6 pole machine: temperature trend during the plough 60 duty cycle.
12 8
7
10
8
5
6 4
3
4
2
1
0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
15 12
10
10 8
5 4
0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
101
20 4.5
18 4
16
3.5
14
3
12
2.5
10
2
8
1.5
6
1
4
2 0.5
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
25 12
10
20
8
15
10
4
5
2
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
30 14
12
25
10
20
15
10
4
5
2
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
102
12 7
6
10
5
8
4
2
2
1
0 0
0 5 10 15 20 25 30 35 40 45 50 55 0 5 10 15 20 25 30 35 40 45 50 55
Figure 5.19: 6 pole machine: temperature trend during the power harrow duty cycle.
15 12
10
10 8
5 4
0 0
0 5 10 15 20 25 30 35 40 45 50 55 0 5 10 15 20 25 30 35 40 45 50 55
20 18
18 16
16
14
14
12
12
10
10
8
8
6
6
4
4
2 2
0 0
0 5 10 15 20 25 30 35 40 45 50 55 0 5 10 15 20 25 30 35 40 45 50 55
103
Conclusion
This work has faced the powertrain electrification for a 280 [Hp] open field farming tractor.
A mild-hybrid parallel architecture with a power management based on a simple threshold
logic has been chosen. It has been presented how the torque and speed trends can be found out
by modeling the power balance at the engine shaft with a one-mesh mechanical circuit which
takes as sources some duty cycles of the original powertrain, whose quantities must be referred
to the engine shaft. To simulate a few hybrid powertrains the original ICE has been supposed
and consequently three downsized engines have been chosen. It has been described how and
how much the mechanical characteristic of the original ICE can be inferred from available
measurements of torque and speed taken at the engine shaft during real working conditions.
The torque specifications needed to begin the preliminary sizing of the EM have been chosen
using the thermal equivalent torque method, that has been implemented by processing the
output of the mechanical circuit model with a low-pass filter. Hence a thermal time constant
has been supposed to select the wanted specifications at rated load and maximum overload.
Moreover the size constrains and other requirements given by the chosen hybrid architecture
have been identified and severely taken into account in the EM design.
Three SPM motor with different number of poles has been designed. The preliminary sizing
procedure has been oriented to the cost-effectiveness of the machines as much as possible, with-
out forget the wanted performances and the imposed constrains. The displacement limitation
in particular has been very challenging and a dedicated preliminary sizing method has been
developed to cope with it. Various sizing attempts have been made for each number of pole
varying the magnet grade (NdFeB), the number of slots and the magnet thickness. The main
dimensions of the machines as well as the number of slots and the magnet parameters have been
finally chosen among several chances. Distributed fractional-slot winding with copper wire con-
ductors have been preferred for all the three motors. Then the stator and the rotor stacks have
been designed and the material costs have been compared between the three motors.
Detailed electromagnetic analysis have been done on the three designed machines using
both analytical estimations and finite element analysis. The unload behavior as well as the
on-load performances are satisfactory: all the three motors generates sinusoidal emfs with low
harmonic contents and the cogging torque is very limited thanks to the chosen winding layout;
the rated torque specifications and the overload capabilities are fulfilled with acceptable ripples.
The machines are less saturated than expected, so they result a bit oversized in terms of power
capabilities and they require a lower feeding voltage. Thus in further designs it is suggested to
size the conductor for a lower current and a higher voltage keeping the induction under 2 [T ]
also at maximum overload. In this way the voltage capabilities of power electronic components
can be better exploited and the overall Joule losses in the electric drive are reduced. The losses
and the efficiency of the three motors have been mapped carefully. The iron and permanent
magnet losses in particular have been computed through a series of FEA considering all the
losses contributions.
In the end the thermal behavior of the machines have been tested both at steady state and
under a real duty cycle. A thermal FEA has been implemented to obtained the temperature
map under rated load at steady state, while the transient simulations have been done by solv-
ing an equivalent lumped parameters thermal network, where both the stator and rotor have
been taken into account. Despite the actual thermal time constants outcomes far below than
the values supposed before the preliminary sizing, the slot overtemperature remains under the
maximum value of 140◦ C even in the most critical duty cycle. Nevertheless the safety mar-
104
gin recommended by the IEC standard is overcome. So, to avoid damages on the insulating
materials, it is recommended to limit the overload capability of the EM. The acceleration per-
formances of the vehicle will be affected but they are not fundamental in farming tractors. The
slot overtemperatures during the other operations remain under the safety margin of 125◦ C
recommended for H class insulating materials. No issues come from the magnets, although the
steady state temperatures computed with FEA are higher than expected in the 8 and 10 pole
machines, due to the higher magnets loss density.
The purpose of this project is to assess the feasibility of the electric motors design through
the equivalent thermal torque method in an application not yet deeply studied. As a beginning
investigation, it offers many further developments. Different hybrid powertrains can be consid-
ered as well as different power management systems. Moreover, even with the same architecture
presented in this thesis, it is possible to choose different specifications for the EM sizing and
other type of electrical machines can be designed, analyzed and compared. Furthermore the
inverter and battery pack sizing has not been covered in this work, but from the few information
given it can be stated that also their design is not straightforward. Finally the complexity of
the application presented in this project involves many engineering fields: a careful mechanical
redesign of the clutches, flanges and engine shaft is needed; the vehicle control system requires
heavy modification; energy consumption evaluations and performances analysis should be done
on a detailed model of the tractor to compare different powertrains and different couplings
between engines and electric motors. In addition the definition of proper standard duty cycles
based on agriculture characteristic operations could make further electrification projects easier
and it could boost the industrialization of electrified agricultural machinery.
This thesis provides valid suggestions and practical examples about the design of SPM
motors for a hybrid farming tractor. Despite the achieved results, there is still a lot of work to
do in this field.
105
Appendix
This appendix reports extract from the NdFeB magnets datasheet and the technical drawings
of the three designed motors. Figures are not labeled.
106
Sez. A-A' Sez. B-B'
12.4
22.2 A
B-B' A-A'
3.2 0.6 1
Slot Detail
Scale: 2:1
13.3° 1.0 12.0
40.0°
394.0 60.0 271.0
107
64.0
186.0
2p Ds De Qs wt A
6 271 394 27 24.3 212 SCALE: 1:2
Gauge: M350-50A Author: Elia Scolaro
9.8
Sez. A-A' Sez. B-B'

A
30.8
B-B' A-A'
2.8 0.6 1
Slot Detail
Scale: 2:1
1.0 12.0
10.0°
30.0°
330.0 60.0 222.0
108
58.0
185.0
2p Ds De Qs wt A
8 222 330 36 15.0 215 SCALE: 1:2

Sez. A-A' 8.2
Sez. B-B'
A
29.6
B-B' A-A'
2.5 0.6 1
Slot Detail
Scale: 2:1
8.0°
1.0
10.0
24.0°
109
332.0 60.0 235.0
54.0
2p Ds De Qs wt A 184.0
10 235 332 45 12.4 178 SCALE: 1:2
Bibliography
[1] W. Aumer, M. Lindner, M. Geißler, and T. Herlitzius, “Electric tractor: Vision or future?,”
Agricultural Engineering, vol. 63, p. 14–15, Feb. 2008.
[2] J. Flint, D. Zhang, and P. Xu, “Preliminary market analysis for a new hybrid electric
farm tractor,” in Proceedings of the 2014 International Conference on Global Economy,
Commerce and Service Science, pp. 98–102, Atlantis Press, 2014/01.
[3] X. Bin, L. Hao, S. Zheng-He, and M. En-Rong, “Powertrain system design of medium-sized
hybrid electric tractor,” ITJ: Information Technology Journal, vol. 23, pp. 7228 – 7233,
2013.
[4] G. Moreda, M. Muñoz-Garcı́a, and P. Barreiro, “High voltage electrification of tractor and
agricultural machinery – a review,” Energy Conversion and Management, vol. 115, pp. 117
– 131, 2016.
[5] “Eu regulation on non-road vehicles emissions.” https://dieselnet.com/standards/eu. Ac-

cessed: 2020-09-01.
[6] “Us regulation on non-road vehicles emissions.” https://www.epa.gov/regulations-

emissions-vehicles-and-engines/regulations-emissions-nonroad-vehicles-and-engines. Ac-
cessed: 2020-09-01.
[7] J. Barthel, D. Gorges, M. Bell, and P. Munch, “Energy management for hybrid electric
tractors combining load point shifting, regeneration and boost,” in 2014 IEEE Vehicle
Power and Propulsion Conference (VPPC), pp. 1–6, 2014.
[8] S. Grammatico, A. Balluchi, and E. Cosoli, “A series-parallel hybrid electric powertrain

for industrial vehicles,” in Vehicle Power and Propulsion Conference (VPPC), 2010 IEEE,
pp. 1–6, 2010.
[9] M. G. de Soto, L. Emmia, C. Benavides, I. Garcia, and P. G. de Santos, “Reducing

air pollution with hybrid-powered robotic tractors for precision agriculture,” Biosystems
Engineering, vol. 143, pp. 79 – 94, 2016.
[10] D. Troncon, L. Alberti, S. Bolognani, F. Bettella, and A. Gatto, “Electrification of agri-

cultural machinery: a feasibility evaluation,” in 2019 Fourteenth International Conference
on Ecological Vehicles and Renewable Energies (EVER), pp. 1–7, 2019.
[11] R. R. Melo, F. L. M. Antunes, S. Daher, H. H. Vogt, D. Albiero, and F. L. Tofoli, “Concep-

tion of an electric propulsion system for a 9 kw electric tractor suitable for family farming,”
IET Electric Power Applications, vol. 13, pp. 1993–2004, 2019.
[12] A. Lajunen, P. Sainio, L. Laurila, J. Pippuri-Mäkeläinen, and K. Tammi, “Overview of

powertrain electrification and future scenarios for non-road mobile machinery,” Energies,
vol. 11, p. 1184, 05 2018.
[13] M. Ehsani, Y. Gao, and A. Emadi, Modern electric, hybrid electric, and fuel cell vehicles:
fundamentals, theory, and design. CRC press, 2009.
110
[14] H. Gao and J. Xue, “Modeling and economic assessment of electric transformation of
agricultural tractors fueled with diesel,” Sustainable Energy Technologies and Assessments,
vol. 39, p. 100697, 2020.
[15] D. Troncon and L. Alberti, “Case of study of the electrification of a tractor: Electric motor
performance requirements and design,” Energies, vol. 13, p. 2197, May 2020.
[16] O. Lagnelöv, G. Larsson, D. Nilsson, A. Larsolle, and P.-A. Hansson, “Performance com-
parison of charging systems for autonomous electric field tractors using dynamic simula-
tion,” Biosystems Engineering, vol. 194, pp. 121 – 137, 2020.
[17] “Finite element method magnetics: website homepage.”
http://www.femm.info/wiki/HomePage. Accessed: 2020-09-01.
[18] O. D. Momoh and M. O. Omoigui, “An overview of hybrid electric vehicle technology,” in
2009 IEEE Vehicle Power and Propulsion Conference, pp. 1286–1292, 2009.
[19] M. Kebriaei, A. H. Niasar, and B. Asaei, “Hybrid electric vehicles: An overview,” in 2015
International Conference on Connected Vehicles and Expo (ICCVE), pp. 299–305, 2015.
[20] A. Somà, F. Bruzzese, and E. Viglietti, “Hybridization factor and performances of hybrid
electric telescopic heavy vehicles,” in 2015 Tenth International Conference on Ecological
Vehicles and Renewable Energies (EVER), pp. 1–9, 2015.
[21] K. Nevala, J. Penttinen, and P. Saavalainen, “Developing of the anti-slip control of hy-
drostatic power transmission for forest tractor and optimisation of the power of diesel
engine,” in AMC’98 - Coimbra. 1998 5th International Workshop on Advanced Motion
Control. Proceedings (Cat. No.98TH8354), pp. 475–480, 1998.
[22] M. Anwar, M. G. Rasul, and N. Ashwath, “Combustion characteristics of an agricul-
tural diesel engine fuelled with papaya and stone fruit biodiesel: A comparison,” in 2019
IEEE 2nd International Conference on Renewable Energy and Power Engineering (REPE),
pp. 26–31, 2019.
[23] P. Dabove and A. M. Manzino, “Gps mass-market receivers for precise farming,” in 2014
IEEE/ION Position, Location and Navigation Symposium - PLANS 2014, pp. 472–477,
2014.
[24] N. Bianchi and S. Bolognani, Metodologie di progettazione delle macchine elettriche.
CLEUP, 2001.
[25] C. Jia, W. Qiao, and L. Qu, “Modeling and control of hybrid electric vehicles: A case
study for agricultural tractors,” in 2018 IEEE Vehicle Power and Propulsion Conference
(VPPC), pp. 1–6, 2018.
[26] D. Troncon, L. Alberti, and M. Mattetti, “A feasibility study for agriculture tractors elec-
trification: Duty cycles simulation and consumption comparison,” in 2019 IEEE Trans-
portation Electrification Conference and Expo (ITEC), pp. 1–6, 2019.
[27] “Cummins engines datasheets.” https://www.cummins.com/engines/qsm. Accessed: 2020-
09-01.
[28] “Perkins engines datasheets.” https://www.perkins.com/en GB/products/new/perkins/in
dustrial/industrial-diesel-engines.html. Accessed: 2020-09-01.
[29] “Deutz engines datasheets.” https://www.deutz.com/en/products/engines. Accessed:
2020-09-01.
[30] “Fpt industrial engines datasheets.” https://www.fptindustrial.com/global/en/engines/agri
culture. Accessed: 2020-09-01.
[31] W. Leonhard, Control of electrical drives. Springer, 2001.
111
[32] Z. Q. Zhu and D. Howe, “Electrical machines and drives for electric, hybrid, and fuel cell
vehicles,” Proceedings of the IEEE, vol. 95, no. 4, pp. 746–765, 2007.
[33] M. Facco, “Progetto di motori a magneti permanenti per powertrain ibrido elettrico di
macchina agricola,” Master’s thesis, Università degli Studi di Padova, Padova, ITA, 2018.
[34] P. Vas, Sensorless vector and direct torque control. Oxford Univ. Press, 1998.
[35] “Ferrite magnets datasheet.” https://www.eclipsemagnetics.com/media/wysiwyg/datashe

ets/magnet materials and assemblies/ferrite ceramic datasheet.pdf. Accessed: 2020-09-
01.
[36] “Alnico magnets datasheet.” https://www.eclipsemagnetics.com/media/wysiwyg/datashe
ets/magnet materials and assemblies/alnico magnets datasheet rev2.pdf. Accessed: 2020-
09-01.
[37] “Neodimium magnets datasheet.” https://www.eclipsemagnetics.com/media/wysiwyg/data
sheets/magnet materials and assemblies/ndfeb neodymium iron boron-
standard ndfeb range datasheet rev1.pdf. Accessed: 2020-09-01.
[38] “Samarium-cobalt magnets datasheet.” https://www.eclipsemagnetics.com/media/wysiwy
g/datasheets/magnet materials and assemblies/samarium cobalt magnets datasheet.pdf.
Accessed: 2020-09-01.
[39] L. Alberti, E. Fornasiero, and N. Bianchi, “Impact of the rotor yoke geometry on ro-
tor losses in permanent-magnet machines,” IEEE Transactions on Industry Applications,
vol. 48, no. 1, pp. 98–105, 2012.
[40] N. Bianchi, Electrical machine analysis using finite elements. CRC Press, 2005.
[41] “Rotating losses in a surface mount permanent magnet motor.”
http://www.femm.info/wiki/SPMLoss. Accessed: 2020-09-01.
[42] D. Ishak, Z. Q. Zhu, and D. Howe, “Eddy-current loss in the rotor magnets of permanent-
magnet brushless machines having a fractional number of slots per pole,” IEEE Transac-
tions on Magnetics, vol. 41, no. 9, pp. 2462–2469, 2005.
[43] “Glycol solution properties.” https://www.engineeringtoolbox.com/ethylene-glycol-
d 146.html. Accessed: 2020-09-01.
[44] C. Bonacina, A. Cavallini, and L. Mattarolo, Trasmissione del Calore. CLEUP, 1992.
[45] P. Shams Ghahfarokhi, A. Kallaste, A. Belahcen, and T. Vaimann, “Analytical thermal
model and flow network analysis suitable for open self-ventilated machines,” IET Electric
Power Applications, vol. 14, no. 6, pp. 929–936, 2020.
112

Scolaro Elia 1197578

Uploaded by

Copyright:

Available Formats

Scolaro Elia 1197578

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Scolaro Elia 1197578

Uploaded by

Copyright:

Available Formats

Università degli Studi di Padova

Dipartimento di Ingegneria Industriale (DII)

Tesi di Laurea Magistrale

Design of Electric Motors

Anno Accademico 2019-2020

Riassunto esteso (summary in italian language) 4

2 Case study: open field farming tractor 15

3 Electrical machines sizing 31

5 Cooling system design and thermal analysis 86

1.1 Hybridization factor

Table 1.1: Electric hybrid vehicles classification

START GEN Hydrualic

Figure 1.1: Original powertrain outline

1.2 Original powertrain

Figure 1.2: Hybrid powertrain outline

1.3 Hybrid powertrain

τL = gICE τICE + gEM τEM (1.2)

Figure 1.3: Torque adder three-gears implementation

Case study: open field farming

2.1 Data analysis

 Transportation: the tractor carried an unknown load in the field.

2.1.1 Plowing tests measurements

 engine speed [rpm]; engine shaft rotational speed;

 vehicle speed [km/h]; tractor linear velocity.

2.1.2 Power harrow test measurements

 engine torque [N m]; torque developed by ICE measured at engine shaft;

 engine speed [rpm]; engine shaft rotational speed;

 PTO torque [N m]; torque measured at PTO shaft;

 PTO speed [rpm]; PTO shaft rotational speed;

 vehicle speed [km/h]; tractor linear velocity.

Engine shaft speed

Figure 2.1: Available data from plough 60 test.

Engine shaft speed

Figure 2.2: Available data from plough 100 test.

2.1.3 Transportation test measurements

 engine torque [N m]; torque developed by ICE measured at engine shaft;

 engine speed [rpm]; engine shaft rotational speed;

 vehicle speed [km/h]; tractor linear velocity.

2.2 Mechanical circuit model and data pre-processing

Engine and PTO shafts speeds

Figure 2.3: Available data from power harrow test.

Engine shaft speed

Figure 2.4: Available data from transportation test.

Figure 2.5: Mechanical circuital model.

pICE + pEM + pT RAC + pP T O = 0

τICE + τEM + τT0 RAC + τP0 T O = 0

τEM = −τICE − τT0 RAC − τP0 T O

2.3 Data post-processing: thermal equivalent torque

τth = Rtheq Ctheq (2.3)

PJtot Rth Cth ϑ

Figure 2.6: Simplified equivalent thermal circuit of an EM.

2.4 Engines selection

Figure 2.7: Original ICE mechanical characteristic.

2.5 Hybridization tests

have been done with the following values of R.

2.6 Hybridization results

Cummins QSM280 210kW torque reduction Downsized ICEs torque

1000 R=75% 1000

Cummins QSM280 210kW power reduction Downsized ICEs power

Transportation: the tractor carried an unknown load in the field.

engine speed [rpm]; engine shaft rotational speed;

vehicle speed [km/h]; tractor linear velocity.

engine torque [N m]; torque developed by ICE measured at engine shaft;

engine speed [rpm]; engine shaft rotational speed;

PTO torque [N m]; torque measured at PTO shaft;

PTO speed [rpm]; PTO shaft rotational speed;

vehicle speed [km/h]; tractor linear velocity.

engine torque [N m]; torque developed by ICE measured at engine shaft;

engine speed [rpm]; engine shaft rotational speed;

vehicle speed [km/h]; tractor linear velocity.

a remaneance Brem = 0.40 [T ];

an intrinsic magnetic field Hcj = 450 [kA/m];

Stator back iron: