2nd Symposium on Lift and Escalator Technologies

The modelling, simulation and experimental testing of vertical

vibrations in an elevator system with 1:1 roping configuration

Xabier Arrasate1, Stefan Kaczmarczyk2, Gaizka Almandoz3, José M. Abete4

and Inge Isasa5
1
Applied Mechanics, Mondragon Unibertsitatea, Loramendi, 4, 20500 Arrasate/Mondragón,
Euskadi, Spain, [email protected]
2
Department of Engineering, School of Science and Technology, University of Northampton, St.
George’s Avenue, Northampton NN2 6JD, UK, [email protected]
3
Electric Drives, Mondragon Unibertsitatea, Loramendi, 4, 20500 Arrasate/Mondragón, Euskadi,
Spain, [email protected]
4
Applied Mechanics, Mondragon Unibertsitatea, Loramendi, 4, 20500 Arrasate/Mondragón,
Euskadi, Spain, [email protected]
5
Mechanical Engineering, Orona EIC S. Coop. Elevator Innovation Center, Pol. Lastaola s/n,
20120 Hernani, Euskadi, Spain, [email protected]

ABSTRACT
Vertical vibrations affect passenger comfort during an elevator travel. In this work the results of a
study to investigate the vertical vibrations caused by torque ripple generated at the drive system and
transmitted through the suspension ropes to the car is presented. The acceleration response at the
suspended masses and at the drive machine end in a laboratory rig are measured during the system
travel. The machine torque ripple and the radial forces generated at the machine air-gap between the
stator and the rotor are computed using the Finite Element Method simulation with the software
FLUX. The torque ripple excitation is then accounted for in a non-stationary lumped-parameter
model of an elevator system with 1:1 roping configuration. The model accommodates the dynamics
and control of the drive system. The model is implemented in the MATLAB/Simulink
computational environment and the dynamic response of the system during travel is determined
through numerical simulation. Both computer simulation and experimental results demonstrate that
the vibration generated at the machine is transmitted to the elevator car becoming magnified if the
excitation frequency is close to the natural frequencies of the system.

INTRODUCTION
One of the main problems in elevator installations is to achieve and to maintain adequate ride
quality standards. Car ride quality can be compromised by excessive vibrations. Hoist ropes due to
their flexibility and loading conditions are particularly affected. The drive machine is the source of
energy supplied to the system but it is as well a source of vibration, caused by imbalance of the
machine, eccentricity of the traction sheave or the torque electromagnetic ripple.

Vertical vibration in elevator systems has been studied in a number of papers. Regarding the
mechanical part, the elevator system is usually modelled as a translating assembly of inertia
elements coupled and constrained by one-dimensional slender continua. The inertial elements are
the rotating components of the machine, the car assembly and the counterweight. The one-
dimensional slender continua are the suspension ropes. They are usually divided into sections,
lumping the mass of each section at corresponding discrete points and joining the mass points by
springs and dampers with the corresponding stiffness and damping properties. Those models are
described by a set of ordinary differential equations. The number of discrete mass points determines
the number of degrees of freedom (DOF) of the system and the number of natural frequencies and
mode shapes that need to be considered [1]. This approach enables one to simulate the vertical
vibration during a travel by updating the models at every time step. Various strategies to suppress
vertical vibration have been tested with time-varying models of this sort [2].

In this work investigation is carried out to identify the torque ripple excitation that is generated in a
synchronous motor used in a laboratory model of an elevator system with a 1:1 roping
configuration. The Finite Element Method (FEM) simulation tests with FLUX are conducted to
determine the magnitude and frequency of the corresponding harmonics. Then, the vibration
response at the main components of the laboratory rig is measured during travel. Furthermore, a
mathematical model of an elevator is developed in order to analyze the response of the system. This
model includes a comprehensive representation of the drive system. The mechanical part of the
system is represented by a model with 3 DOF: the vertical displacement of the car, the vertical
displacement of the counterweight and the angle of rotation of the traction sheave, measured relative
to the equilibrium configuration of the system. The model comprising a set of ordinary differential
equations is solved numerically to predict the response of the elevator components to the torque
ripple excitation generated at the drive system. Experimental and simulation results are then
compared and discussed.

THE LABORATORY SETUP

The schematic of the laboratory model and the experimental setup are shown in Fig. 1.

Fig. 1. Schematic diagram of the laboratory setup

Two equal rigid rigs, whose masses are mc = mw = 33 kg, are suspended at each side of the traction
sheave and guided vertically, together with the machine assembly with the traction sheave and the

9
diverting pulley. The suspended rig at the diverting pulley side will be referred to as the
counterweight, and that one suspended directly from the traction sheave as the car.

The lengths from the counterweight to the diverting pulley and from the car to the traction sheave
are lw and lc respectively. The sum lw + lc equals 8 m. Other approximate values of the system
parameters are I = 0.3385 kg·m2, R = 0.065 m, m = 0.095 kg/m and the product EA between the
Young’s modulus and the cross-section area of the rope is 106 N.

ELEVATOR SYSTEM MODEL

In order to describe and determine the dynamic response of the elevator system to any vertical
excitation originated from the drive system, an elevator model has been developed. The model is
composed of two main parts (see Fig. 2): the first part represents the drive system and the second is
the model of the vertical dynamics of the car-sheave-counterweight-suspension rope system.

Drive System Model

The input to the drive system (see Fig. 2) is a desired velocity profile and the output is the machine
torque. It comprises a permanent magnet synchronous motor powered via an inverter that supplies a
pulse width modulated (PWM) voltage. The motor shaft speed is controlled in order for the car to
follow a prescribed velocity profile ωm* to achieve good ride quality.

In order to simplify the analysis of three-phase synchronous machine of the setup, a mathematical
transformation known as the direct quadrature zero (dqo) transformation [3] is used. In the case of
balanced three-phase circuits, application of the dqo transform reduces the three AC quantities to
two DC quantities. Simplified calculations can then be carried out on these imaginary DC quantities
before performing the inverse transform to recover the actual three-phase AC results.

According to the reference frame transformation theory, the three phase variables, currents and
voltages, are transformed to the d-q reference frame, which is rotating at the stator current
frequency ωs , representing the fundamental frequency. In this setup, the d-q components of the
currents and the voltages are constant.

A well known vector control strategy oriented to the magnets flux has been implemented in the
computer simulation [4]. Such control scheme consists of two control loops: outer and inner loops.
In the outer control loop the speed of the motor is regulated by a conventional Proportional Integral
(PI) controller, which sets the torque reference τ * with the aim of minimizing the speed
error ωm* − ωm . In the inner control loop two PI controllers are implemented in order to regulate the
d-q axes currents id and iq . These two controllers set the d-q axes voltage references vd and vq in
order to minimize the current errors.

10
Fig. 2. The drive system, including the control diagram, the power converter and the electric motor

The q axis current reference iq is obtained directly from the torque reference τ * , because in
permanent magnet synchronous motors the relationship between the torque and the q axis current is
practically constant and known as the torque constant (see Eqn. (8)). The field weakening strategy is
not implemented so that the d axis current reference id is set to zero [5].
Therefore, the PI controllers try to regulate the d-q currents to constant reference values so that the
stationary error in the currents can be compensated by the integral action of the controllers.

Electric motor model

The machine is a 12 pole permanent magnet synchronous motor comprising a stator with 72 slots.
The electric model of the motor is described by two differential equations

dΨd
vd = Rsid + − pωm Lqiq (1)
dt
dΨ
vq = Rsiq + q − pωm (Ldid + Ψpm ) (2)
dt

where vd , vq are the d-q axes voltages, id , iq are the d-q axes currents, Ψd , Ψq are the d-q axes flux
linkages, Ψpm is the magnet flux linkage, that is constant, Rs is the stator resistance, Ld , Lq are the
d-q axes inductances, ωm is the mechanical speed and p is the number of pole pairs. From [3] it can
be deduced that the d-q axes flux linkages are defined as,

Ψd = Ld id + Ψpm (3)
Ψq = Lq iq (4)

Consequently,

did
vd = Rsid + Ld − pωm Lqiq (5)
dt
11
 − pωm (Ldid + Ψpm )
diq
vq = Rsiq + Lq (6)
dt

The torque τ generated by a motor is composed of two components: the average torque τ 0 and the
torque ripple τ r . In addition, the torque ripple is composed of another two components: the
electromagnetic torque ripple and the cogging torque. The electromagnetic torque ripple is mainly
due to the spatial distribution of stator windings and the magnets shape, while the cogging torque
depends mainly on the number of stator slots and pole pairs [6].
The average value of the torque produced by the motor can be modelled by the next equation.

p (Ψdiq − Ψqid )
3
τ0 = (7)
2

If the d axis current id is set to 0 the following results

3
τ0 = pΨpmiq (8)
2

Eqn. (8) shows that the average torque generated by the motor τ 0 is proportional to the q axis
current iq . The magnitude of the electromagnetic torque ripple τ r is also proportional to the q axis
current iq as it is demonstrated in [7]. Nevertheless, the magnitude of the cogging torque does not
depend on the current magnitude and it is constant, even when the motor is not supplied.
Then, the overall torque τ is computed as follows

pΨpmiq + Kτ k iq sin (kpωmt ) + ∆τ c sin (npωmt )

3
τ = τ 0 +τ r = (9)
2

where Kτ k is the torque constant for the main component of the electromagnetic torque ripple, of
order k , and ∆τ c is the magnitude of the main cogging torque component, whose order is n . The
main electromagnetic torque ripple and the cogging torque frequency values are respectively k = 6
and n = 12 times the fundamental frequency, given by ωs = ωm p , for the particular motor of the
laboratory setup. Those components of the torque ripple have been computed by the finite element
analysis (FEA).
In addition to the components of the torque ripple described, radial forces are generated at the air
gap between the stator and the rotor of the machine [8] that causes vibration of the stator core and
yoke. Those forces are decomposed in Fourier series, so that the inner surface of the stator is
subjected to several sinusoidally distributed loads. The main components in the series contribute as
well to the torque ripple, especially when the spatial order and frequency of the excitation force are
close to a stator mode shape and corresponding natural frequency respectively. The torque ripple
components are added to the machine torque generated at the drive system as a perturbation in the
MATLAB-Simulink model.

Vertical Vibration Model

The input to the vertical vibration model (see Fig. 2) is the machine torque and the output is the
machine shaft velocity. The assembly composed of the car, the counterweight, the sheave and the

12
ropes is represented by a 3 DOF model where the variable are the counterweight and car
displacements and the sheave rotation angle, xw , xc ,θ respectively. The ropes are divided into two
sections: the car-ropes subsystem and the counterweight-ropes subsystem. Both subsystems are
represented by a mass-less spring whose stiffness is calculated by Eqn. (10)

EA
k= (10)
l

where l is the length of the corresponding rope subsystem.

The ordinary differential equation set defining the dynamics is given by Eqn. (11).

M&x& + Cx& + Kx = Τ (11)

The inertia, damping and stiffness matrixes are given by Eqns. (12); the machine torque τ is
included in the vector T that together with vector x are given by Eqn. (13). The parameters k w , kc
correspond to the counterweight- and car-ropes subsystems respectively and are calculated
according to Eqn. (10) and updated every time instant during an elevator travel. The damping matrix
accounts for the friction at the guides-rails contact. The values of the coefficients cw , cc have been
calculated from the vibration decay observed when impacting the rig with a hammer. The
coefficient cs has been neglected.

mw 0 0 cw 0 0   kw 0 kw R 
M =  0 mc 0; C =  0 cc 0 ; K =  0 kc − kc R  (12)
 0 0 I   0 0 cs  k w R − kc R (kc + kw )R 2 
T = [0 0 τ ] ; x = [xw xc θ ]
T T
(13)

The three natural frequencies calculated by solving the eigenvalue problem for the case of the
laboratory setup as a function of the counterweight-rope subsystem length lw are shown in Fig. 3
(natural frequency is 0 Hz and corresponds to the overall transport motion of the system).
35

30

25
Frequency [Hz]

20

15

10

0
1 2 3 4 5 6 7

Length [m]

Fig. 3. The natural frequencies of the setup

13
COMPUTER SIMULATION AND EXPERIMENTAL TESTS
The dynamic behaviour of the elevator system will be simulated by employing the model developed
above and the results will be compared to those obtained from the experimental tests.

Computation of the motor parameters by Finite Element Analysis

The electric motor model uses some characteristic parameters such as inductances, resistance,
magnet flux linkage and torque constants. All those parameters of the motor have been computed by
Finite Element Analysis (FEA) with the software FLUX. In Fig. 4 1) the simulated motor geometry
along with the mesh distribution are shown. As the motor consist of p = 6 pole pairs, the simulation
geometry can be simplified to one pole pair domain. In Fig. 4 2) the magnetic field induced by the
magnets around the geometry of the motor is shown. Table 1 summarises the main motor
parameters computed by FEA simulations.

Fig. 4. 1) Mesh distribution, 2) Spatial distribution of the magnetic field induced by the magnets

Coil resistance Rs 0.33 Ω

d-axis inductance Ld 8.5 mH
q-axis inductance Lq 11.7 mH
Magnet flux linkage Ψpm (rms) 0.83 Wb
Average torque Kτ 10.44 N·m/A
constant
Torque ripple Kτ 12 0.1354 N·m/A
constant
Cogging torque ∆τ c 1.25 N·m/A
magnitude

Table 1. Parameters computed by FEA simulations

Apart from those components of the torque ripple considered, additional harmonics appear due to
the radial magnetic forces between the stator and the rotor generated at the air-gap. The radial
magnetic force per unit area or magnetic pressure waveform at any point of the air gap is obtained
by means of the Maxwell’s stress tensor theorem [8] given by Eqn. (14).

pr (θ , t ) =
1
2µ0
(B (θ , t ) − B (θ , t ))
2
n
2
t (14)

where θ is the rotation angle with respect to the axis of symmetry of the machine, µ0 is the
magnetic permeability, t is the time, and B n and B t respectively the normal and the tangential
components of the magnetic field around the air-gap. Fig. 5 a) shows the waveform of the magnetic

14
radial pressure at a certain point of the stator core as a function of the rotor position. The spatial
period of this signal is pi/6 for symmetry reasons. Fig. 5 b) shows the components (spatial orders) in
the corresponding Fourier series. The highest component corresponds to the spatial order 0 and it is
a constant pressure; the spatial order 2 is a sinusoidal pressure distribution of spatial period pi/6 and
corresponding excitation frequency twice the fundamental one 2ωs and it is the main harmonic of
the radial force.
Simulations and tests
Three accelerometers, each with its corresponding charge amplifier (see Table 2) have been placed
on both masses and on the machine, as it is shown in Fig. 1.

Acquisition system B&K Pulse

Accelerometers B & K 4371, s.n.
1573419
Charge amplifiers B & K 2635, s.n.
1602883

Table 2. Details of the data acquisition and measurement system

5 5
x 10 x 10
9

a) 4.5 b)
8

7
Force per Unit Area [N/m ]

Force per Unit Area [N/m ]

3.5
2

6
3

5
2.5

4
2

3
1.5

2
1

1 0.5

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 5 10 15 20 25 30

Theta [Rad] Spatial order

Fig. 5. a) Magnetic radial pressure at a certain point of the stator core as a function of the rotor
position and b) components (spatial orders) of the radial pressure in the Fourier series

Fig. 6 a) and b) show the variation of the car-side rope length lc and the velocity profile of the
elevator model during a travel. The model has been implemented in MATLAB-Simulink and the
vibration during a travel simulated. The ripple components expressed in Eqn. (9) with values given
by Table 1 have been included. According to Fig. 5 b), excitation harmonic force due to 2nd order
component in the radial force has been added as well, as it could arise due to any eccentricity in the
motor. The magnitude of this component is uncertain and has been given an arbitrary value similar
to the amplitude of the cogging torque (see Eqn. (9)). At the setup rated speed of roughly 0.4 m/s the
fundamental frequency ω s is around 6 Hz.

Fig. 7 a) and b) show the measured (red) and the simulation (blue) accelerations of the car and the
counterweight respectively during the travel. The recorded signals have been sampled at 512 Hz.

15
 6.5

5.5

Length [m]
5 a)
4.5

3.5

2.5
0 1 2 3 4 5 6 7 8 9 10 11
Time [s]

0.6

0.5

Velocity [m/s]
0.4 b)
0.3

0.2

0.1

-0.1
0 1 2 3 4 5 6 7 8 9 10

Time [s]

Fig. 6. Evolution of the a) car-side rope length and b) the velocity

5
Test
Acceleration [m/s]

Simulation
a)

-5
0 1 2 3 4 5 6 7 8 9 10 11

Time [s]

5
Test
Acceleration [m/s]

Simulation
b)

-5
0 1 2 3 4 5 6 7 8 9 10 11

Time [s]

Fig. 7. Measured and simulated acceleration of the a) car and the b) counterweight

Figs. 8, 9 and 10 show spectrograms of the accelerations measured by the sensors placed on the
machine, the car and the counterweight respectively, during the constant velocity stage. The
spectrograms, in dB/Hz (dB relative to the reference value of 1 m/s2), have been calculated by the
Burg algorithm [9].
120
-10

-20
100

-30
Frequency [Hz]

80

-40

60 -50

-60
40

-70

20
-80

-90
0
2 3 4 5 6 7 8

Time [s]

Fig. 8. Evolution of the PSD of the acceleration measured on the machine

Figs. 11 and 12 show the evolution of the PSDs of the simulated accelerations of the car and the
counterweight respectively.
Discussion on experimental and simulation results
The results shown in Fig.7 demonstrate that the acceleration response predicted by the simulation
model does not capture all measured vibration components and the vibration levels predicted by the
model are much smaller than the actual levels. This would indicate that the actual excitation has
16
been underestimated in the model. Furthermore, more mechanical degrees of freedom would have to
be applied to obtain a more accurate simulation model.

As described before, the main electromagnetic torque ripple and the cogging torque frequency
values are respectively k = 6 and n = 12 times the fundamental frequency, that is around 6 Hz at the
setup rated speed. Those frequencies are 36 Hz and 72 Hz respectively. Furthermore, the 2nd order
component of the radial force is around 12 Hz. All these frequencies appear in the spectrograms
presented in Figs. 8-11.

As expected, the resonance regions are evident in the spectrograms. This is manifested by the
presence of the frequency components corresponding to the excitation frequencies that are near the
natural frequencies of the system (see Fig. 3). At the frequency of 12 Hz, de amplitude of vibration
of the car decreases and that one of the counterweight increases during the travel (see Figs. 7, 9-12).
This is because in the 1st modal shape the amplitude of vibration of the rig suspended from the
longest rope part is higher than that one suspended from the shortest. The experimental results (see
Figs. 8-10) show as well that there is an excitation frequency at 24 Hz; it could correspond to the
component of spatial order 4 in the radial force (see Fig. 5 b)). Although its amplitude is not so high
in the FEM calculation, it could be amplified because it is close to the 2nd natural frequency (see
Fig. 3). The simulation results show another excitation around 108 Hz (18 times ω s ) that is not clear
enough in the experimental spectrograms (see Figs. 8-10).
120 0

-10
100

-20
Frequency [Hz]

80 -30

-40

60
-50

40 -60

-70

20
-80

-90
0
2 3 4 5 6 7 8

Time [s]

Fig. 9. Evolution of the PSD of the acceleration measured on the car

120
0

100

-20
Frequency [Hz]

80

-40

60

-60

40

-80

20

-100

0
2 3 4 5 6 7 8

Time [s]

Fig. 10. Spectrogram of the acceleration measured on the counterweight

17
 120
-20

-40
100

-60

80

Frequency [Hz]
-80

-100
60

-120

40
-140

20 -160

-180

0
2 3 4 5 6 7 8 9

Time [s]

Fig. 11. Spectrogram of the acceleration of the car predicted by the model

-20
120

-40

100
-60

80 -80
Frequency [Hz]

-100
60

-120

40
-140

20 -160

-180
0
2 3 4 5 6 7 8 9

Time [s]

Fig. 12. Spectrogram of the acceleration of the counterweight predicted by the model

CONCLUSIONS
Experimental tests confirm that the hoist ropes, due to their elasticity, are a key component of an
elevator system, as they transmit to the car, and even amplify, the excitation forces that originate at
the drive system. This leads to excessive vibrations at the car end that compromise ride quality of an
elevator system. The FEM model employed in the simulations provides an estimate of the machine
torque ripple and the radial forces generated at the machine air-gap between the stator and the rotor.
Those estimates can be introduced in a comprehensive drive system model of an elevator, where the
car-counterweight-sheave-ropes assembly is represented by a 3 DOF lumped parameter model. The
simulation results are compared to those obtained from the experimental tests performed using a
laboratory model. The comparison shows that the simulation model underestimates the excitation
forces and the dynamic response of the actual setup. Nevertheless, the simulation results help to
identify which the main frequency components of the excitation forces are. The dynamic model can
then be employed to predict the resonance regions and the dynamic behaviour of the system.

REFERENCES
[1] K. Nai, W. Forsythe, R. M. Goodall, “Modelling and Simulation of a Lift System” in IMA
International Conference on Control, September 1992
[2] S. R. Venkatesh, Y. M. Cho, J. Kim, “Robust control of vertical motions in ultra-high rise
elevators”, Control Engineering Practice, Vol. 10, pp. 121-132, 2002
[3] R.H. Park, “Two Reaction Theory of Synchronous Machines”, AIEE Transactions, Vol. 48, pp.
716-730, 1929.
18
[4] P. Vas, Sensorless Control and Direct torque Control, Oxford Science publications, 1998
[5] H. Liu, Z. Q. Zhu, E. Mohamed, Y. Fu, X. Qi, “Flux-weakening control of nonsalient pole
pmsm having large winding inductance, accounting for resistive voltage drop and inverter
nonlinearities”, IEEE Transactions on Power Electronics, Vol. 27 (2), February 2012
[6] Z.Q. Zhu, S. Ruangsinchaiwanich, Y. Chen, D. Howe, “Evaluation of Superposition Technique
for Calculating Cogging Torque in Permanent-Magnet Brushless Machines”, IEEE Transactions on
Magnetics, Vol. 42, 2006
[7] G. Almandoz, J. Poza, M. A. Rodriguez, A. Gonzalez, “Analytic Model of a PMSM
Considering Spatial Harmonics”, International Symposium on Power Electronics, Electrical Drives,
Automation and Motion SPEEDAM, 2008
[8] J. F. Gieras, C. Wang, J. C. Lai, Noise of polyphase Electric Motors, Taylor & Francis Group,
LLC, 2006
[9] J. G. Proakis, D. G. Manolakis, Digital signal processing, Prentice Hall, 1996.

19