Week 1

Lo1-2
1. Market Segments
- A set of customers who have similar utility function
2. Trades-o s
- The need to sacri ce one capability in orders to increase another one
3. Pareto Dominated
- Means that a lm’s product or service is inferior to one or multiple competitors on all dimensions
of the customer utility function
4. E cient Frontier
- The set of rms that are not Pareto dominated
5. Customer Utility Function

Lo2-2
Resource : A group of people and/or equipment that transforms inputs into outputs
Flow units : The basic unit that moves through a process, starting as input, and later leaving the
process as output
Inventory : Number of units within a process at any given time ( Units )
Flow Rate : The average rate at which ow units travel through a process ( Units/time)
Flow Time ( Throughput Time) : The time a ow unit spends in a process, from start to nish
( Weeks/hrs/mins/secs)
Demand Rate : No. Fun customers want per unit of time
Output rate ( Supply rate) : No. Fun supplied per unit of time
Steady State Equilibrium (SSE): FR(in)=FR(out)
1
Processing Time (p) : the time it takes a resources to process one FUN. p =
Ca p a bilit i esResouces
1
Cycle Time : C T = The time between completing two consecutive ow units. Cycle Time depends
FR
on process capacity
Lead Time : LT-Time between when order placed and received
Ava ilT i m e 1
Takt Time: = Ratio between time available and quantity to be produced to
Req u ir edQ t y FRDemand
serve demand
- Aim: CT→TT. CT < TT = producing faster than demand, waste
- CT > TT = producing slower than demand, lose pro ts.
1
Resource Capacity : = Max of resource =
p
ffi
fi
ff
fi
fi
𝐶
𝐹
𝑅
fl
fl
fi
fl
fi
 1 1
Process Capacity : = Max of process = = Also equal to min of all resource
Pma x PBottleneck
capacities. Also equal to bottleneck capacity.
FR Fl o wR a te
Utilisation = =
C Ca p a cit y
FRp Fl o wR a teProcess Fl o wR a teBot tl e n eck
Process Utilisation = = =
Cp Ca p a cit yProcess Pr o cessCa p a cit y
Little’s Law : Inventory = Flow Rate X Flow Time
FR↑ – pro t ↑ assuming demand is strong. Productive. Good for business. Usually higher Fixed
Cost (FC) but somewhat lower Variable Cost (VC)
FT↓ – if FR is constant, lower FT is better for business, but superbly expensive to achieve (eg
investments in super high end machines, expert consultants who know how to cut production steps,
etc). Usually FT↑ over
time or due to poor process design – bad.
WIP ↓ – current inventory cost ↓ (cost of writing-o all current WIP ↓) responsiveness ↑, so is
good for
business. Extreme: 0 WIP➔Just-In-Time.

Waste – Input resources used by system which contribute zero or negative value to product or end-
customer. E ciency ↓.
◦ Waste ↑➔pro t ↓, e ciency ↓, cost ↑, environmental risk ↑ (eg food→ pests,
factory→rust, overheating, explosion)
Variability – Fluctuations in and , and uctuations in processing times at
resource .
◦ Variability ↑➔waste ↑, pro t ↓, cost ↑
◦ In exibility – Unable to adapt to changes (or uctuations) in or . Usually due to
ine cient process (eg high WIP), high xed costs (costly machines, untrainable sta ).
E ectiveness ↓.
◦ In exibility ↑ to ➔ poor customer service ↑, revenue ↓, competitiveness ↓
◦ In exibility ↑ to ➔ product inavailability ↑, revenue ↓, competitiveness ↓
fl
fl
ff
ffi
fl
fi
ffi
𝑖
𝐹
𝑅
𝐼
𝐶
𝑛
fi
𝐹
𝑅
𝐷
𝐹
𝐹
𝑒
𝑅
𝑅
ffi
𝑚
𝑂
𝑎
𝑢
𝑛
𝑡
𝑑
fi
𝐹
𝑅
𝐷
𝑒
𝑚
fi
𝑎
𝑛
fl
𝑑
ff
fl
𝐹
𝑅
𝐼
𝑛
𝐹
𝑅
𝐷
𝑒
𝑚
𝑎
𝑛
𝑑
𝑝
ff
𝑖
 Week 2
Multiple Process - Serial Vs Parallel Process

Serial Process : Capacity = min 1 , 2, 3 , ie the min of 1 2 3 all resource capacities.

CR1 CR 2 CR 3
Parallel Process : m i n( , )
0.3 0.6 0.1
Time Through Empty Process = FT : Time for rst FUN to ow through an empty process i.e., for a
process that has no WIP
Time For Units At Output (Through Empty Process) = + −1 (ie time required to obtain
units at process output is the time to get rst unit plus time to get remaining ( − 1) units.
Cost of Direct Labour : The labour cost associated with serving one customer (unit: $/customer)
Cost
WagesPerUn ito f T i m e No . E m pl o yees * ( Hour )PerE m pl o yee
cos to f d ir ec tl a b our = =
Fl o wR a te Fl o wPerUn it T i m e

Labour Content : The amount of work that goes into serving one customer( Generally one FUN)
which is the sum of all processing time involving labor for each resource j across all n resources
Labour Content = Sum of the process times involving labour ( Add all p)

Total Idle Time : Flow Time * Number of Workers - Labour Content or Cycle time - Processing Time

l a b our Con te n t L a b our Con te n t

Average Labour Utilisation = Mean (Ui)= =
L a b our Con te n t + Tot a l Id l eT i m e n * CT

Choosing Sta Level : Matching supply with demand always starts with taking the demand rate as
given and attempting to sta a su cient number of resources in order to meet the demand
L a b our Con te n t
Target Manpower : = L a b our Con te n t * FRDemand Determining the minimum number
Ta k t T i m e
of resources required to meet the demand
O oading the Bottleneck : Reassign, Automate & Outsource

Week 3
Batch : A collection of FUNs such that no FUN can be taken out from batch until all FUNs in the
batch are processed.
Batch Size : The number of FUNs in a batch
Production Cycle - A repeating duration of sequence of activities to produce units. Can include: Step
up time, production time and idle time.
Setup Time -Non-processing time required for production of ow units (FUN) Is incurred for every
batch
Batch, Setup Time& Processing - Each Time a new batch is to be processed, setup time is incurred.
Each time a FUN is to be processed, processing time is incurred.
Capacity of Resources ( Machine Label ) CML is the manufactured capacity speci cation on a
machine label. For Human resource, the “ machine label” cannot be seen, but can be estimated
Processing Time (Of a Resource) For 1 FUN : p=1/CML
Tot a lUn it sPr o cessed B
Capacity With Setup S and Batch B : CBS = =
Tot a lT i m eTa k e n S + pB
ffl
𝑿
ff
𝐶
ff
ffi
𝐶
𝐶
fi
𝐶
fi
𝐹
𝑇
fl
𝑋
𝑅
fl
𝑅
𝐶
𝑅
𝑇
𝑋
fi
𝑋
 Setup Exists: We never get CML, but get CBS, which is perhaps 50% CML
1
Very Large Batching : When batch size goes in nite, it can be shown that CBS = = CML
p
Target Capacity : CTarget = Mi n(De m a n d R a te, Bot tl e n eck Ca p a cit y) = Mi n(FRDemand , CBottleneck )
CTarget * S
Once C target is determined, we determine the suitable batch size B using : B =
1 − (CTarget * P )
Setup Time and Capacity : When ↑, ↓.
Batch Size and Capacity : When ↑, ↑.
Setup Time and FT : When ↑, ↑. Batch Size and FT : When ↑, ↑.
Setup Time and WIP : ↑ ↑. So WIP ↑. Batch Size and WIP : ↑ ↑. So WIP ↑.
Setup Time and FR : When ↑, capacity ↓. remains the same if it is already below the lowered
capacity. Process crashes if current is higher than new lowered capacity.
Batch Size and FR : When ↑, stays the same. and are two separate convenient levers an
ops manager has over the process. Do not confuse tray feed-rate with –

FRPrevailing
Utilisation : With batch size and setup time, UResource = , Where FR prevailing is not higher
CML
CBS
than CBS. If FR prevailing is not stated, assume FR prevailing =CBS, which UResource =
CML
Since CBS<<CML, U resource is often <<1, resource are very under utilised
Setup Time and Utilization : When ↑, BS ↓ and so ↓
Batch Size and Utilization : When ↑, BS ↑ and so ↑
Max Inventory : Max Inventory in a production cycle =pB(CML − FR ) = B(1 − p * FR ))
Average Inventory =Max/2

Setup Time and Product Variety : When ↑, ↓ and so Product Variety ↓. Very expensive for
such process to produce large quantities. Should compete more on Volume rather than Variety.
When ↓, ↑ and so Product Variety ↑ even with small batches. May compete on both
Volume and Variety.

Setup Time and Idle Time: Setup time is NOT idle time. Setup time is more like “processing time”.
The qualitative and practical e ects are di erent.
◦ Resource (man or machine) is unproductive during setup, but some activities (eg preparation,
diagnosis, warm-up, etc) are occurring in order for resource to produce a batch.
◦ Resource is unproductive during idle and nothing else contributory in nature occurs.
◦ Setup disappears when Batch Size=0. Idle disappears when FR =
◦ Setup can NOT be reduced by changing FR, but idle can be reduced with higher FR or lower CT.
◦ Setup reduction is VERY expensive (change of machine, work ow, etc). Reducing idle is relatively
cheaper.

Reduce Setups – Always seek to reduce setup times

𝑆
𝐶
𝐵
𝑆
𝑆
𝐵
𝑆
𝑆
𝐹
ff
𝑇
𝐹
𝐹
𝐵
𝑅
𝑆
𝑇
𝐵
𝑆
𝐹
𝑅
𝐶
𝐶
𝐶
𝐵
𝐶
𝑆
𝐵
𝑆
ff
𝑆
𝐹
𝐶
𝑅
𝐵
fi
𝑆
𝐵
𝑈
𝑈
𝐹
𝑅
fl
𝐵
𝐶
𝐵
𝐹
𝐹
𝑇
𝑅
𝐹
𝑇
 ◦ SMED – Single-Minute Exchange of Dies. A goal to reduce the setup time to a single-digit
number
of minutes.
◦ Internal setups – Activities that can only be done during the actual setup; eg when the
machine is not producing. This is our usual sense of setup: prep machine, check con g, self-test,
etc.
◦ External setups – Activities that be done while production continues to occur. Setup time is
not removed, but overlapped concurrently with trailing processing times of previous resource, so that
when previous resource is done with a batch, current resource is already prepped and ready to
process. Eg, usually man-machine op setup, or resource-to-resource connectivity setup, etc.

Mixed-model Assembly – An assembly process in which production resembles the true rate of
demand, even over a short time horizon.

Heijunka – (平準化) Levelling production by reducing variation in the work schedule

that arises from either demand variation or the desire to run production or
transports in large batches.

Week 4
Queue Process : Made up of Boundary, Queue and servicing Station
Arrival Rate (avg) = , Flow Rate
1
Flow Time = Tq +
μ
Service Time (Avg) = , Capacity of server resource, CML
1
Inter-arrival Time (avg) = a= . Time between 2 consecutive customers arriving at a queue system
λ
1
Service time (avg) p = , Processing time of 1 FUN. Also the time customer spends with one server
μ
Unstable Queue : > m or equivalently a < p. An unstable queue cannot reach Steady State
Equilibrium ( SSE )
Stable Queue : < m or equivalently a > p. A queue system in which the demand rate is less than
capacity

For queue only, by little’s Law,

WIP = FR * FT, So Queue Inventory Iq = λ * p
For processing area only,
WIP = FR * FT. So, Ip = λ * p
For entire system, total inventory I = Ip + IQ
λ
I = λ * (Tq + P) = Iq + λp = Iq +
μ
𝜆
𝜆
𝜇
𝜆
𝜇
𝜇
fi
 General Process : arrival process in which inter-arrival times does not follow any
known distribution such as Exponential Distribution.

Sta n d ard De viat ion

Coe cient of Variation ( CV ) = CV =
Mea n

Stde vofAr r ivalT im e σα

CV of interval arrival time = CVa = =
Mea n Ar r ivalT im e α

Stde vof ProcessingT im e σρ

CV of service time = CVp = =
Mea n ProcessingT im e ρ

Poisson Process : Arrival (Service) process in which Arrival (Service) rates uctuate
according to Poisson distribution when mean rate (or ). The inter-arrival time
MUST follow exponential
For poisson Arrival Process, CVa=1
For poisson Service Process, CVp=1

p Utilization 2(m + 1)−1 CVa2 + CVp2

Time-in-Queue = Tq = *( )*( )
m 1 − Utilization 2

m= number of servers
p= service time CVa=CV of inter-arrival times, CVp=CV of service time

Utilisation
Utilisation (One Server) FlowRate 1/a λ
= =
Capacit y 1/p μ
Utilisation (m Servers) FlowRate p λ
= =
Capacit y a*m μm
WIP in Queue
# of waiting customers (avg in queue) Tq
Iq = λ * Tq =
a
p
# of customers in Service (avg in servers) Iq = λ * p =
a
# of Customers in Queue System (avg) Is = λ(Tq + p)

Queue Growth Rate Demand - Capacity

Length of queue at time T T * (Demand - Capacity)

ffi
𝜆
𝜇
fl
Time to serve Qth person in the queue Q/capacity

Time to serve the person arriving at time Dem a n d

T *( − 1)
T Capacit y

Average time to serve a unit 1 Dem a n d

*T *( − 1)
2 Capacit y

TC=Cs*m+Cw*I
Total Cost for Queue Owner=
Cost of service ($/hr) * number of servers + Cost of waiting ($/hr) * total
average WIP in Queue system