Analysis and Management

of Production System
Lesson 11: Variability of processing time

Prof. Giulia Bruno

Department of Management and

Production Engineering

[email protected]
Variability

• Definition: not uniformity of a class of entities

• It causes a departure from regularity and predictability of the
system behaviour

• Deterministic or stochastic:
 Worst case: completely predictable variability (consequence
of an inappropriate management)
 Practical worst case: not predictable variability (due to
randomness)
Variability

• Examples of variability sources:

 Machine failures
 Setups
 Material shortages
 Yield loss
 Rework
 Operator unavailability
 Workspace variation
 Different skill levels
 Material handling
 Product variety
Variability

• Stochastic variability is studied through probability

• Random variable: consequence of events out of our immediate
control (we don’t know in advance when a machine could fail)
• A random variable is characterized by:
1. Type pf distribution: Exponential, Normal, Poisson’s,….
2. Parameters:
 Mean value 𝜇: most probable value
 Variance 𝜎 2 : values distribution with respect to the mean value
 Skewness: bend top position with respect to the mean value
 Kurtosis: how bend seems sharp
Variability

Random variable:
• The “quantity” is measured statistically through the
mean value 𝜇
• The “variability” is measured statistically through the
variance 𝜎 2 and the standard deviation 𝜎
• To classify and to quantify the variability of a random
variables, dimensionless coefficients are used:

1. Coefficient of Variation (CV):

2. Squared coefficient of variation (SCV):

Variability

To allow the comparison, systems are classified according to the

level of variability expressed through the coefficient of variation:
• Low variability (LV): c < 0,75
• Moderate variability (MV): 0,75 < c < 1,33
• High variability (HV): c ≥ 1,33

Relation to performance cases for balanced systems:

• MV – Practical Worst Case
• LV – between Best Case and Practical Worst Case
• HV – between Practical Worst Case and Worst Case
Variability

Mean Value = 25 min

StDev = 5.65 min
CV = 0.226

Mean Value = 73.6 min

StDev = 8.4 min
CV = 0.114
Variability sources in a process

The main variability causes for process times are:

1. Natural variability
2. Variability due to not planned interruptions (pre-emptive
outages: out of orders, breakdowns)
3. Variability due to planned interruptions (non pre-
emptive outages: setups, tools changes)
Natural variability

• Definition: variability related to the natural process

• not explicitly related to a cause, but implicit in the natural
process
• Natural variability of process time is usually due to
human factors, but also to others sources as operator
skill, product type variation, product quality
• Natural variability is, usually, of LV type:
𝜎0
𝑐0 = < 0,75
𝑡0
where 𝜎0 and 𝑡0 represent the mean value and the standard
deviation of natural process time respectively
Pre-emptive outages (failures)

• Definition: variability of process time due to failures,

which can happen at each moment (break of a tool,
lack of consumable parts,….) and cause an
interruption to the process
• Failure features:
 mf = mean time to failure (MTTF), or also called mean time
between failures (MTBF), it is the mean time between two
failures
 cf = coefficient of variability of time between failures (we
assume cf = 1)
• Repair features:
 mr = mean time to repair (MTTR)
 cr = coefficient of variability of the time to repair a broken
machine
Pre-emptive outages (failures)
• Failures change the machine availability which is given by:

 Availability is the long-run average fraction of time in which the

machine is available for processing jobs
• Machine availability changes the natural mean process time,
so that the effective mean process time (> natural process
time) is given by:

• The variance of the effective process time is given by:

Natural variability Variability due to failures repair time

Pre-emptive outages (failures)

• The quadratic coefficient of variation of

effective process time is given by:

Natural variability Variability due Variability due

to repair time to repair time
variability
Pre-emptive outages (failures)
Pre-emptive outages – Example

Let’s consider a line consisting of 2 machines (M1, M2) with

mean process time t0 = 15 min and natural standard deviation
equal to 𝜎0 =3,35 min. Both machines undergoes failures
which limit their availability (but in different ways according to
the following table). Which machine is to be preferred?

Variation
Machine Failure type MTTF MTTR
coefficient

Long and
M1 744 min 248 min 1
rare

Short and
M2 144 min 38 min 1
frequent
Pre-emptive outages – Example

• Machine availability depends on mf and mr

• Correcting natural mean process time with the availability term of
the machine, it is possible to obtain the effective process mean
time:
A = 0,75 te = 15/0,75 = 20 min
• Effective capacity of a workstation:
re = 1/te = A/t0 = A*r0 = 0,75 * a pcs/hr = 3 pcs/hr
• The two machines seem to be equivalent by using the mean
values both when it comes to as, to te and to re, but we also need
to consider variability:

M1:
M2:
Pre-emptive outages – Example

Pre-emptive maintenance procedures allow to convert

long and rare breaks into more frequent short pauses,
obtaining, in this way, a lower variability
Not Pre-emptive outages (setups)

• Definition: variability of process time due to setups

(e.g., tools substitution or planned maintenances)
• They allow to conclude the job process before
switching off the machine
• Features:
 Ns = average number of items processed by the machine
between one setup and the other
 ts = mean setup time
 cs = coefficient of variation of setup time
• It is assumed that the probability to run a setup after
each piece is the same
• The setup increases the effective process time of each
job by the quantity ts/Ns
Not Pre-emptive outages (setups)

• Effective process time is given by:

Probability that the setup won’t Probability that the setup will
take place after the current job take place after the current job

• Variance of effective process time is given by:

Natural variability Setup Interaction between the

time different setups
variability

• The quadratic variation coefficient of effective

process time is given by:
Not Pre-emptive outages (setups)
Not Pre-emptive outages – Example

Let’s consider the machines M1 and M2. M1

doesn’t need setups, while M2 does. Data are
displayed in the following table. Which machine
is to be preferred?

t0 c0 Ns ts cs

M1 1,2 h 0,5 / / /

M2 1,0 h 0,25 10 pcs 2h 0,25

Not Pre-emptive outages – Example

So, the first machine is preferable to the second one,

even if it has greater t0 because it has a lower variability
Variant 1

t0 c0 Ns ts cs

M1 1,2 h 0,5 / / /

M2 1,0 h 0,25 5 pcs 1h 0,25

Variant 1

The second machine is preferable to the first one

because it has a lower variability
Variant 2

t0 c0 Ns ts cs

M1 1,2 h O,5 / / /

5 pcs 1h 0,25
M2 1,0 h 0,25
1 pc 6 min 0,25

The second machines undergoes a second setup

operation, executed on each piece and which
lasts 10 minutes
Variant 2
Summary table
Exercise

In a system for the construction of electrical circuits and boards, the mean
time of exposition of a single panel is 2 minutes with a standard deviation
equal to a minute and a half (t0=2 min, 𝜎0=1.5 min)
a) Compute the process natural coefficient of variation (CV)
b) Compute the effective time, variance, and quadratic variation
coefficient (SCV), supposing the machine undergoes failures and that
distributions of mean time to failures and mean time to repair are
exponential (cr=1), respectively with mean value 60 and 5 minutes
c) Compute the effective time, variance, and quadratic variation
coefficient (SCV), supposing the machine undergoes planned
interruptions (e.g. setup, pre-emptive maintenance), which lasts 5
minutes, has CV=1 and repeat every 30 pieces
d) Compute the effective time, variance, and quadratic variation
coefficient (SCV), supposing the machine undergoes both detractors
described at points b and c
Exercise 1a – Natural Variability

● In a system for the construction of electrical circuits and

boards, the mean time of exposition of a single panel is 2
minutes with a standard deviation equal to a minute and a half
(t0=2 min, 𝜎0=1.5 min)
a) Compute the process natural coefficient of variation (CV)

Formula:

Solution:
Exercise 1b – Failures

• In a system for the construction of electrical circuits and boards,

the mean time of exposition of a single panel is 2 minutes with a
standard deviation equal to a minute and a half (t0=2 min, 𝜎0=1.5
min)
b) Compute the effective time, variance, and quadratic variation
coefficient (SCV), supposing the machine undergoes failures and
that distributions of mean time to failures and mean time to repair
are exponential (cr=1), respectively with mean value 60 and 5
minutes

Formulas:

Solutions:
Exercise 1c – Setup

• In a system for the construction of electrical circuits and

boards, the mean time of exposition of a single panel is 2
minutes with a standard deviation equal to a minute and a half
(t0=2 min, σ0=1.5 min)
c) Compute the effective time, variance, and quadratic variation
coefficient (SCV), supposing the machine undergoes planned
interruptions (e.g. setup, pre-emptive maintenance), which lasts
5 minutes, has CV=1 and repeat every 30 pieces

Formulas: Solutions:
Exercise 1d – Failures and setup

● In a system for the construction of electrical circuits and boards,

the mean time of exposition of a single panel is 2 minutes with a
standard deviation equal to a minute and a half (t0=2 min,
𝜎0=1.5 min)
e) Compute the effective time, variance, and quadratic variation
coefficient (SCV), supposing the machine undergoes both
detractors described at points b and c

Formulas:

Solutions:
Exercise 1d – Setup and failures

● In a system for the construction of electrical circuits and boards,

the mean time of exposition of a single panel is 2 minutes with a
standard deviation equal to a minute and a half (t0=2 min,
𝜎0=1.5 min)
e) Compute the effective time, variance, and quadratic variation
coefficient (SCV), supposing the machine undergoes both
detractors described at points b and c

Formulas:

Solutions:
Painting Station

Compute the effective processing time of a cars painting station.

• Each car is arranged for the painting, when some components
are disassembled and surfaces to be painted are prepared. The
time required for this operation follows a normal distribution with
mean value equal to 116 minutes and a standard deviation of 30
minutes.
• The time require for the painting follows a normal distribution
with mean value of 120 minutes and standard deviation of 12
minutes
• Once a year (composed of 40 working weeks, each consisting of
5 days of 8 hours) a failure occurs on the drying system and it
requires the substitution of one ore more components. Such
machine stops requires on average 3 work hours with a
standard deviation of 20 minutes (according to a normal
distribution)