Escolar Documentos
Profissional Documentos
Cultura Documentos
1. The problem
A manufacturing process with an x control chart to monitor the process
The process starts with in-control and could become out-of-control after a while
due to assignable causes
The occurrence of assignable causes of the manufacturing process follows
Poisson distribution
The production is a continuous process and does not stop
The control chart has Types I and II errors in detecting the assignable causes
It has certain cost and time to sample the products
The decisions are the size of the sample, the control chart limits and the
frequency of sampling
The objective is to minimize the expected cost related to product quality and
quality control per unit time (hour) of production
The decision variables:
o h time interval in hours to take samples
o n the sample size
o k the parameter for control limit of the x chart as in 0 k / n
and 0 k / n
The parameters:
o the rate that the production system fails due to assignable causes.
The number of system failures in a certain time period follows
Poisson distribution (or the time that the process remains in control
follows exponential distribution with mean 1 / hour)
113
o The expected number of inspections during the in-control state can be
estimated by:
1
h
o probability of type I error (the false alarm) with
2 (k )
z z2
where ( z )
1
2 e
2
dz , the standard normal cumulative
distribution function
(k n ) (k n)
where indicates the magnitude of the shift caused by the assignable
cause so that the mean will be moved from 0 to 0 or to
0
o D the time needed to restore the process following an action signal
o g time to inspect 1 unit of the products, so the time to inspect a
sample of n units is gn
o the elapsed time between the last inspection in the in-control
period and the process shifts to out-of-control state
114
o The expected cycle time E (T ) can be found as:
1 h
E (T ) gn D
1
1* 1* 1* 1* 1*
h h h h
gn gn D
inspection
interval
sampling
time 3*
h /(1 )
4*
115
Other parameters:
o V0 income per hour of operation in in-control state
1 h a ' E (T )
E (C ) V 0 V 1 gn D a 3 3 ( a 1 a 2 n )
1 h h
Let a 4 V 0 V 1 , the penalty cost for out-of-control production, then the above
equation can be written as:
h a'
a4 gn D a 3 3
E ( A)
E (C ) a a2n
V0 1 1 h
or
E (T ) h E (T )
116
h h h 2 a '
a 4 gn D a 3 3
a a2n
E ( A) V0 1 1 2 12 h
V0 E ( L)
h E (T )
where
a1 a 2 n a4 h h h 2 a3 a 3'
E (L) gn D
h E (T ) 1 2 12 E ( T ) hE ( T )
and
1 h h h 2
E (T ) gn D
1 2 12
2. Example 9.5
117
The penalty cost for the process to operate in the out-of-control state is
$100/hour ( a4 100 )
With the given values of the parameters, the expected cycle time can be calculated
by:
1 h h h 2
E (T ) gn D
1 2 12
h h 0 . 05 h 2
20 0 . 0167 n 1
1 2 12
h h
21 0 . 00417 h 2 0 . 0167 n
1 2
or
h h
E ( T ) 21 0 . 00417 h 2 0 . 0167 n
1 2
The expected loss per hour can be calculated by:
a1 a 2 n a4 h h h 2 a3 a 3'
E (L) gn D
h E ( T ) 1 2 12 E ( T ) hE ( T )
1 0 . 1n 100 h h 0 . 05 h 2 25 50
0 . 0167 n 1
h E (T ) 1 2 12 E ( T ) 0 . 05 hE ( T )
Or
1 0 . 1n 100 h h 25 1000
E (L) 0 . 00417 h 2 0 . 0167 n 1
h E (T ) 1 2 E ( T ) hE ( T )
118
1 0 . 1n 100 h h 25 1000
min E ( L ) 0 . 00417 h 2 0 . 0167 n 1
n ,k ,h h E (T ) 1 2 E ( T ) hE ( T )
119
The sample size n is mainly affected by the shift, .
The penalty costs mainly affect the sampling frequency, h.
The costs related to looking for assignable causes mainly affect the width of the
control limit, k. They also affect the sample size n. As they increase, the chance
for false alarms reduces.
The sampling costs affect all design parameter.
The system failure rate due to assignable causes mainly affects the sampling
frequency, h.
h 2 h h 2 h
We can treat E (T ) as constant and let 0 . Then . If n
12 2 12 2
and k are given, then the optimal sampling interval h * can be found by letting
d
E ( L ) 0 to obtain:
dh
( a 1 a 2 n ) E ( T ) a 3' /
h*
a 4 1 1 12
There are other results under different assumptions as in W-H Chen and D.
Tirupati (International Journal of Quality & Reliability Management, Vol. 14,
pp. 234-259, 1997)
120