Minimum Average Fraction Inspected For CSP-M Plan: Chung-Ho Chen and Min-Tsai Lai

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

Tamkang Journal of Science and Engineering, Vol. 9, No 2, pp.

151-154 (2006) 151

Minimum Average Fraction Inspected for CSP-M Plan


Chung-Ho Chen1* and Min-Tsai Lai2
1
Department of Industrial Management, Southern Taiwan University of Technology,
Yung-Kang, Taiwan 710, R.O.C.
2
Department of Business Adminstration, Southern Tainan University of Technology,
Yung-Kang, Taiwan 710, R.O.C.

Abstract
In this paper, a simple method is developed to find the unique combination (i, f) that will meet
the average outgoing quality limit (AOQL) requirement, while also minimizing the average fraction
inspected (AFI) for a CSP-M plan with specified number of inspection levels when the process average
p(> AOQL) is known.

Key Words: Multi-Level Continuous Sampling Plan (CSP-M Plan), Average Outgoing Quality (AOQ),
Average Outgoing Quality Limit (AOQL), Average Fraction Inspected (AFI)

1. Introduction 4.If i sampled units are found free of defects, increase


k by one and go to step 3. However, k must not ex-
The continuous sampling plan can be used in the some ceed k0, that is k £ k0.
production processes when the lots are not clearly defined. 5.If a defective is found, decrease k by one and go to
An example might be welded leads emanating from a step 3. If k = 0, go to step 2.
welding operation. The continuous sampling plan was
first proposed by Dodge [1] which called the CSP-1 plan. From Schilling [5], the advantages of the CSP-M
Dodge and Torrey [2] improved CSP-1 plan and intro- plan are as follows: (1) it allows for smoother transi-
duced the CSP-2 and the CSP-3 plan. In 1955, Lieberman tion between sampling inspection and 100% inspec-
and Solomon [3] presented the theory of multi-level con- tion; (2) it requires 100% inspection only when the qu-
tinuous sampling plan (CSP-M plan) which reduces the ality submitted is quite inferior; (3) it allows for a min-
sampling frequency as successively more product is pas- imum amount of inspection when quality is definitely
sed without finding a defective. In 1957, Derman et al. [4] good.
extended the CSP-M plan as tightened multi-level plan The inspection cost is proportional to the average fra-
(MLP-T plan). The MLP-T plan is simply a CSP-M plan ction inspected (AFI) for a continuous sampling plans.
which requires a switch all the way back to 100% inspec- The average outgoing quality limit (AOQL) is one of the
tion, at any level, whenever a defective unit is found. indices to measurement the performance of the general
From Schilling [5], the procedure of CSP-M plan al- continuous sampling plans.
lowing the possibility of infinite levels is as follows: Lieberman and Solomon [3] presented the method of
1.Specify the clearing interval i, the initial sampling minimum AFI for CSP-M plan with k = 1 and k®¥. How-
frequency f, and the maximum number of inspec- ever, Lieberman and Solomon [3] didn’t propose the prob-
tion levels k0. lem of minimum AFI for CSP-M plan with specified num-
2. Set k = 1 and begin 100% inspection. bers of inspection levels k. In this paper, a simple method is
3.After i units in succession have been found without developed to find the unique combination (i, f) that will
a defective, start at a rate of f k. meet the AOQL requirement, while also minimizing the
AFI for a CSP-M plan with specified number of inspection
*Corresponding author. E-mail: chench@mail.stut.edu.tw levels when the process average p (> AOQL) is known.
152 Chung-Ho Chen and Min-Tsai Lai

2. Mathematical Model and Solution By integrating Elliff and Foster’s [6] and Lieberman
Procedure and Solomon’s [3] results, Eqs. (5)-(6) can be rewritten
as
Elliff and Foster [6] presented the straightforward minimize
equation defining AFI as a function of the primary pa-
rameters of the CSP-M plan. From Elliff and Foster [6], k
f k å q ij (1 - q i ) k - j
we have j =0
AFI = k (7)
k å f k - n q in (1 - q i ) k - n
f k å q ij (1 - q i ) k - j n=0

j =0
AFI = k
(1) subject to
åf
n=0
k -n i k -n
q (1 - q )
in

1 1
1 1
where p is the probability of a unit being defective (it f = f1 ( ) 3 + f 2 [1 - ( ) 3 ] (8)
k k
can be estimated by the process average p when the pro-
cess is in control); q = 1-p; i is the length of the consecu- i ³ 1, integer
tive non-defective run required before the CSP-M plan 0£f£1
shifts to the next inspection level; f is the initial sam- where
pling frequency; k is the number of inspection levels.
Lieberman and Solomon [3] adopted the harmonic (1 - pL )i
f1 = (9)
cube root interpolation for obtaining the initial sampling 1 p
(1 - pL )i + (1 + )i (1 + i ) L
frequency f of specified AOQL for any fixed number of i 1 - pL
inspection levels k. From Lieberman and Solomon [3, p.
701], we have (1 - pL )i
f2 = (10)
1 - (1 - pL )i
1 1
1 1
f = f1 ( ) 3 + f 2 [1 - ( ) 3 ] (2) The incoming defective p can be estimated by the
k k
process average p when the process is in control. For the
where pL is the specified AOQL; given parameters k, p, and pL, we can adopt the direct
search method for obtaining the optimal combination (i,
(1 - pL )i f). The solution procedure is as follows:
f1 =
1 p (3) Step 1.
(1 - pL )i + (1 + )i (1 + i ) L
i 1 - pL Estimate the process average p.
Step 2.
(1 - pL )i Search the minimum i value with minimum AFI.
f2 = (4) (i)Let i = 1. Substituting this i into Eqs. (7)-(8) to find
1 - (1 - pL )i
the corresponding f and AFI.
Hence, the problem of minimizing AFI for a CSP-M (ii)Repeat (i) of Step 2 with i replaced by i+1. Then repeat
plan with specified number of inspection levels is as follows: (i) of Step 2 with i replaced by i+2, etc. Terminate the
Minimize procedure when it is obvious that a global minimum
AFI has been found. The optimal solution is (i*, f*).
AFI (5)

max AOQ = pL (6) 3. Numerical Example and Sensitivity Analysis


0 £ p £1

i ³ 1, integer 3.1 Numerical Example


0 £ f £1 Assume that the process is in control, the process av-
Minimum Average Fraction Inspected for CSP-M Plan 153

erage p = 0.04, pL = 0.02 and k = 3. By solving the above Table 3. Effect of the process average ( k = 5, p L = 0.02)
Eqs. (7)-(8), we have i* = 42 and f* = 0.3360331. Table 1
p i f AFI
lists the optimal solution for CSP-M plan with specified
number of inspection levels k. From Table 1, we obtain 0.025 69 0.1723433 0.1198891
0.030 51 0.2953483 0.2856967
that the AFI increases with k.
0.035 44 0.3733267 0.4556394
0.040 41 0.4152558 0.5948373
3.2 Sensitivity Analysis 0.045 39 0.4468639 0.6977808
1.Effect of the specified average outgoing quality li- 0.050 38 0.4639269 0.7715461
mit pL
For a given set of parameters k and p, it is obvious
that i increases and AFI decreases with pL (as shown in For a given set of parameters k and pL, it is obvious
Table 2). that i decreases and AFI increases with (as shown in Ta-
2.Effect of the process average p ble 3).

Table 1. Values of parameters for given k 4. Conclusions


( p L = 0.02, p = 0.04)
Lieberman and Solomon [3, p. 686] point out that
k i f AFI
one of the primary purpose of CSP-M plan is to allow a
1 48 0.1235259 0.500001 minimum amount of inspection when quality is defini-
2 43 0.2666184 0.5223535
3 42 0.3360331 0.5418016 tely good. Usually, we use CSP-M plan with the small
4 41 0.3874367 0.5677643 number of inspection levels in order to avoid scheduling
5 41 0.4152558 0.5948379 of workloads difficult. Hence, it is important that the
6 42 0.4211058 0.6196808 problem of minimizing AFI for a CSP-M plan with spec-
7 42 0.4374989 0.6408135 ified number of inspection levels. In this paper, we have
8 43 0.4353581 0.6582727
presented that the above problem for a CSP-M with spec-
9 44 0.4311101 0.672568
10 45 0.425421 0.684354 ified number of inspection levels when the process aver-
11 45 0.4332161 0.6941821 age p(> AOQL) is known. Further study will emphasis
12 46 0.4254196 0.702463 the economic design of CSP-M plan with other inspec-
13 46 0.4313957 0.70963 tion cost.
14 47 0.4223845 0.7157925
15 47 0.4271301 0.7212586
16 48 0.4174177 0.7260935
References
17 48 0.4212862 0.7304108
18 48 0.4248625 0.734332 [1] Dodge, H. F., “A Sampling Inspection Plan for Contin-
19 49 0.4144062 0.7378851 uous Production,” Annals of Mathematical Statistics,
20 49 0.4174058 0.7411015
Vol. 14, pp. 264-279 (1943).
[2] Dodge, H. F. and Torrey, M. N., “Additional Continu-
Table 2. Effect of the average outgoing quality limit ous Sampling Inspection Plans,” Industrial Quality Con-
( k = 5, p = 0.04)
trol, Vol. 7, pp. 7-12 (1951).
pL i f AFI [3] Lieberman, G. J. and Solomon, H., “Multi-level Con-
0.010 69 0.5363110 0.9412744 tinuous Sampling Plans,” Annals of Mathematical Sta-
0.015 50 0.4750274 0.8068521 tistics, Vol. 26, pp. 686-704 (1955).
0.020 41 0.4152558 0.5948379
[4] Derman, C., Littauer, S. and Solomon, H., “Tightened
0.025 38 0.3296439 0.3569569
0.030 40 0.2210629 0.1718522 Multi-level Continuous Sampling Plans,” Annals of
0.035 52 0.0944231 0.0643128 Mathematical Statistics, Vol. 28, pp. 395-404 (1957).
0.039 1090 0.0061388 0.0187375 [5] Schilling, E. G., Acceptance Sampling in Quality Con-
154 Chung-Ho Chen and Min-Tsai Lai

trol, Marcel Dekker, Inc. (1982). Research, Vol. 13, pp. 423-425 (1975).
[6] Elliff, G. A. and Foster, J. W., “A Note on Calculation
of the Average Fraction Inspected for a Continuous Manuscript Received: Oct. 8, 2004
Sampling Plan,” International Journal of Production Accepted: Apr. 27, 2005

You might also like