A new model for the representation of the ISO 2859 standard

George Nenesa,, Yiannis Nikolaidisb

a
Department of Mechanical Engineering,

University of Western Macedonia, Bakola & Sialvera, 50100 Kozani, Greece

b
Department of Technology Management,

University of Macedonia, 59200 Naoussa, Greece

Abstract

In order to carry out Acceptance Sampling, companies often use sampling schemes

recommended by easy-to-use quality standards. These standards, however, do not take quality

costs directly into account. This research proposes a new model for the representation of the

use of ISO 2859-1 (1999) standard. Using this new model, it is feasible to evaluate

economically the use of the standard. A real case example is presented and the quality cost of

the standard’s implementation is evaluated.

Key words: Quality Standard; ISO 2859 – 1; Acceptance Sampling; Quality Cost


Corresponding author. Tel.: +302461056665; fax: +302461056601
E-mail address: [email protected] (George Nenes).
Department of Mechanical Engineering, University of Western Macedonia, Bakola & Sialvera, 50100 Kozani,
Greece
1. Introduction

Over the last years quality issues and especially the cost aspects of quality are

increasingly gaining attention from the research community. According to Montgomery

(2001) quality cost may be as high as 40% of a company’s turnover and, therefore, its

accurate estimation, analysis and reduction (if possible) plays a major role to any company’s

development. In particular, quality cost includes all costs that result from imperfect systems,

processes and products, and it usually consists of prevention, appraisal, internal and external

failure costs. In many cases, each quality cost category can be expressed as a function of the

actual quality of products (e.g. fraction nonconforming of a lot of products). In these cases it

is possible to develop a total quality cost function, namely a sum of the partial quality costs,

and optimize it in order to find the economically optimum quality level of products.

This paper focuses on Acceptance Sampling (AS) and more specifically on single-

sampling by attributes. In order to perform AS and monitor the quality of acquired or

produced lots, companies often use sampling plans (n,c)1 that are recommended by specific

standards, which are simple and easy to use. The most popular international standard is ISO

2859 which in fact is a series of standards for various AS cases. For the purposes of this paper

we study the sampling schemes and procedures that are recommended by ISO 2859-1 (1999),

which will be mentioned hereafter for brevity as ISO 2859.

More specifically, this paper aims at determining the economic consequences of using the

ISO 2859 quality standard. To this end, we develop a new, Markov chain model for the

representation of the use of the sampling plans that are recommended by the ISO 2859 and

then by using this model we evaluate the economic impact of the standard on a company.

The remainder of the paper is structured as follows. The following section outlines the

related literature. In Section 3 we set the notation and we develop the expected total quality

1
n is the sample size and c is the acceptance number

1
cost function of a sampling plan. Section 4 presents in detail the model that represents the

integrated use of the ISO 2859 standard, while in Section 5 a real case study is presented. The

final section of the paper summarizes the main conclusions of our research.

2. Related Literature

Extensive references to economically optimum sampling plans can be found in Wetherrill

and Chiu (1975) and Wall and Elshennawy (1989). Tagaras and Lee (1987) use the Bayes’

theorem to develop an algorithm for the computation of the economically optimum single-

sampling plan. Ferrell and Chhoker (2002) propose an economic model for the design of AS

plans adopting the Taguchi approach. A year later, González and Palomo (2003) use a

Bayesian analysis in order to derive AS plans regarding the number of defects per unit and

apply their methodology to the paper pulp industry. In that paper the sampling plans are

obtained through the minimization of the expected total quality cost. At the same time,

Cassady and Nachlas (2003) define a generic framework for establishing three-level AS plans,

using quality value functions. They note that there are many cases in which the quality of a

product can be classified in three or more discrete levels; for example, a food product may be

classified as good, marginal or bad. Chen et al. (2004b) present and investigate a general

model of AS plan with exponentially distributed random censoring, based on Bayesian

decision theory. In order to determine the optimal sampling plan they consider a loss function,

which includes the sampling cost, the time-consuming cost and the decision loss to determine

the optimal AS plan. At the same time and in a similar study Chen et al. (2004a) develop a

general Bayesian framework for designing a variable AS plan with mixed censoring. A

general loss function including the three partial costs of Chen et al. (2004b) as well as the

salvage value is introduced to determine the corresponding optimal sampling plan. Stout and

Hardwick (2005) present a unified approach to the problem of response adaptive screening,

when multiple costs and constraints need to be incorporated. In particular, they describe a

2
cost- and constraint-based approach, which is suitable in AS where sample products must be

tested before an entire lot is accepted.

The forerunner of the present paper is some earlier work by Nikolaidis and Nenes (2009),

regarding the economic consequences of using the ISO 2859 standard. However, in that paper

the switching procedures that the ISO 2859 recommends are not taken into account. Thus, this

paper revisits the problem of the economic evaluation of AS procedures by extending that

earlier work in order to account for the complete usage of the standard, including the

switching procedures between normal, tightened and reduced inspection.

3. Cost functions of single-sampling by attributes

In general, single-sampling by attributes is carried out as follows: from a lot of N units, a

sole sample of n units is taken and every unit is checked and characterized as conforming or

nonconforming, based on an attribute-type quality characteristic. The lot is accepted or

rejected based on the outcome of the control and the maximum allowable number of

nonconforming items in the sample (acceptance number c). Regarding the economic elements

of AS by attributes, these are usually the inspection cost per item ci, the replacement/repair

cost cr and the cost per nonconforming item that is not detected during inspection cd.

The average total quality cost per lot that corresponds to the use of a single-sampling plan

(n,c) for the inspection of a lot with a fraction nonconforming p, is in general:

K  n,c p   [cost of accepting a lot]  Pa  p  +[cost of rejecting a lot]  1  Pa  p   , (1)

where Pa  p  is the probability of accepting a lot with a fraction nonconforming p. In

practice, every rejected lot is usually submitted to 100% inspection (rectifying inspection). In

this case, the analytical form of (1) becomes:

K  n ,c p    nci  npcr   N  n  pcd   Pa  p    Nci  Npcr   1  Pa  p   . (2)

When the fraction nonconforming p per lot is stochastic - let φ(p) be the pdf of its

3
distribution - the average total quality cost K  n ,c  is given by:

K  n ,c    K  n ,c p  φ p  dp .
p
(3)

Nikolaidis and Nenes (2009) have investigated numerically the effect of a variety of

parameters, cost elements and quality levels of lots on the economic behavior of the ISO 2859

sampling plans. They have presented some rules of thumb which ensure the minimization of

the economic loss that arises from using the plans recommended by ISO 2859 instead of the

economically optimum ones. The derivation of these rules, however, has not been based on

the integrated use of the standard, namely the use with the switching procedures between

normal, tightened and reduced inspection. Thus, their rules of thumb hold only for the case

where normal inspection is applied and the switching procedures are ignored in practice.

In the following section, we develop a Markovian model that can be used for the

complete representation of the ISO 2859 standard and, thus, when combined with the

aforementioned cost functions, it can be exploited for the economic evaluation of the

“integrated” use of the standard, which involves the consideration of the switching procedures

between normal, tightened and reduced sampling.

4. A model for the representation of the complete use of the ISO 2859 standard

In order to evaluate the use of the ISO 2859 standard under an economic viewpoint, it is

necessary first to understand the switching procedures of the standard and then create a model

for their representation. At the beginning of the standard implementation, the inspection

should be set to normal, unless otherwise directed by the responsible authority. The normal

sampling plan will be applied to successive lots until the switching procedure indicates that

the inspection should be more/less severe. In brief, if two out of five (or fewer) lots are

rejected, tightened inspection should be implemented. Normal inspection will be reinstated as

soon as five out of five lots are accepted during tightened inspection. In order to switch from

4
normal to reduced inspection, the “switching score” must be at least 30 (assuming that

reduced inspection is desirable by the responsible authority and the production rate is steady).

The switching score in sampling epoch i (ssi) is computed in the following way:

 Case 1 - c ≥ 2: ssi = ssi-1 + 3 if the lot is accepted and would also be accepted for one step

tighter AQL; otherwise reset the switching score to zero.

 Case 2 - c < 2: ssi = ssi-1 + 2 if the lot is accepted; otherwise ssi = 0.

Finally, normal inspection should be reinstated if a lot is rejected under reduced inspection.

In order to compute the average quality cost of using the sampling schemes of ISO 2859

it is necessary to know:

a. the average cost of implementing the standard under normal, reduced or tightened

inspection and

b. the estimated proportion of lots that are inspected in the long run, using normal,

reduced or tightened inspection.

Then, the average cost of using ISO 2859 per lot, taking into account the switching

procedures, is given by:

ECL   N  K (nN , cN )   T  K (nT , cT )   R  K (nR , cR ) (4)

where πΝ (πT, πR) is the proportion of lots that are inspected in the long run using normal

(tightened, reduced) inspection, K  n ,c  the expected quality cost of using the sampling plan

(n,c) and nΝ,cΝ (nT,cT, nR,cR) the sample size and acceptance number of the normal (tightened,

reduced) sampling plan.

A Markov chain has been employed in order to model the switching procedures of the

examined standard and eventually compute the values of  N ,  T ,  R and ECL. Two models

have been developed; one for Case 1 and another for Case 2, since the computational

5
procedure of the switching score is different in each case.

4.1 Case 1

In each inspection epoch i, and specifically just after the inspection of a lot, three questions

need to be answered:

(i) What is the current (updated) value of ssi?

(ii) How many lots have cumulatively been rejected for the period of time that the

standard indicates?

(iii) Will the sampling severity be changed in the following lot?

The used Markov chain consists of the following 26 states:

 N(0,ss=j), for j = 0, 3, 6, 9, 12, 15, 18, 21, 24, 27

 N(1,ss=j), for j = 0

 N(2,ss=j), for j = 0, 3

 N(3,ss=j), for j = 0, 3, 6

 N(4,ss=j), for j = 0, 3, 6, 9

 T(a), for a = 0, 1, 2, 3, 4 and

 R

At the first 20 states - which have the form N(h,ss=j) - the decision that should be taken

after lot i is inspected, regarding the severity of the next sampling, is to use normal inspection.

The integer j expresses the updated value of the switching score after the inspection of lot i.

The integer h is a counter that counts how many lots have been inspected after the last

rejection, including the lot that has been rejected. It is 0  h  4. In this sense, h = 1 in case of

a rejection of the last lot. Similarly, h = 2 in case of a rejection of the next to last lot (while the

6
last lot is accepted). If a lot is rejected and then the next two (three) are accepted, then h = 3

(h = 4). The counter h is set equal to zero

a. at the beginning of the process,

b. after reinstating normal inspection (from reduced or tightened) and, finally,

c. after accepting a lot, provided that for previous value of the counter h was 4.

At the five states T(a) the decision for the severity of the next sampling is to use tightened

inspection. Parameter a represents the number of consecutive lots that have been accepted

when tightened inspection is implemented. In this sense, whenever the procedure is at state

T(0) and the next lot is accepted, a transition to T(1) takes place. Similarly, as long as the

following lots are accepted, the state of the procedure gradually moves from T(1) to T(2), T(3)

and, eventually, T(4). If, being at state T(4), the examined (current) lot is again accepted, then

normal inspection should be implemented in the next lot, since five consecutive lots will have

been accepted, namely, a transition to N(0,ss=0) should take place. The rejection of a lot that

is inspected using tightened inspection, means that the state of the procedure returns (or

remains) to T(0).

The final state of the Markov chain (R) recommends reduced sampling.

Based on the above, let us examine step by step all possibilities from the beginning of the

process, in order to understand all possible state transitions that may occur in AS when using

ISO 2859. Assume that the process starts using normal inspection and the switching score is

set to zero: ss0 = 0. State N(0,ss=0) is the initial state of the procedure, namely, the state of the

procedure before the arrival of any lot. It indicates that the first lot to come will be inspected

using normal inspection while the current switching score is zero, since no lot has been inspe-

cted yet. In this sense, after the inspection of the first lot, there are three potential transitions:

a. If the lot is accepted and it would also be accepted for one step tighter AQL, then ss1 = 3

7
 and a transition to N(0,ss=3) state takes place.

b. If the lot is accepted, but it would be rejected for one step tighter AQL, then the process

remains in the N(0,ss=0) state (i.e., ss1 = 0).

c. Finally, if the lot is rejected, then a transition to N(1,ss=0) state occurs. N(1,ss=0) state

indicates that:

(i) The previous lot is rejected.

(ii) The current switching score is zero.

(iii) The following lot will be inspected using normal inspection.

In the same sense, the transition from N(0,ss=3)

a. to N(0,ss=6) occurs if the next lot is accepted and would also be accepted even if the AQL

was one step tighter.

b. to N(0,ss=0) takes place if the next lot is accepted, but it would not be accepted if the AQL

was one step tighter and, finally,

c. to N(1,ss=0) state occurs if the lot is rejected.

Similarly to states N(0,ss=0) and N(0,ss=3), for all states N(0,ss=6), N(0,ss=9),

N(0,ss=12), N(0,ss=15), N(0,ss=18), N(0,ss=21), N(0,ss=24), a transition to N(1,ss=0) occurs

if the next lot is rejected, a transition to N(0,ss=0) occurs if the next lot is accepted, but would

not be accepted for one step tighter AQL, while if the next lot is accepted for the current AQL

and would also be accepted for one step tighter AQL, then a transition to state N(0,ss+) occurs,

where ss+ = ss + 3.

Note that state N(0,ss=27) has not been included in the above analysis. The transitions

from this state differ from the other states in the following manner: if, being at state

N(0,ss=27), the next lot is accepted (and would also be accepted even if the AQL was one step

8
tighter) then a transition to state R occurs.

The possible transitions from N(1,ss=0) are the following:

a. If the next lot is accepted and would be accepted even if the AQL was one step tighter,

then the next state will be N(2,ss=3).

b. On the other hand, if the next lot is accepted but would not be accepted if the AQL was

one step tighter, the next state will be N(2,ss=0).

c. Finally, a transition to T(0) state will occur if the lot is rejected.

The states N(2,ss=0) and N(2,ss=3) indicate that:

(i) The next to last lot is rejected.

(ii) The current switching score is zero (or three).

(iii) The following lot will be inspected using normal inspection.

Using the same reasoning, if the next lot is accepted and would be accepted even if the

AQL was one step tighter, then the following transitions may occur, according to the current

state: N(2,ss=0) → N(3,ss=3), N(2,ss=3) → N(3,ss=6), N(3,ss=0) → N(4,ss=3), N(3,ss=3) →

N(4,ss=6) and N(3,ss=6) → N(4,ss=9). Similarly, if the next lot is accepted, but would not be

accepted for one step tighter AQL, then the following transitions may occur: N(2,ss=0) →

N(3,ss=0), N(2,ss=3) → N(3,ss=0), N(3,ss=0) → N(4,ss=0), N(3,ss=3) → N(4,ss=0) and

N(3,ss=6) → N(4,ss=0). Finally, if the lot is rejected, a transition to T(0) will occur. The same

transition to T(0) will also occur from states N(4,ss=0), N(4,ss=3), N(4,ss=6), N(4,ss=9) if the

lot is rejected. On the other hand, if the lot is not rejected, then a transition to N(0,ss=0) will

take place if the lot would be rejected for one step tighter AQL, or to N(0,ss+) if the lot would

be accepted for one step tighter AQL, where ss+ = ss + 3.

Finally, as long as the procedure remains to state R, the next lot is inspected with the

9
reduced sampling plan. As soon as a lot is rejected, the normal plan is implemented -

transition to N(0,ss=0).

The procedure described above is represented graphically in Figure 1.

[Insert Figure 1 about here]

4.2 Case 2

For case 2 (c < 2), the model is similar to the one developed so far, but in this case the

switching score increment is two instead of three. Thus, the possible states, the explanation of

which is similar to the respective one of Case 1, are the following:

 N(0,ss=j), j = 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28

 N(1,ss=0), N(2,ss=2), N(3,ss=4), N(4,ss=6)

 T(j), j = 0, 1, 2, 3, 4 and

R

Figure 2 represents graphically the states of the Markov chain and all possible transitions

for this particular case.

[Insert Figure 2 about here]

It should be noted that we disregard the case of discontinuing the inspection procedure (in

case that five lots are rejected, while using the tightened inspection plans).

5. Real case example

In this section we present a real case example of the industry in order to illustrate the use

of the new model. The company under study is an aluminium opening systems (windows and

doors) manufacturing company, which will be called hereafter by the fictitious name “Alum”,

for reasons of confidentiality. Alum uses specific types of locking mechanisms in the

windows and doors that it manufactures. Every so often Alum acquires lots of mechanisms.

10
For the period of January 2008 to June 2009, Alum has been keeping records of all incoming

lots, their size, the sampling process, as well as the nonconforming items found within every

sample. The results for this period of time, for lots of a certain type of mechanisms, acquired

from a specific supplier, are provided in Table 1.

[Insert Table 1 about here]

Based on the nonconforming items that have been found in the samples, Alum has

estimated that the average fraction of nonconforming mechanisms in the lots is about 1%.

Furthermore, as there seems to be no apparent trend around this value, Alum has assumed that

p is uniformly distributed between 0 and 2%, p~U(0,0.02).

Based on these information, the Quality Department of Alum has come to an agreement

with the locking mechanisms’ supplier that the incoming lots of mechanisms will be inspected

using the sampling plans and the switching procedures recommended by the ISO 2859

standard, for AQL = 1% and general inspection level II. Moreover, according to the

agreement, the size of lots will be constant and equal to 800, i.e. N = 800, and the deliveries

will take place on a regular, 10-day basis. Finally, the procurement arrangement specifies that

each time a lot is rejected, the supplier will be obliged to provide a discount of 500 €, besides

the obligation to replace all nonconforming mechanisms that will be revealed during the

100% rectifying inspection that follows the rejection.

According to this agreement, the sampling plans that the ISO 2859 standard recommends

for normal, tightened and reduced inspection are ( nN , cN )  (80, 2) , (nT , cT )  (80,1) and

( nR , cR )  (32,1) respectively. In addition, the recommended sampling plan when using

normal inspection but for one step tighter AQL - namely, for AQL = 0.65% - is (80,1), i.e. the

same as the tightened inspection sampling plan.

Alum has also estimated that the cost of inspecting each locking mechanism is about ci =

1.55 €, while the cost of each nonconforming mechanism that escapes the inspection process

11
cd has been estimated 60 €. The value of cd is considerably high, since every time a

nonconforming mechanism is discovered after having been mounted on the door, an

immediate disassembly of the door is necessary, in order to replace the nonconforming lock.

This has a significant effect on both the labor cost and on the cost of lost production time. The

repair cost of each locking mechanism that is discovered in the inspection process is trivial,

i.e. cr = 0, since according to the arrangement, the supplier is obliged to replace every

nonconforming lock with a conforming one at no cost for Alum.

Based on the above, the average total quality cost per lot determined by (2) and

incorporating the 500 € discount, is:

K  n, c p    n 1.55   800  n  p  60   Pa  p    800 1.55  500   1  Pa  p   . (5)

Using (5), the average costs for normal, tightened and reduced inspection are K (nN , cN ) 

0.02
0 K ( p nN , cN ) ( p )dp  552.65 € and, in the same way, K(nT,cT) = 555.99 € and K(nR,cR) =

510.97 € respectively.

For p~U(0,0.02), the average probabilities of acceptance for normal, tightened and

0.02
reduced inspection are Pa (nN  80, cN  2)   Pa ( p ) ( p )dp  0.932 and, in the same way,
0

Pa (nT  80, cT  1)  0.796 and Pa (nR  32, cR  1)  0.951 , respectively. Thus, for this

particular problem setting, the Markov chain is described by the transition probability matrix

shown in Table 2.

[Insert Table 2 about here]

The steady-state transition probabilities, denoted by  i ( i  1, 2,..., 26 ), can be obtained

either by solving the respective system of linear steady-state equations or by repeatedly

multiplying the transition probability matrix by itself until it reaches the steady state. Then,

12
the steady-state probabilities can be used to evaluate the long-run expected cost per lot. After

having computed the values of  i for i  1, 2,..., 26 , the values of  N ,  T and  R are given

20 25
by the equations:  N    i ,  T   i and  R   26 . In this way, for this particular
i 1 i  21

example, we find that  N  0.620 ,  T  0.087 and  R  0.293 , while from (4) we get that

ECLISO 2859 = 540.73 €.

Ιn this case study, the optimum decision for AS from an economic point of view would

be to accept the lot without any inspection at all, i.e. n = 0. Consequently, the resulting

average cost would be ECL* = K  n  0  

0.02
  Npc   ( p)dp  Npc
0 d d  480 €, namely

ECLISO 2859 is 12.7% larger compared to the minimum quality cost ECL*.

It is of major importance for both supplier and company (Alum, in our case) to be able to

know the proportion of lots that are inspected in the long run, under each of the three different

severities of the standard. This information can be easily obtained through our model.

More importantly, the example of Alum is indicative of the usefulness of our model for

evaluating economically the use of the ISO 2859 standard. It is not at all obvious that the use

of this standard is the best choice regarding the selection of AS plans. For the particular case

of Alum and the lots of locking mechanisms, it is evident that the best alternative, from an

economic point of view, is to accept all lots and just replace the nonconforming mechanisms

from the doors upon detection.

In the long run, this alternative is economically optimum for the supplier of locking

mechanisms as well. Based on the procurement agreement described previously, all

nonconforming mechanisms will be eventually replaced by the supplier (since all

nonconforming will be sooner or later revealed either during inspection or after having been

13
mounted on the doors). However, if Alum accepts all lots without inspection, the supplier will

never have to offer the 500 € discount. Therefore, it is obvious that the alternative of no

inspection, for the particular type of incoming raw materials, leads to a win-win situation

under an economic point of view, for both Alum and its supplier.

It should be noted that the reason for this research is not to degrade the importance of ISO

2859 standard. The existence of such standards is of significant importance for industry since

they provide a common and simple basis that companies may use in order to perform AS.

Nevertheless, it is true that these quality standards neglect the economic impact of the Quality

Control process on companies or, at the very least, do not take it explicitly into account. Thus,

when the Quality Practitioners are in position to use more sophisticated techniques to estimate

the cost elements of the Quality Control process and evaluate their economic impact, then,

economically better alternatives may arise.

6. Conclusions

We have presented a new model that can be used for the economic evaluation of the ISO

2859 standard. Using this model, it is now feasible to evaluate the economic behavior of the

standard. Besides, the derivation of the steady state probabilities are helpful in giving insights

about the proportion of lots that are inspected using each type of inspection severity. It should

be finally mentioned that the accuracy of the model depends on the estimation accuracy of all

parameters of the AS procedure. Nevertheless, when accurate estimation of all parameters is

not possible, the model does not necessarily become redundant. Even a rough estimation of

the parameters can be used in the model to provide useful (though not very accurate) insights.

References

Cassady, R., Nachlas, J.A. (2003). Evaluating and Implementing 3-Level Acceptance

Sampling Plans, Quality Engineering, 15(3):361-369.

14
Chen, J.W., Chou, W., Wu, H., Zhou, H. (2004a). Designing Acceptance Sampling Schemes

for Life Testing with Mixed Censoring, Naval Research Logistics, 51(4):597-612.

Chen, J.W., Choy, S.T.B., Li K.H. (2004b). Optimal Bayesian Sampling Acceptance Plan

with Random Censoring, European Journal of Operational Research, 155(3):683-694.

Ferrell, W.G., Chhoker, A. (2002). Design of Economically Optimal Acceptance Sampling

Plans with Inspection Error, Computers & Operations Research, 29(10):1283-1300.

González, C., Palomo, G. (2003). Bayesian Acceptance Sampling Plans Following Economic

Criteria: An Application To Paper Pulp Manufacturing, Journal of Applied Statistics,

30(3):319-333.

ISO 2859-1 (1999). Sampling Procedures for Inspection by Attributes - Part 1: Sampling

Schemes Indexed by Acceptance Quality Limit (AQL) for Lot-by-lot Inspection,

International Organization for Standardization.

Kounias, A., Nenes, G., Nikolaidis, Y., 2010, “Economic evaluation of the ISO 2859:2006

Acceptance Sampling plans: the case of a plastics company”, Proceedings of the 4 th

Panhellenic Conference for Standardization, Standards and Quality, Thessaloniki,

Greece, 19-20 November. (in Greek)

Montgomery, D.C. (2001). Introduction to Statistical Quality Control; John Wiley, NY.

Nikolaidis, Y., Nenes, G. (2009). Economic Evaluation of ISO 2859 Acceptance Sampling

Plans used with Rectifying Inspection of Rejected Lots, Quality Engineering, 21(1):10-23

Stout, Q.F., Hardwick, J. (2005). Optimal Screening Designs With Flexible Cost and

Constraint Structures, Journal of Statistical Planning and Inference, 132(1-2):149-162.

Tagaras, G., Lee, H.L. (1987). Optimal Bayesian Single-Sampling Attribute Plans with

Modified Beta Prior Distribution, Naval Research Logistics, 34(6):789-801.

Wall, M.S., Elshennawy, A.K. (1989). Economically-based Acceptance Sampling

Plans, Computers & Industrial Engineering, 17(1-4):340-346.

15
Wetherrill, G.B., Chiu, W.K. (1975). A Review of Acceptance Sampling Schemes with

Emphasis on the Economic Aspect, International Statistical Review, 43(2):91-210.

16
Tables and figures

Table 1: Records of incoming lots of locking mechanisms, January 2008 to June 2009
No Date Lot size N Sample n Nonconforming d p (%)
1 January 12, 2008 1000 100 1 1.00
2 February 6, 2008 1800 150 2 1.33
3 February 13, 2008 500 100 2 2.00
4 March 19, 2008 2400 200 0 0.00
5 March 27, 2008 600 100 2 2.00
6 April 13, 2008 1400 100 1 1.00
7 April 16, 2008 400 50 0 0.00
8 May 16, 2008 2000 200 3 1.50
9 May 31, 2008 1000 100 0 0.00
10 June 29, 2008 2000 200 1 0.50
11 July 16, 2008 1200 100 2 2.00
12 August 9, 2008 1600 150 1 0.67
13 September 10, 2008 2200 200 1 0.50
14 September 18, 2008 600 100 2 2.00
15 October 1, 2008 900 100 0 0.00
16 October 8, 2008 500 100 0 0.00
17 October 24, 2008 1200 100 0 0.00
18 November 12, 2008 1300 100 2 2.00
19 December 11, 2008 2000 200 3 1.50
20 January 3, 2009 1500 150 2 1.33
21 January 15, 2009 900 100 1 1.00
22 January 25, 2009 700 100 2 2.00
23 January 29, 2009 500 100 1 1.00
24 February 20, 2009 1500 150 3 2.00
25 March 25, 2009 2400 250 1 0.40
26 April 24, 2009 2100 200 2 1.00
27 April 29, 2009 500 100 1 1.00
28 May 22, 2009 1700 150 1 0.67
29 June 20, 2009 2000 200 1 0.50

17
 Table 2: Transition probability matrix of the Markov chain for the examined case

N(0,ss=12)

N(0,ss=15)

N(0,ss=18)

N(0,ss=21)

N(0,ss=24)

N(0,ss=27)
N(0,ss=0)

N(0,ss=3)

N(0,ss=6)

N(0,ss=9)

N(1,ss=0)

N(2,ss=0)

N(2,ss=3)

N(3,ss=0)

N(3,ss=3)

N(3,ss=6)

N(4,ss=0)

N(4,ss=3)

N(4,ss=6)

N(4,ss=9)

T(1)

T(2)

T(3)
T(0)

T(4)

R
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
N(0,ss=0) 1 .136 .796 0 0 0 0 0 0 0 0 .068 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
N(0,ss=3) 2 .136 0 .796 0 0 0 0 0 0 0 .068 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
N(0,ss=6) 3 .136 0 0 .796 0 0 0 0 0 0 .068 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
N(0,ss=9) 4 .136 0 0 0 .796 0 0 0 0 0 .068 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
N(0,ss=12) 5 .136 0 0 0 0 .796 0 0 0 0 .068 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
N(0,ss=15) 6 .136 0 0 0 0 0 .796 0 0 0 .068 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
N(0,ss=18) 7 .136 0 0 0 0 0 0 .796 0 0 .068 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
N(0,ss=21) 8 .136 0 0 0 0 0 0 0 .796 0 .068 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
N(0,ss=24) 9 .136 0 0 0 0 0 0 0 0 .796 .068 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
N(0,ss=27) 10 .136 0 0 0 0 0 0 0 0 0 .068 0 0 0 0 0 0 0 0 0 0 0 0 0 0 .796
N(1,ss=0) 11 0 0 0 0 0 0 0 0 0 0 0 .136 .796 0 0 0 0 0 0 0 .068 0 0 0 0 0
N(2,ss=0) 12 0 0 0 0 0 0 0 0 0 0 0 0 0 .136 .796 0 0 0 0 0 .068 0 0 0 0 0
N(2,ss=3) 13 0 0 0 0 0 0 0 0 0 0 0 0 0 .136 0 .796 0 0 0 0 .068 0 0 0 0 0
N(3,ss=0) 14 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 .136 .796 0 0 .068 0 0 0 0 0
N(3,ss=3) 15 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 .136 0 .796 0 .068 0 0 0 0 0
N(3,ss=6) 16 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 .136 0 0 .796 .068 0 0 0 0 0
N(4,ss=0) 17 .136 .796 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 .068 0 0 0 0 0
N(4,ss=3) 18 .136 0 .796 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 .068 0 0 0 0 0
N(4,ss=6) 19 .136 0 0 .796 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 .068 0 0 0 0 0
N(4,ss=9) 20 .136 0 0 0 .796 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 .068 0 0 0 0 0
T(0) 21 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 .204 .796 0 0 0 0
T(1) 22 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 .204 0 .796 0 0 0
T(2) 23 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 .204 0 0 .796 0 0
T(3) 24 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 .204 0 0 0 .796 0
T(4) 25 .796 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 .204 0 0 0 0 0
R 26 .049 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 .951

18
 Figure 1: Transitions for all possible states for Case 1 (c ≥ 2)

a
a A
a
a
N(0,ss=0) r R a
a a
a a
A A a a a a A
r a

N(0,ss=3) A N(0,ss=6) A N(0,ss=9) A N(0,ss=12) A N(0,ss=15) A N(0,ss=18) A N(0,ss=21) A N(0,ss=24) A N(0,ss=27)

r
r
r

r r
r r r r

A
N(1,ss=0) A A A
A
a
N(2,ss=0) N(2,ss=3) A

a A
a
N(3,ss=0) N(3,ss=3) N(3,ss=6) A
r
r
r a
a a A
T(4) r r r
r
A N(4,ss=0) N(4,ss=3) N(4,ss=6) N(4,ss=9)

T(3) A T(2) A T(1) A T(0) A

r r
r
r r
r r
r

A: the lot is accepted and would have been accepted even if the AQL had been one step tighter (for normal inspection)
the lot is accepted (for tightened or reduced inspection)
a: the lot is accepted but would have been rejected if the AQL had been one step tighter (for normal inspection)
r: the lot is rejected

Figure 2: Transitions for all possible states for Case 2 (c < 2)

a

N(0,ss=0) r R

a a
a

r N(0,ss=2) a N(0,ss=4) a N(0,ss=6) a N(0,ss=8) a N(0,ss=10) a N(0,ss=12) a N(0,ss=14) a N(0,ss=16) a N(0,ss=18) a N(0,ss=20) a N(0,ss=22) a N(0,ss=24) a N(0,ss=26) a N(0,ss=28)

r a
r r r
r r r r

r r r r
r r

N(1,ss=0) a N(2,ss=2) a N(3,ss=4) a N(4,ss=6)

r
T(4)
r
r r
a a: the lot is accepted
r r: the lot is rejected
T(3) a T(2) a T(1) a T(0)

r r
r
r

19