Some pros and cons of MIL-STD-414

MIL-STD-414 has been in existence in its original form since 1957. Obviously, there are some things that are right about 414. But there are some changes to the standard that are long overdue. This article presents some pros and cons of 414 and a recommendation that it be replaced by a MIL-STD-414A. Two concerns regarding ANSI/ASQC Z1.9 are discussed.     

Qmp/USP: A modern alternative to MIL-STD-105D

The Quality Measurement Plan (QMP) and the Universal Sampling Plan (USP) are the data analysis and sampling plans for the AT&T Technologies quality audit. This article describes QMP/USP, an acceptance sampling plan based on QMP and USP principles. QMPIUSP is a complete acceptance sampling system. It combines the elements of classical rectification inspection plans with those of MIL-STD-IOSD. There is no switching between plans, no tables of numbers to look through, and no discontinue state. QMP/USP is a computerized, self-contained system that features:

• Acceptance decisions based on the QMP Bayes empirical Bayes analysis of current and past sampling result
• Sample size selection based on USP, i.e., lot size, AQL, a cost ratio, the QMP analysis, and a budget constraint
• Guaranteed AOQ
• A complete statistical analysis of the quality process.

In this article, we describe the operation of QMP/USP and compare its performance with that of MIL-STD-IOSD. The comparison is made under many different quality environments with many metrics. Our results show that QMP/USP and MIL/STD/IOSD perform similarly for the environments where quality could be described as “in control”; and that QMPlUSP is superior in the environments where quality is “out of control”.     

The effect of nonnormality on variables sampling plans

Acceptance sampling plans based on variables have been in use for many years. Recently, there has been a renewal of interest in these plans, because of the relative efficiencies that they offer with respect to attributes sampling regarding sample size. Furthermore, in situations where acceptable quality levels are very small, and a high level of protection is desired, variables sampling is often much more efficient than attributes sampling. An important disadvantage of variables sampling is that the distribution of the parameter being inspected must be known. Most standard variables sampling plans assume that the distribution of this parameter is normal. This article examines the effect of the normality assumption in variables sampling. Methods to detect departures from normality are reviewed.     

Stepwise inspection in Bayesian multiattribute acceptance sampling

Bayesian models for multiattribute acceptance sampling have been developed under the assumption that sampling inspection is carried to completion. A Bayesian multiattribute model for stepwise sampling inspection is proposed, whereby sampling inspection is terminated as soon as the disposition of the inspection lot is determined. An iterative solution procedure is developed for obtaining optimal or near-optimal multiattribute acceptance sampling plans under stepwise sampling inspection. The effect of stepwise sampling inspection on the characteristics of an optimal sampling plan is investigated. It is shown that stepwise sampling inspection achieves a sampling plan with lower total expected cost than complete sampling inspection. In addition, it is shown that the sequence of attributes in a stepwise sampling inspection substantially affects the sampling plan and resultant expected cost. The proposed methodology is used to evaluate various heuristics which may be used to determine the sequence of attributes in a stepwise inspection procedure.     

Economic design of sampling plans with multi-decision alternatives

For situations where there are several markets with different profit-cost structures, economic attributes sampling plans are developed for determining the market to ship the lot to. Two sampling plans are considered; the fixed-size sampling plan with several levels of acceptance numbers, and the inverse sampling plan with several levels of sample sizes. Linear profit models are constructed which involve four profit/cost components; profit from a conforming item, inspection cost, replacement cost, and cost from an accepted nonconforming item. For fixed-size and inverse sampling plans, methods of finding optimal sampling plans are presented and examples with beta-binomial prior distributions are given.     

Some experiences using MIL-STD-105 in the united states air force

The historical and, to some extent, the current role of attribute inspection sampling in the United States Air Force is described. Some insight into consideration culminating in revisions made to the original MIL-STD-105 sampling plans is provided. These revisions were motivated from a practitioner's perspective rather than that of a statistical expert. Some significant administrative applications within the Air Force are described, and a practical method, based on the total amount of inspection, is given for distinguishing between sampling plans that have a fixed risk level (i.e., common value on the OC curves).     

Present practices and future plans for MIL-STD-781

MIL-STD-781 specifies reliability acceptance test procedures based on both fixed-length tests and probability ratio sequential tests. The assumption underlying MIL-STD-781 is a constant mean time between failures (MTBF) and typical practice applies MIL-STD-781 to electrical, electronic, and mechanical equipment. This article discusses some of the difficulties that have prompted the C and D revisions of MIL-STD-781. In addition, it discusses the relationship of MIL-STD-781 with MIL-STD-1635(EC) which deals with reliability growth testing. The article concludes with a discussion of needed research in reliability growth testing, in support of MIL-STD-781, and in the area of stress and its impact of MTBF.     

An analysis of the long range operating characteristics of the MIL-STD-105D sampling scheme and some suggested modifications

An integral part of the MIL–STD–105D scheme for sampling inspection by attributes is the transfer from normal inspection to reduced or tightened inspection when the historical record of inspected lots suggests unusually good or bad quality. The switching rules in MIL–STD–105D have been criticized, especially by Japanese manufacturers, as being too severe when what is defined as acceptable quality material is submitted. This paper examines the long range fraction of lots rejected for several MIL–STD–105D sampling plans by using the MIL–STD–105D switching rules, using a modification suggested by the Japanese Standards Association, and by using a second modification developed by the authors. The Japanese Standards Association switching rules are more complex than those in MIL–STD–105D. It is demonstrated that they lead to improved long–range properties for Normal–Tightened–Reduced schemes, but to poorer properties for Normal–Tightened schemes. A simplified set of switching rules is suggested, wherein the “limit numbers” in MIL–STD–105D are eliminated. In comparison to MIL–STD–105D, the simpler rules lead to a lower probability of rejection for good lots and a comparable probability of rejection for bad lots.     

Some recent findings in bayesian attribute single and double sampling

This article deals with several items, including theoretical and applied results. Specific topics include (1) a discrete, economically based, attributes acceptance sampling model and its adaptations, (2) relevant costs, (3) relevant prior distributions, (4) comparison of single- and double-sampling results, and (5) reasons for marginal implementation success following excellent implementation efforts. The basic model used is one developed by Guthrie and Johns; adaptations include provisions for fixed costs as well as modifications to permit double sampling. Optimization is exact, rather than approximate. Costs incorporated into the model are for sampling inspection, lot acceptance, and lot rejection. For each of these three categories a fixed cost is included as well as two variable costs, one for each item and the other for each defective item. Discrete prior distributions for the number of defectives in a lot are used exclusively. These include the mixed binomial and Polya distributions. Single- and double-sampling results are compared. Double sampling regularly performs at only slightly lower cost per lot than single sampling. Also, some cost and prior distribution sensitivity results are presented. Comments are provided regarding actual implementation experiences in industry. Practical deficiencies with the Bayesian approach are described, and a recommendation for future research is offered.     

计数序贯抽样检验方案的OC函数和ASN

根据序贯概率比检验的理论,给出计数序贯抽样检验方案的设计原理,得到国家标准GB／T8051—2008中计数序贯抽样检验方案的OC函数和ASN的计算方法。     

比较的视角：中美研究生招考制度研究

本文以比较的视角,在招考管理权,招生人数控制、入学奖学金制度、考试内容以及招生录取标准等几个方面探讨了中美研究生招考制度的异同,分析目前我国研究生招生考试中出现的问题,最后得出解决问题的启示。     

The optimum design of multivariate acceptance sampling plans

A model is developed which may be used to determine the expected total cost of quality control per inspection lot under acceptance sampling by variables where several characteristics are to be simultaneously controlled. Optimization of the model is accomplished through the application of a conventional search procedure. The sensitivity of the model and the optimum solution to the shape of the underlying probability distributions is discussed and associated analyses are presented through an example.     

浅析建筑物鉴定中的计数抽样方案

在介绍计数型抽样检验原理的基础上,从建筑物鉴定检测的特点出发提出了采用相应抽样检验国家标准的建议。指出GB/T50344-2004《建筑结构检测技术标准》中关于计数抽样检测方案的规定失之过松,有可能给建筑物的检测或鉴定留下安全隐患,不宜采用。     

成败型产品验收试验方案研究

在可靠性定型试验的基础上,利用多层贝叶斯方法确定批产品可靠性指标的先验分布,从而制定出成败型产品可靠性验收试验的一种贝叶斯方案.给出利用定型试验信息确定产品可靠性指标先验分布的方法.这种验收方法充分利用了产品定型试验中的先验信息,在确保有较好验收效果的前提下,与传统的验收试验相比,可以大大减少试验量,从而得到可观的经济效益.     

The applicability of acceptance sampling with respect to process stability and control

This article examines the applicability of acceptance sampling and the effectiveness of Deming's kp rule in relation to the degree of process stability achieved through statistical process control techniques. A discrete-event simulation model is used to characterize the correlation between the number of defective units in a randomly drawn sample versus in the remainder of a lot, in response to a number of system and control chart parameters. The model reveals that such correlation is typically present when special causes of variation affect the production process from time to time, even though the process is tightly monitored through statistical process control. Comparison of these results to an analogous mixed binomial scenario reveals that the mixed binomial model overstates the correlation in question if the state of the process is not necessarily constant during lot production. A generalization of the kp analysis is presented that incorporates the possibility of dependence between a sample and the unsampled portion of the lot. This analysis demonstrates that acceptance sampling is generally ineffective for lots generated by a process subject to statistical process control, despite the fact that the number of defectives in the sample and in the remainder of the lot are not strictly independent. © 1994 John Wiley & Sons, Inc.     

A cost analysis of sampling inspection under Military Standard 105D

Military Standard 105D has been almost universally adopted by government and private consumers for the lot-by-lot sampling inspection of product which may be inspected on a dichotomoun basis The plan specifies, for each lot size, a random sample size and set of acceptance numbers (maximum allowable number of defectives in each sample). The acceptance numbers are based upon the binomial distribution and depend upon the quality required by the purchaser. Where several consecutive lots are submitted, a shift to less severe (“reduced”) inspection or more severe (“tightened”) inspection is specified when the ongoing quality is very high or low. Further experience permits a return to normal sampling from either of these states This paper examines the long range costs of such a sampling scheme. The three inspection types are considered as three distinct Markov chains, with periodic transitions from chain to chain. The expected sample size and the expected proportion of rejected product are determined as a function of the two parameters under control of the manufacturer, lot size and product quality. Some numerical examples are given which illustrate how to compute the overall cost of sampling inspection. Suggestions are made concerning the choice of parameters to minimize this cost.     

Minimum cost sampling plans using bayesian methods

This article considers a general method for acceptance/rejection decisions in lot-by-lot sampling situations. Given arbitrary cost functions for sampling, accepting, and rejecting (where the cost can depend on the quality of the item) and a prior distribution on supplier quality, formulas are derived that lead to the minimal cost single-staged inspection plan. For the Bernoulli case, where each item is classified as acceptable or defective, the formulas simplify immensely. A computer code for solving the Bernoulli case is given.     

Optimal selection of the t best of a sequence with sampling cost

This article deals with the problem of selecting the t best of n independent and identically distributed random variables which are observed sequentially with sampling cost c per unit. Assume that a decision for acceptance or rejection must be made after each sampling and that the reward for each observation with value x is given by px - c, where p is 1 if the observation is accepted, or 0 otherwise. The optimal decision procedure (strategy) for maximizing the total expected reward is obtained. The critical numbers which are necessary to carry out the optimal decision procedure is presented by two recursive equations. The limit values of the critical numbers and the expected sample size are also studied.     

On the robustness of the modified beta distribution for acceptance sampling in statistical quality control

The purpose of this article is to examine the robustness of the modified Beta distribution as a probability distribution of lot fraction defectives in Bayesian acceptance sampling for statistical quality control. In complex manufacturing systems, a production process may consist of multiple production stages in a serial or nonserial fashion. Hence, inputs to a production station can be a result of subassembly of several inputs, or outputs of some inspection stations for some prior work stages. We investigate the effectiveness of the modified Beta distribution as an approximation to the lot fraction defectives probability of inputs at intermediate work stations. The robustness of the modified Beta distribution simplifies both the determination of the optimal sampling plan for acceptance sampling, and the calibration of distributions resulted from subassembly or inspection operations in complex manufacturing systems.     

A backward recursive technique for optimal sequential sampling plans

This paper presents a procedure akin to dynamic programming for designing optimal acceptance sampling plans for item-by-item inspection. Using a Bayesian procedure, a prior distribution is specified, and a suitable cost model is employed depicting the cost of sampling, accepting or rejecting the lot. An algorithm is supplied which is digital computer oriented.