Sampling Plans for Batch and Sequential Inspection

Traditional supervision consists of keeping close control of operations and progress, the focus of attention being the product or process outputs. A direct implication of this approach is to guarantee product quality through inspection and screening. The chapter discusses sampling techniques and measures of inspection effectiveness. Performance characteristics of sampling plans are discussed and guidelines for choosing economic sampling plans are presented. The basic theory of single-stage acceptance sampling plans for attributes is first presented including the concepts of Acceptable Quality Level and Limiting Quality Level. Formulas for determining sample size, acceptance levels, and operating characteristic functions are provided. Moving on from single-stage sampling, the chapter covers double sampling and sequential sampling using Wald's sequential probability ratio test. One section deals with acceptance sampling for variable data. Other topics covered include computations of Average Sample Numbers and Average Total Inspection for rectifying inspection plans. Modern Skip-Lot sampling procedures are introduced and compared to the standard application of sampling plans where every lot is inspected. The Deming "all or nothing" inspection criterion is presented and the connection between sampling inspection and statistical process control is made. Special sections are dedicated to sequential methods of software applications such as one- and two-arm bandit models used in A/B testing and software reliability models used in determining release readiness of software versions. Throughout the chapter we show Python code which is used to perform various calculations and generate appropriate tables and graphs.