Track: Quality Control and Quality Management

Abstract

In recent years, technology has improved dramatically, and consumers’ demands for quality have become higher. Therefore, many scholars have investigated in designing advanced process capability indices for measuring the process performance with extremely low defect rate.  Process yield is a commonly used tool in manufacturing industry to evaluate process performance. However, process yield only consider if the product falls within the specification limits and assume the products which meet the specification limits have the same quality. To remedy for this situation, a more advanced and customer-orientated index, called quality yield index (Yq), was proposed. Quality yield index takes quality loss into consideration by penalizing yield when the product characteristic deviates from the target value. Unfortunately, the sampling distribution of the estimated Yq  is too complicated, which makes it difficult to establish the exact confidence interval. Therefore, several approaches for constructing confidence intervals of the index Yq have been presented including approximate distribution approach (AD) and bootstrap approach. Approximate distribution method is constructed by subtracting the upper bound of expected relative process loss from the lower bound of process yield and can also be view as a conservative estimation method. Bootstrap method is a useful technique for estimating sampling distribution no matter what the exact population distribution is. In this paper, we apply the concept of generalized pivotal quantities to construct the generalized confidence interval (GCI) for Yq. In order to examine the performance of GCI method, a series of simulations are conducted and compared with the existing methods in terms of coverage rate and the average value of lower confidence bound (LCB). The simulation results indicate that GCI method always performs the best for calculating LCB of Yq. As a result, the GCI method is recommended in practical situations for assessing the process performance based on the quality yield index.