Models for continuous improvement of productivity and quality

Ishpal Singh Rekhi, Purdue University

Abstract

During the last decade there has been a tremendous revolution in quality management practices. These changes have come as a result of the necessity to change in a globally competitive environment rather as a result of any major breakthrough in science or technology. This dissertation focuses on problems that will assist the firm in increasing its productivity and improving the quality of the product. We first consider the problem of optimal allocation of work in an assembly system, we examine a PUSH system. The problem corresponds to moving work between feeder stations and the assembly station to achieve the optimal workload that will maximize expected throughput. For this configuration, earlier studies showed that optimal expected throughput decreased as the number of feeder stations increased for exponential and log-normal processing times. However, the pattern was not similar for uniform processing times, where, the optimal throughput initially decreased and then increased as the number of components increased. We provide insights as to the causes of this seemingly anomalous behavior described for the uniform processing time distribution. We also investigate the impact of unbalancing on the variability of interdeparture time (reciprocal of throughput) and observe that the variability of interdeparture time decreases due to unbalancing. Further, we investigate a more realistic case that considers the assembly time as a function of the number of components. Next, we consider a manufacturing environment where process improvement activities require use of the productive capacity of the firm in addition to other investments. Thus, the firm must allocate its productive capacity between production activities and improvement activities. The output of production activities is used to meet customer demand. Process improvement activities improve the quality of the output, which in turn leads to lower quality related costs (both internal and external) and possibly lower per-unit production cost. A continuous time, finite horizon, profit maximization, resource allocation model is developed to find an optimal time path for process improvement activities and production activities. Lastly, we consider acceptance sampling plans using Bayesian techniques and subject to measurement error. Prior work has shown that Taguchi's quadratic loss function is superior to the step-loss function and Bayesian sampling plans are more appropriate. Impact of measurement error is significant and the assumption of no measurement error should be reconsidered. In this section, we develop two Bayesian lot-by-lot variable acceptance sampling plans, subject to measurement error, using the Taguchi loss function for measuring the loss due to imperfect quality. We conduct a sensitivity analysis of the models with respect to the cost of rejection or rework, the quadratic cost of quality, the correlation between the observed deviation and the actual deviation of the performance variable, and the measurement error, and obtain some interesting counterintuitive results. These models incorporate the concept of continuous improvement of productivity and quality through learning.

Degree

Ph.D.

Advisors

Moskowitz, Purdue University.

Subject Area

Management|Industrial engineering

COinS