Access Type

Open Access Dissertation

Date of Award

January 2012

Degree Type

Dissertation

Degree Name

Ph.D.

Department

Industrial and Manufacturing Engineering

First Advisor

Ratna Babu Chinnam

Second Advisor

Ekrem Alper Murat

Abstract

Companies are increasing their manufacturing excellence in order to stay competitive in the globalizing market. Plants are becoming more complex year by year due to increasing product classes, hardware complexity, etc... The design and operation of manufacturing systems is of greater importance today than it was in the past. Many studies have been carried out on the design and operation of manufacturing systems by academicians and practitioners over the years, however, there is still no agreement on how to best predict and improve the factory performance (Gershwin, 2000). The studies are based on either analytical approaches or simulation-based approaches. Success stories from some companies, for instance General Motors, which applied these techniques in combination, motivate our study.

In the dissertation, our main focus area is the automotive industry. Maintenance, being the most critical component of the automotive industry, has a direct impact on the improvement of the overall production performance. Therefore, we introduce an anticipative plant level maintenance decision support system (APMDSS), which gives guidance on the corrective and the preventive maintenance priorities, and the times for doing preventive maintenance tasks based on the bottleneck ranks with an objective of improving the throughput of a plant which consists deteriorating machinery. Unlike the previous bottleneck management approaches, APMDSS anticipates the system dynamics (i.e., bottlenecks, hourly buffer levels, and machine health) of the upcoming shift by using initial state information such as machine ages, operational status of machines, buffer levels, and model mix. In order to make a more realistic and detailed analysis, we model the factory dynamics using a simulation model.

We also propose two analytic models for throughput evaluation. First one is an exact formula for a deteriorating two-machine system. In the model, the machines degrade with usage and the reliability behavior of each machine changes depending on the machine's health condition. The model considers both perfect and imperfect repairs simultaneously.

The second one is hybrid aggregation-decomposition algorithm that approximates the throughput of longer production lines. The algorithm selectively aggregates the parts of the line based on the location of the bottlenecks. In this model, we engage the existing aggregation and decomposition methods. The basic idea of making a hybrid of these two throughput evaluation approaches is to benefit from the speed of the aggregation method and the accuracy of the decomposition method.

We obtained promising results in the experiments that we tested our models using real data from a major automotive company. We also used synthetic data in the experiments to investigate different scenarios.