Date of Award

August 2015

Degree Type


Degree Name

Master of Science



First Advisor

Liang Zhang





Yang Sun

The University of Wisconsin-Milwaukee, 2015

Under the Supervision of Professor Liang Zhang

Production lines with unreliable machines and finite buffers are characterized by both steady-state performance and transient behavior. The steady-state performance has been analyzed extensively for over 50 years. Transient behavior, however, is rarely studied and remains less explored. In practice, a lot of the real production systems are running partially or entirely in transient periods. Therefore, transient analysis is of significant practical importance.

Most of the past research on production systems focuses on discrete materials flow which utilities Markov chain analysis. This dissertation is devoted to investigate the effects of system parameters on performance measures for transient serial production line with other machine reliability models. The reliability models investigated in this dissertation are exponential and no-exponential (Weibull, Gamma, Log-normal).

In a real production line system, machine reliability models are much more difficult to identify. Strictly speaking, it requires the identifications of the histograms of up- and downtime, which requires a very large number of measurements during a long period of time. The result may be that the machines' real reliability model on the factory floor are, practically, never known. Therefore, it is of significant practical importance to investigate the general effects of system parameters on performance measures for transient serial production line with different reliability models. The system parameters include machine efficiency e, ratio of N and Tdown (K), machines' average downtime Tdown, and coefficient of variation CV. The performance measures include settling time of production rate (t_sPR), settling time of work-in-process (t_sWIP), total production (TP), production loss (PL). The relationship between the performance measures and system parameters reveals the fundamental principles that characterize the behavior of such systems and can be used as a guideline for product lines' management and improvement.

Most previous research studies are limited to two or three machine system due to the technical complexity. Furthermore, presently there are no analytical tools to address the problems with multiple machines and buffers during transient periods. This dissertation addresses this problem by using simulations with C++ programming to evaluate the multiple machines (up to 10) and buffers and demonstrate the transient performance at different conditions. The simulation method does not only provide quantified transient performance results for a given production line, but also provides a valuable tool to investigate the system parameter effects and how to manage and improve the existing production line.