Manufacturing Systems Modeling and Analysis
Open Access
- Author:
- Julaiti, Juxihong
- Graduate Program:
- Industrial Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- May 06, 2021
- Committee Members:
- Soundar Rajan Tirupatikumara, Dissertation Advisor/Co-Advisor
Guodong Pang, Committee Chair/Co-Chair
Eunhye Song, Committee Member
Vasant Gajanan Honavar, Outside Member
Bicheng Chen, Special Member
Steven James Landry, Program Head/Chair
Soundar Rajan Tirupatikumara, Committee Chair/Co-Chair
Guodong Pang, Dissertation Advisor/Co-Advisor - Keywords:
- stochastic scheduling
job shop scheduling
optimal control
reinforcement learning - Abstract:
- In the United States, an increasing amount of industries are becoming high-mix and low-volume (HMLV) facilities to provide various products and to stay competitive. The heterogeneity of products introduces frequent reconfiguration to the production line and therefore increases the chance of unplanned downtime. Because of the significant cost of unplanned downtime, any effort to reduce it is a welcome endeavor to U.S. manufacturers. This thesis contains three chapters, the first two address HMLV reconfiguration problem. The third chapter deals with adaptive controls in unreliable single server queues, which can represent a typical HMLV environment. In the first chapter, a parallel machine scheduling problem with stochastic machine breakdowns is studied to minimize the weighted tardiness. We propose a reinforcement learning-based framework with a novel sampling method. The results indicate that the new sampling approach expedites the learning process and the resulting policy significantly outperforms static dispatching rules. In a HMLV facility, similar jobs are often scheduled together to decrease additional setup times. However, in dry machining processes, utilizing a tool for a prolonged period of time overheats the tool and increases chances of tool damage and scrapped parts. Therefore, the optimal schedule should avoid tool overheating. In the second chapter, we propose a mixed integer programming model to minimize the makespan in a job-shop scheduling environment with overheating constraints. Numerical studies are conducted to validate the model. In queuing systems, the unplanned machine downtime is commonly modeled using queues with interruptions and it has received considerable attention since the late 1950s. In the third chapter, we study adaptive service rate control problems for single server queues with server breakdowns. We focus on a particular dependent structure where the breakdown rate of the server is a linear function of the service rate. Since the relation between these two rates might be unknown in practice, we develop online algorithms to obtain the optimal policy. Numerical studies are conducted to analyze the optimal policy and validate proposed algorithms.