Discrete event simulation based analysis to improve performance of a manufacturing system

• Main Author: Lee, Bernice Xing Ye
• Other Authors: Sivakumar Appa Iyer,Siva
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Engineering::Manufacturing::Flexible manufacturing systems; Engineering::Industrial engineering::Supply chain; Engineering::Mechanical engineering
• Online Access: https://hdl.handle.net/10356/141035
• Institution: Nanyang Technological University

This project covers an analysis of a backend assembly test manufacturing system by using discrete event-based simulation model to improve its productivity. The system consists of 9 operation processes with multiple parallel operating machines to be conducted for analysis. From the actual data collected from the former National Semiconductor operation by a research team, the bottleneck was identified to be in the Test station achieving a throughput, rb of 6 lots per hour and a raw process time of 16 hours per lot. The warm-up time period of 14 days was required in the model settings and multiple replications were conducted to obtain a reliable analysis of the actual production. A discrete event simulation model built using the Simul8 software was verified and validated to be adequately reliable to run for analyses as the % difference in throughput and cycle time with the actual production was minimal. Using the model, some parameters that affect the performance of the manufacturing system were studied. Although production flow was smooth without presence of blockage, hourly input release system produces a poor performance compared to the daily input release due to machine starvation. Stochastic input release system with uncertain demand performs poorer than the deterministic system and the higher the input release variability, the poorer the performance as WIP accumulates in the constraint stations. Moreover, the higher the average input volume release to the system, the better the system performs and it performs best at the bottleneck rate. It starts to deteriorate when the input volume further increases causing system deterioration. In addition, further analysis on the system with low variability of machine breakdown with low breakdown frequency demonstrates a better performance than a high breakdown variability system. Additional machines might not be ideal in improving the system as the result demonstrates that the gap between the actual production and the critical point remains constant. On the other hand, reduction in the repair time of the bottleneck station improves the performance with an increase in the constraint machine utilisation. Thus, this modification can be possibly implemented in the real-life scenario of the actual production.