Performance Analysis Of Multi-loop Closed Semiconductor Manufacturing

Semiconductor assembly and test is the back end of the semiconductor manufacturing, which has a very important role in semiconductor manufacturing. Especially in china, it took more than 50 percent of the whole semiconductor manufacturing market. Along with the progressing of the market competition, semiconductor system performance evaluation and related system planning and design have been more and more critical.In this thesis, we clarified the special characters of the complex semiconductor assembly and test production system in detail and pointed out the important indicators for system performance evaluation such as the throughput, cycle time and WIP level. To understand the behavior of the semiconductor system, we designed a graphic method named"O_L Graph"to analyze the semiconductor performance evaluation process and the improvement process.For the performance analysis of the two-loop closed semiconductor system formed by the limited assistant resources circling on the line, we proposed the exact method for solving smaller ideal lines and the approximate method for analyzing of the larger and practical systems. In the exact method, we constructed a two-loop closed production system by means of block-structured Markov Chains, and proposed an effective RG-factorization solution to evaluate performance measures of the smaller system; while in the approximate method, an overlapping decomposition process based on the performance analysis algorithms for single loop closed system was designed to solve the problem of the larger systems.By means of the simulation software Witness, a lot of numerical simulation experiments were implemented. The performance measure results based on simulation and the results based on the proposed approximate method for different production lines were compared and showed that the approximate algorithm has an acceptable accuracy. Moreover, we analyzed the performance when changing resource numbers, machines parameters and system structures, and found some important behavior rules for this type of system including: (1) in a two-loop unreliable closed system with no buffer limits, the system throughput will increase with the increasing of the resource numbers;(2) in a two-loop unreliable closed system with buffer limits, the system throughput increases first and then decreases with respect to the increasing of the resource numbers; (3) in both types of two-loop closed system, the WIP level of the production line increases with respect to the increasing of resource numbers; (4) in both types of two-loop closed system, machine of the resource change station has more important role to the system throughput than other stations, for its improvement can bring more increase to the throughput.A case study in a real semiconductor assembly and test factory was described including the process of data pretreatment and the comparision between the results achieved from the approximate method and results from the practice. Through this case study, we found that the performance measure method proposed in this thesis can be applied to some practical systems effectively. Various algorithms application ways such as resource optimization, performance forecast, and objectives setting, were discussed to indicat that the algorithms we designed have a great application foreground.