Optimization Of Semiconductor Manufacturing System Based On Agent Features

As the key to the third scientific and technological revolution,the semiconductor industry has developed rapidly with the continuous progress of science and technology.As a representative of NP-hard reentrant production systems,it has attracted much attention from academics and industry for many years.Meanwhile,as the United States has suppressed and sanctioned the development of the Chinese semiconductor industry since 2018,efficient management and resource utilization efficiency have become important factors for domestic companies to break monopoly and win such fierce competition.Therefore,a comprehensive study of the semiconductor production system not only has important theoretical and economic values,but also has certain social significance in the current international context.However,most of the current researches on this production system focus on scheduling strategies and scheduling algorithms.The modeling is relatively complicated and the computational requirements are relatively high.In addition,the current research has not obtained extensive and consistent conclusions.Based on the above situation,this thesis introduces the characteristics,composition,and status scheduling of this production system,models a typical semiconductor production system,and analyzes the performance of equipment utilization rate,production cycle,work in process inventory,and production capacity under various scheduling and feeding strategies based on simulation results,and further explore the main factors affecting the overall production efficiency of the semiconductor manufacturing system based on a variety of combined scheduling and feeding strategies.The following three conclusions were obtained:(1)As for scheduling strategies,the shortest remaining process time strategy(SRPT)is optimal in terms of production capacity,work in process inventory,and equipment utilization,and it has a good adaptability to speeding up the feeding rate,which will be normally regarded as the optimal scheduling strategy.The longest queue buffer priority strategy(LB)has a significantly lower average manufacturing cycle of workpieces than the other three strategies,and it does not change significantly with the speed of feeding,so it is more suitable for stable large-scale production scenarios that do not require high output extremes.The shortest step of next-visit priority strategy(STNV)performs far worse than the other three methods in all indicators.(2)As for feeding strategy,the optimization effect of the fixed WIP feeding strategy is the most obvious.It can maintain high production capacity and reduce the number of work-inprocess.Anti-hunger strategy can also improve work-in-process conditions indirectly through queuing control of a certain part of the production line,static feeding strategy is the worst because the feeding speed is not adjusted due to the feedback of the real-time situation of the production line.(3)The comparison of the combined optimization effects of scheduling and feeding strategies further reveals that the key factor that reduces production efficiency is the overall work-in-process inventory level.Therefore,the feeding strategy that can directly affect the number of work-in-process,especially the fixed work-in-process strategy has the greatest impact.A reasonable scheduling strategy can indirectly optimize the overall efficiency of the production line by accelerating the output of work-in-process.