Research On Automatic Generation Methods Of Dynamic Job Shop Dispatching Rules

Job shop scheduling is one of the important research contents of industrial production system optimization,and it has received extensive attention from academia in recent years.Due to the variability of the state of production nodes under complex working conditions and the randomness of production disturbances in the network environment,the scheduling problem is very challenging.In the existing job shop scheduling problems,optimization methods are usually dominant,but these methods mainly focus on static problems and simplified workshop environments,which often cannot be adapted to the dynamic characteristics of the actual workshop,and are too time-consuming and impractical.Therefore,the dispatching rules are proposed to find a fast and efficient solution,which is a very effective method to deal with actual constraints,and can cope with the dynamic changes of the workshop.However,designing an effective dispatching rule is not a simple task and requires a lot of knowledge about scheduling.The emergence of automated heuristics algorithm solves this difficulty.The goal of this method is to explore the "heuristic search space" of the problem,rather than the search space of the solution in the case of heuristics and meta-heuristics.The purpose of this method is to reduce the time required for manual design of dispatching rules,and to increase the opportunity to discover more powerful and undiscovered dispatching rules.Aiming at the actual production demand of dynamic job shop,the hyper-heuristic algorithm based on genetic programming is studied and applied to the automatic heuristic design of job shop dispatching rules.This article is mainly divided into the following four charpters:(1)Aiming at the actual production process of the dynamic job shop,we establish a mathematical model of multi-objective dynamic job shop scheduling problem,clarify the correlation between the optimization goal and the process parameters,and give the constraint conditions.The definitions of multiple types of dispatching rules,such as single rule,combined rule and weighted combination rule.The production characteristics of the dynamic job shop is futher analyzed,and the terminal feature set required for the automatic design of dynamic production dispatching rules is gived in this chapter.(2)Due to the uncertainty of the production status of the dynamic job shop and the randomness of tasks,it is difficult to find a general dispatching rule problem suitable for a variety of complex production scenarios.An automated generation method of dynamic job shop dispatching rules based on hyper-heuristic genetic programming is proposed to improve the dynamic adaptability of dispatching rules under different production scenarios.Through semantic analysis of the generated dispatching rules,the effect of GP tree’s features on different optimization goals is further studied.The experimental results show that the proposed algorithm can generate suitable dispatching rules for different production scenarios,and its performance is better than that of manually designed benchmark dispatching rules.(3)For large search space and long calculation time caused by the large and complex features of the above methods,a three-stage adaptive algorithm based on genetic programming is proposed,which selects feature sets through weighted voting and sorting,and keeps important features and eliminates irrelevant features.Compared with the dispatching rule before feature selection,the structure of the resulting dispatching rule is simpler and easier to understand,and it has a strong adaptive ability for the production of the actual dynamic workshop.The experimental results show that only using the remaining features can get better dispatching rules.(4)Based on the chapters above,we use the information integration platform independently developed by our team to develop a production scheduling software for machine tool plant in Wuxi.Based on Matlab,Oracle database,IDEA and other development software to complete the development of system function modules.The feasibility and effectiveness of the algorithm is verified by industrial application in the actual enterprise workshop.