Research On Production Scheduling Algorithm For Mechanical Products Based On Assembly Constraints

Production scheduling is an important part in mechanical products manufacturing, and it has direct relationship with production cycle, production cost and enterprise survivability. Excellent production sheduling schemes are helpful to enterprise in allocating different type production resource reasonably, reducing production cost, realizing just-in-time production and improving macket competitiveness. In the past, most of research on production scheduling focused on Job-shop Scheduling Problem-JSP without considering assembly stage of products. Assembly constraints of mechanical products must be considered in production scheduling because mechanical products are mostly assembly type ones. So another scheduling problem derives-AJSP (Assembly Job-shop Scheduling Problem) which is less investigated. Research in ths paper focuses on related algorithm for AJSP and has very important practical significance. In addition, problem modeling and algorithm operators designing are more difficult for existance of strong assembly constraints, so the research also has great theory value.Firstly, the paper describes assembly job-shop scheduling problem in detail through mathematical modeling and gives the classical assembly structure and assembly constraints relationship. Then the encoding scheme and selection operator are both determined when summarizing the key technology and evolution parameters of genetic algorithm (GA), and several kinds of fitness functions are designed. Two genetic algorithms are proposed for AJSP-Entire Solution Space Genetic Algorithm and Feasible Solution Space Genetic Algorithm (FSSGA). The design of repair operator is the most difficult point in designing entire solution space genetic algorithm. Two new repair operators TDRA (Top-Down Recursively Adjustment) and GE (Genes Exchange Based on Father Link-List) are proposed in the research. They are compared in detail with operator based on Design Structure Matrix-DSM in information entropy loss and chromosome map, and the results show that TDRA and GE can reserve population diversity better than DSM in GA. Subsequently, solution space size is analyzed and computing method is illustrated by example, and it's proved that for AJSP the feasible solution space size is much smaller than the entire one. Based on the above conclusion, the feasible solution space genetic algorithm is proposed and corresponding feasible solution space operators are designed to serve FSSGA. In that case, the searching solution space is much smaller than before and the repair operation is saved, so the algorithm efficiency is improved greatly. In addition, large scale problem experiment, literature problems comparison experiment and engineering instance experiment are performed.Finally, very good results are obtained in a series of experiments for new algorithm and operators. Therein the result of repair operators comparison experiment proves that two new repair operators reserve more population diversity and are helpful for GA to evolve and search optimal solution. The GA experiment result indicates that entire solution space genetic algorithm has preferable feasibility and optimization ability. The results of algorithms comparison experiment and last literature problems comparison experiment both demonstrate that FSSGA is much better than other algorithms not in performance and quality. And the result of engineering instance experiment indicates that the designed algorithms can be applied to practical situation very well.