双基球扁发射药生产线的改进遗传排产算法研究

Research on improved genetic scheduling algorithm for double based ball flat propellant production line

摘要:

针对传统遗传算法在解决批次生产问题中存在的“早熟收敛”以及“局部搜索能力差”等问题,设计了基于预处理技术的改进遗传算法,实现对批次生产过程的处理.采用随机数法、定则生成法和块基因插补法三种方法,按照合适的比例,进行种群的初始化,在保证初始化种群多样性同时提高其个体质量;通过精英保留策略和锦标赛选择策略进行选择操作,实现优质种群个体的选择;运用专家打分法对产品进行优先级排序;采用基于位置和优先级相结合的方法选择交叉位点,进行交叉操作,保留父代优良基因,避免“早熟收敛”;采用邻域重组策略进行变异操作,保证优质解种群的产生和质量解的继承.以最大化最小交货提前期为目标函数,实现排产算法研究.最后,以双基球扁发射药生产线为例,实现了改进遗传算法排产过程,大大提高公司的接单预估效率和产线的生产组织效率,然后运用单一随机初始化种群法和混合初始种群法进行比较分析,证明了改进算法的优越性.

Abstract:

Aiming at the problems of premature convergence and poor local search ability of traditional Genetic Algorithm(GA) in batch production process optimization, an Improved Genetic Algorithm(IGA) is designed by means of pretreatment technology. Random number method(RNM), rule generation method(RGM), and block gene interpolation method(BGIM) are used to initialize the population according to the appropriate ratio, which not only ensures the initial population diversity but also improves the individual quality. Elite retention strategies(ERS) and tournament selection strategies(TSS) are used to select individuals with good performance. A method is used to select crossover sites and take crossover operation based on the combination of location and priority obtained by expert marking, which preserves the superior genes and avoids “premature convergence”. The mutation operation is carried out via the neighborhood reorganization strategy(NRS) to ensure the generation of highquality solution populations and the inheritance of quality solutions. IGA regards maximizing the minimum delivery lead time as the objective function and realizes the scheduling algorithm research. Finally, the doublebase ball flat propellant production line is taken as an example to realize the scheduling process based on IGA, which greatly improved the order estimation efficiency of the company and the organization efficiency of the production line. The superiority of IGA is demonstrated by comparing both single and mixed initial population method.