Solution of Co-evolution Self-adaptive Genetic Algorithm to Production Scheduling Problem of Flexible Manufacturing System

Abstract

Improved random selection strategy, self-adaptive crossover and mutation operators, population co-evolution algorithm are adopted in this paper to set up a new genetic algorithm(GA), guaranteeing the universality of population, the thoroughness of solution and the execution efficiency of algorithm in solving multi-process, process crossover, work-piece flow and other condition based scheduling problem of flexible manufacturing system(FMS). The problem of GA in solving FMS scheduling is summarized in this paper, reasons and countermeasures for GA’s pre-maturing are studied, longitudinal and transverse comprehensive state-based correlative self-adaptive crossover and mutation operators are raised according to the analysis on GA to avoid local optimization and pre-maturity, the co-evolutionary GA is designed to solve the scheduling model for FMS, and the algorithm is validated for the scheduling of box-type parts for precise machine tools. 1. Overview The FMS scheduling problem is further researched constantly with FMS development. Not only work pieces, machine tool, cutter, checking-tool, pallet and other factors, but also disturbance information inside and outside the system must also be integrated into system in the real-time manner in the process of scheduling, researchers conducted researches on FMS scheduling from multiple aspects, and existing scheduling theories were supplemented, perfected and improved while new basic theories were proposed. Researches with variable process paths, diversified objectives, scheduling dynamics and algorithm complex were initiated; constantly improved self-adaptive GA, reactive and other intelligent scheduling strategies were built; processing objects, object grouping, equipment load balance and reasonable processing sequence suitable for FMS and others issues were researched in literatures; dynamics with different processing paths and identical processing under the premise that different machine tools are selectable are researched in literatures to conduct machine tool selection and process allocation; documentary research changes the sequencing method of the fewest parts in order through effective management on free time of equipment and rearrangement of task sequence; documentary study the way of multi-objective optimization in the production planning and scheduling. Researches on effects of longitudinal and transverse comprehensive state of on scheduling in evolutionary process are seldom though fruitful and comprehensive achieved are made. This paper is aimed at the demand of FMS dynamic scheduling and the internal and external interference factors to FMS scheduling, in the shortest production time, maximum utilization as the optimization goal, intelligent algorithm is chosen as the main technical means, is presented based on the state of vertical and horizontal to the evolutionarily related adaptive selection, crossover and mutation operators of co-evolution GA, and are studied the FMS production intelligent scheduling problem based on multi-procedure, crossing process, the work-piece flow combination and so on. 2nd Workshop on Advanced Research and Technology in Industry Applications (WARTIA 2016) © 2016. The authors Published by Atlantis Press 813 2. Self-adaptive Co-evolutionary GA to Solve FMS Scheduling Problem 2.1 Reasons and Countermeasures for GA Premature.GA serves as the model to a simulate natural selection and genetic mechanism of Darwin’s biological evolutionism, it is earlier applied in research in the field of production scheduling, and substantial corresponding research achievements are available, but it is vulnerable to weak local search ability, premature and slow late search as well as significant effects of the population size, probability of selection (Ps), probability of crossover(Pc), and mutation probability(Pm) on the algorithm performance, the two important embodiments of which are “population diversity’’ and “selective pressure”. Extra selection pressure can accelerate the rate of algorithm convergence, but it can make individuals in a population which are adverse to problem solving “end rapidly, destroying the diversity of population and having selection and crossover operators loss due effects, thus being trapped in local optimal solution and being “premature”. The decrease in selection pressure can increase the probability of search to the global optimal value, but will it reduce the search efficiency. Furthermore, the crossover operator and mutation operator are actually designed for specific variables, on one hand effects due to inaccurate fitness function can be accelerated, making that some individuals close to global optimal points are individual near local optimal points thereby decreasing opportunities to be selected; on the other hand the evolution of individual close to local extreme points selected are gradually perfected, thereby gradually increasing and approaching local extreme points, further decreasing selected opportunities of newborn individuals close to global optimal points in the evolution and causing the “premature” of populations. And biological evolution, the evolution of a species contains good genes can individuals in a population rapidly occupy the dominant position, but it does not lose its overall. In GA design, therefore, should be dynamically add new individual strategy, and with selective pressure and the diversity of population, crossover and mutation operators of phase equilibrium mechanism. In order to realize the requirement of intellectualization, through combined the thought of biological co-evolution, established the cooperative co-evolution GA to solve FMS scheduling problem. Population were randomly divided into several subgroups, various subgroups according to certain mode independently evolved, lateral groups for timely horizontal information sharing and co-evolution, embedded inherited the evolution of the vertical information within population factor, balances the global selectivity and population diversity, suppresses the premature phenomena, at the same time has high global optimization ability and faster convergence speed. 2.2 “Premature” Evaluation Methods for GA Population. Methods evaluating the GA population “premature” include spatial distribution variance method of population individual, entropy method of population, optimum fitness of population, differentials of average fitness of method, etc. However, spatial distribution variance method of population individual reflects the space dispersion degree of individual distribution of population but cannot fully reflect the diversity of population; the entropy method of population, optimal fitness of individuals and differentials of average fitness can't reflect “premature” degree of population individuals in a timely manner. Researches show that the fitness of individuals varies with different population distributions, hence the dispersion degree of population fitness distribution can be utilized to demonstrate the diversity of individual distribution within the population. Suppose that a population is comprised by individual M i X X X X , , , , , 2 1   (M denotes chromosome number in each population), and the fitness of each individual represents M i f f f f , , , , , 2 1   respectively. Therefore: ) , , , min( 2 1 min M f f f f  = (1) ) , , , max( 2 1 max M f f f f  = (2) max min f f f f i i − = ′ (3)