Multi-objective Cross-training Programming With Human Factors In Assembly Cell

Uncertainty of production demand caused by rapid development of globalization brings rise to cross-training, which becomes a necessary means for a flexible assembly cell working effectively. Cross-training, as one of the key methods to realize workforce flexible (labor with multiple skills), concentrates on which tasks labors should be cross-trained, meeting with task coverage, labor multi-functionality and cost constraints, in order to achieve optimal operational cost, operational efficiency and humanized management.Existing factors considered in cross-training programming are mainly cost and operational efficiency. Human factors are less considered, which can not reflect practical problems completely. Researches on cross-training programming considering human factors and intelligent-based optimization methods can broaden and enrich theory and method of cross-training programming, which has application prospects for labor intensive assembly cell enterprises, and has profound significance of improving humanized management methods.This work was financially supported by the National Natural Science Foundation of China (NSFC)(70971019), the National Science Fund for Innovative Research Groups (70721001) and the Fundamental Research Funds for the Central Universities (N100404026). The background of this dissertation is cross-training in assembly cell environment. After analyzing factors of implementing cross-training, we focus on cross-training programming problem with human factors and designing evolutionary algorithms (MPSO and NSGA-II) to solve the cross-training problems. The main work contents are as follows:First of all, labor satisfaction, work efficiency and learning efficiency are proposed and quantitatively described as factors of cross-training programming in flexible assembly cell enterprises. The benefit and necessity of implementing cross-training programming in assembly cell are analyzed.Secondly, cross-training programming in flexible assembly cell environment taking labor satisfaction and the number of cross-trained workers into account is addressed. A multi-objective0-1programming model is presented to implement the cross-training programming. The model is solved by a two-stage method. A two-phase fractional transformation method is applied to convert the0-1fractional programming model to a mixed0-1linear programming model. Finally, optimization software, CPLEX is used to solve the problem to prove the model validity.Thirdly, cross-training programming problem from the point view of labor satisfaction and human cost is addressed. A multi-objective0-1integer programming model is presented to implement the cross-training programming. The model is solved by Multi-objective Particle Swarm Optimization Algorithm (MPSO) and Non-dominated Sorting Genetic Algorithm (NSGA-â…¡). The influence of operators of MPSO and NSGA-is analyzed, and then the performances of the two algorithms are compared. Finally, the influences of model parameters on cross-training programming are analyzed, and detailed strategies are given to provide guidance for enterpreneurs.Fourthly, cross-training programming problem considering labor satisfaction and work efficiency is addressed. A multi-objective0-1integer programming model is presented to implement the cross-training programming. The model is solved by MPSO using external archive and adaptive grid and NSGA-â…¡using double termination conditions and Pareto solution filter. The influence of operators of MPSO and NSGA-â…¡is analyzed, and then the performances of the two algorithms are compared.Finally, cross-training programming problem considering learning efficiency and cross-training cost is addressed. A multi-objective0-1mix integer programming model is presented to implement the cross-training programming. The model is solved by MPSO and NSGA-â…¡with elitist strategy. The influence of operators of MPSO and NSGA-â…¡is analyzed, and then the performances of the two algorithms are compared and evaluated by multiple metrics. A number of computational experiments show model and the two algorithms are effective.