Design Agent-oriented Product Development Process Modeling And Simulation

With product complexity increasing and new product deveoplment model widely used,how to make a efficiency plan for product development process to improve product competitiveness, becames a important issue to academes and enterprises. Product development process(PDP)is a typical knowledge processing course, the effects of design agents and design organizations shouldn't be ignored. However, in most researches, design agents are treated as general resources, and the PDP model is task-oriented. In this model, the design agent's initiative and autonomy actions didn't be took account of and the design agent actions and organization strategies can't be analysed. According to the characteristics of PDP, and taking account of the initiative actions of design agents during PDP, a deeply study which concentrates on the PDP modeling and simulation method is implemented in this dissertation. The main contents consist of following parts:Firstly, through analyse the characteristics of PDP, base on CAS theory and MAS modeling principle, an agent-orientied PDP modeling method is proposed. According to the method, a PDP integration model is built which consists of design agent model, task model, resource model and product information model. The integration model centers on design agent model, and describes PDP as a continuance process, in which design-agents change the status of all tasks until the task is finished under the constraints of resources and information. Meanwhile, the integration model presents person-centered characteristics of PDP, and the initiative, autonomy and collaborative of design agents. In order to describe the design agent actions, a INTERRAP agent model is proposed. In this model, the behaviors of agent are divided into three layers—rebehavior layer, plan layer and collaboration layer. Then, the process of agent working and collaboration is analysed and described by hierarchical colored stochastic petri net. Furthermore, under the basis of modeling analysis, the main behaviors of design-agent in three layers are strictly defined by predicate logic of artificial intelligence methods. The definition is a foundation for further simulation system building.Secondly, organization performance evaluation of PDP is studied, and a evaluation indices system is proposed. The influencing factors of organization performance can be divided into two aspects, macro-level and micro-level. The factors in Macro-level include organization structure, task breakdown, work enlargement, work enrichment and job rotation. Micro-level factors mainly include the behavior of design agent, such as agent failures which include different types of design failure during working, agent decision behaviors which include task scheduling and collaboration partner selection behaviors. Then, base on influencing factors, an organization performance evaluation indices system which include Macro-level and micro-level as well is proposed. The indices system can not be used to analyse and optimize the organization performance but the behaviors of individual.Thirdly, an agent-oriented product development process simulation system(A-PDPS) is developed. According to PDP integration model, a multi-agents system which reflects the interbehaviors during PDP is built, Then a modeling and simulation platform of PDP is developed by Visual C++. The simulation system takes design agent-driving as simulation mechanism, and fixed increment advanced discrete simulation as foundation method. It realizes the dynamic simulation of design agents, tasks, resources and information, and can support the organization-oriented PDP performance evaluation. Furthermore, a railway design project example is simulated to testified the correction and validity of the simulation system. The simulation results indicate that the A-PDPS and development method can not only describe the initiative, autonomy and collaboration behaviors of design agents, but offer a more scientific basis for organization behaviors optimization.Finally, the influencing factors of organization performance simulation study is carried out. Then, base on organization strategies such as team configuration, organization structure, task scheduling behaviors and human reliability, we conduct experiment and analysis. Meanwhile, a task benefit-based agent-oriented dynamic task scheduling method is proposed. According to simulation results, design agent-oriented modeling and simulation method can be a useful approach for PDP forecasting, optimization, and organization performance improvement.