Research On Quality Design Model For Products With Complex Quality Structure Based On MAS

The designing process determines the quality of products. In order to manufacture products ofhign quality, enterprises pay more and more attention to design quality, thus to get advantage inmarket competition. Strongly promoted by market competition and people’s requirement for productscontinually raised and, the quantity and complexity degree of comlex products increase sharply.The quality characteristics of complex products emerge in large numbers, and traditional robustdesign can no longer meet the demand of its quality design. Products with complex quality structureas the object, this paper studies the problems of its quality design. The framework of quality designsystem based on MAS is constructed and related models are studied for key agents of MAS to realizetheir functions. Aiming at a certain number of key problems during the process for quality design, aseries of research is conducted with a unique perspective, which provides a new train of thought forcomplex products’s quality design. The main work and result of this paper is as follows:(1) The idea of multi-agent system(MAS) is introduced into quality design for complex productsto coordinate the possible problems during complex products designing process. A framedwork forquality design of products with complex quality structure base on MAS is constructed, which consistsof three layers: organizing layer, coordinating layer and executing layer. There are five kinds of agentsin the quality design system based on MAS, which are M-agent, S-agent, C-agent, T-agent andR-agent respectively.(2) Based on the description and definition of common characteristics and modular for variouskinds of agents, as well as their fuction, specific structure of the various kinds of agents is continuedto be designed in this paper, and the modulars which should be included in each kind of agent is alsodetermined. The work flow of the quality design system based on MAS for products with complexquality structure is summerized and proposed.(3) According to the specific modular of M-agent, a model to realize the function of building andanalysing quality structure is constructed, as well as specific algorithms.Based on the concept ofquality structure, DSM is applied to express the complex interaction between numerous qualityelements. Improved PSO algorithm is used to realize the clustering for DSM, thus getting theclassification for quality elements of complex products. The analysis of products’ quality structure isthen conducted based on that, and the quality elements and their relationships are systimaticlymastered, which makes decompositon and allocation of designing Tasks more easy.(4) The model for M-agents to realize function of quality checking is built, and the algorithm to realize that is studied. On basis of ensuring the generalization ability of single BP Neural networkthrough optimization it by PSO algorithm, A quality prediction model based on neural networkensembles is constructed, which builds the input-output relationships between the numerous designingparameters and the overall quality characteristics. The M-agent is then able to realize its qualitychecking function successfully.(5) Bayesian network is applied to express the complex correlativity in quality for products withcomplex quality structure. The bayesian network is first developed based on the quality structure, andaccording to bayesian reasoning, the key nodes which affect the root node significantly are figured outwith the effective use of empirical data and prior probability. The influence degree of each qualityelement to the whole product’s quality is also analyzed through the bayesian network’s modeling andanalysis process. The modeling and analysis function of the S-agent laid a foundation for taskallocation and completion of the products with complex quality structure.(6) In order to coordinate the conflicts between T-agents, on basis of ordinary Taguchi parameterdesign method, the signal to noise ratio(SNR) is transformed into standard quality loss. Multipleregression analysis as the tool, minimization of comprehensive quality loss of the complex productsystem as the goal, level of each designing parameter is determined. Based on quality structure,ANOVA is used to determine the degree of importance of components’ quality variation to theresultant products’ robustness. The tolerances of various parameters under the best parametercombination program is determined when both manufacturing cost and quality level are considered.The method for parameter design and tolerance design for products with crossed quality structureproposed in this paper provides the basis and source of algorithm for C-agent to resolve the conflictsand to realize its coordinating function.Finally, on basis of summarizing the content of the whole paper, a prospect of possible researchtopics is made.