Scheme Generation And Evaluation Methods Of Conceptual Design Of Mechanical Products

Conceptual design is the beginning of product design process, and is also the key stage of product innovation, which has become the kernel step of product development and innovation. Conceptual design is the main measure to improve the quality and reduce the cost of products, and make the enterprise more competitive. According to the process of conceptual design, conceptual design of mechanical products mainly includes the following two phases: scheme generation and design candidate evaluation. With in-depth research on conceptual design, a number of tools and techniques have been proposed to address scheme generation and evaluation problems in conceptual design of mechanical products. But the intellectualization of conceptual design is always a bottleneck to restrict its development. In the recent years, intelligent optimization techniques have been developed rapidly, which attempt to set a theoretical basis to address the highly intelligentized and nonlinear scheme generation and evaluation problems of conceptual design.In this dissertation, based on the existing works, begin with the general process of conceptual design of mechanical products, scheme generation, especially design evaluation methods under different conditions are studied by using computational intelligence technique of fuzzy logic, neural network and genetic algorithm and intelligence optimization techniques of physical programming. The main research of this dissertation is as follows:(1) A scheme generation technique of conceptual design of mechanical products is studied. In view of the convergence speed of simple genetic algorithm is very low and the exploring process is easy to be trapped into local optimum, according to the known research, a modified adaptive genetic algorithm is proposed. On the basis of function analysis and morphological matrix of solution principles combination, a hybrid optimization model based on modified adaptive genetic algorithm and a BP neural network is constructed to perform the scheme generation of conceptual design process. An improved adaptive genetic algorithm is applied to explore schemes in the searching space of conceptual design, and a BP neural network is used to evaluate the fitness of the population in modified adaptive genetic algorithm. So the combination explosion and solving difficulty in knowledge representation of traditional morphological matrix can be avoided.(2) A design evaluation and decision-making problem when evaluation criteria can be quantified is investigated and discussed. To avoid the main drawback of typical decisionmatrix that the decision maker must specify physically meaningless weights, the linear physical programming (LPP) multi-attribute evaluation model proposed by professor Messac was used to design evaluation and decision making in conceptual design on the condition that the evaluation criteria can be quantified. Based on the specified interval of preference regions of every criterion, the weights used in aggregate objective function and the evaluation values of every design candidate can be calculated by LPP model automatically.(3) A design evaluation and decision-making method when evaluation criteria can't be quantified is studied. The information managed in conceptual design is incomplete, uncertain and imprecise. In most cases, design criteria are difficult to be quantified and are only described with fuzzy linguistic variables. Based on this consideration, a ranking based adaptive evolutionary operator genetic algorithm (RAOGA)-based fuzzy neural network (FNN) model is developed to evaluate design candidates in conceptual design under the condition that design criteria can't be quantified. In FNN model, a feedforward neural network is used to construct the network structure and a RAOGA-based learning algorithm is adopted to adjust fuzzy weights and thresholds with fuzzy inputs and outputs of FNN. Based on fuzzy set theory, evaluation criteria are fuzzified first, and then fuzzy decision-making is performed according to certain reasoning rules by FNN model.(4) A multi-level design evaluation and decision-making problem is studied when too many evaluation criteria and hierarchy of criteria exist. The disadvantages of multi-attribute evaluation model based on LPP and RAOGA-based FNN evaluation model are analyzed when the problems with many hierarchical criteria are evaluated. On the basis of the work of Vanegas and Labib, a multi-level conceptual design evaluation model based on NFWA and fuzzy compromise decision making method is constructed. In the multi-level evaluation model, the group AHP based on fuzzy Delphi method is used to identify the fuzzy number of weights of all the criteria, the NFWA and fuzzy compromise decision making method is applied to calculate the overall desirability of each design alternative level by level. The use of fuzzy Delphi analytical hierarchical process can guarantee expert knowledge is reflected to decision making process and make decision making process more practical. Fuzzy compromise decision making method can calculate relative approach degrees between design candidates and fuzzy ideal solution, which can improve the resolving power of the evaluation model validly.