An Investigation To Module Knowledge Mining And A Modular Design Methodology For Products

For a typical modular product design,it is essential to consider modularity requirements and architectural constraints from multiple knowledge viewpoints(function,structure,and design parameter,etc.),and from several modularity perspectives(coupling-based and similarity-based).The design problem solving process is iterative.For the above,there lacks effective design process framework to support this iterative process from the studies published in literature.Whilst most of the existing module identification methods based on the Design Structure Matrix(DSM)aim to address these,they cannot deal with architectural constraints effectively,and even multi-objective product module identification methods often generate the solutions which cannot satisfy the practical partial grouping requirements of the special components.These limitations often lead to unreasonable module partition solutions for product architectures.Besides,there are problems of the dispersed distribution of protected parameters and the narrow range of choices of shared parameters in the existing parametermodule identification methods used for supporting secure collaborative design.Aiming to address the above problems,this dissertation is oriented towards the task requirements of modularization at the design stage from conceptual design to embodiment design for the redesign of mechantronic products,and conducts the research on the design methodology based on the module mining,coupling-based module identification method considering architectural constraints,parameter-module identification method used for supporting secure collaboration,and multi-objective product module identification method incorporating component grouping constraints.The main research contents are as follows:(1)A module-knowledge mining based design methodology is proposed.Firstly,the comprehensive product knowledge model is constructed.This model includes: the couplingbased design structure matrices(DSMs)within the function and structure viewpoint;the design function-design parameter matrix(FP matrix)within the design parameter viewpoint;mapping matrices across viewpoints;the similarity-based design structure matrix(DSM)and Design Product Matrix/Module Indication Matrix(DPM/MIM)within the structure viewpoint;and the architectural constraint sets within multiple viewpoints.Secondly,a module-knowledge mining based integrated design process framework is constructed.The framework includes relatively independent module identification units,which is effective to express the knowledge integration mechanisms between related units and to guide the current module mining process.(2)Two kinds of coupling-based module identification methods incorporating architectural constraints are developed.In the case of module identification problems incorporating both types of architectural constraints,a visual matrix based method is proposed.First,a genetic algorithm incorporating architectural constraints is used to generate automatically a set of optimized module partition solutions.Then,a grouping likelihood matrix(GLM)is obtained using these optimized solutions and is diagonalized to form a diagonalized GLM(DGLM).The color-encoded DGLM can help designers identify typical system structures and potential modules.In the case of the identification problems incorporating the special type of architectural constraints,a module identification method by diagonalizing the original DSM is proposed.In this method,in order to deal with architectural constraints,the two-dimensional solution encoding mode and two-dimensional neighborhood searching operators are developed;the artificial bee colony algorithm is utilized to optimize the sorting clustering criterion and obtain the diagonalized DSM.The generated diagonalized DSM and its corresponding module structure matrix visually display coarse-grained modules and overlaps among modules.(3)A design-parameter module identification method is proposed for supporting secure collaborative designs.It aims at protecting the critical design parameter information of original manufacturers from leaking in collaborative designs,thus the critical design parameters and related design functions should be placed in the same protected module in the FP matrix.Firstly,the FP matrix is clustered under the given architectural constraints.Secondly,the initial module partition solution is adjusted according to a measure called the degree of relative participation of the protected function in the protected module.Compared with the traditional methods,the proposed method can reduce the number of protected modules and enlarge the range of choice of shared parameters.This method also provides a module partition scheme adjustment tool to further reduce information leaking risk.(4)A multi-objective module identification method for product architecture incorporating component grouping constraints is proposed.Firstly,desired component partial grouping schemes are formulated according to results of DSM clustering and DPM/MIM clustering and other priori knowledge.Secondly,a clustering optimization model with partial grouping schemes as clustering constraints is constructed.In this model,three optimization objectives are considered: the modularity quality for DSM,the reangularity and the module coherence for DPM/MIM.Thirdly,the elitist non-dominated sorting genetic algorithm(NSGA-Ⅱ)is used to search for a Pareto optimal set for this problem.Subsequently,the technique for order preference by similarity to ideal solution(TOPSIS)is utilized to help decision makers select the proper compromise solution from the Pareto front.This method effectively solves the problems of generating optimal solutions which cannot satisfy the practical partial grouping requirements of the special components,and generating too coarse-grained modules in the traditional methods based on the DSM and DPM/MIM.(5)The high-speed turbo-pump is used as the application case study,and the proposed design method and the related module identification approaches are implemented and verified.The results show that the knowledge module partition solutions are reasonable,which can effectively support product update,maintenance and the collaborative design of the product system.