Abstract: Traditional modular design method seldom considers uncertainties throughout the full design process. A novel modular design method based on the fuzzy correlation analysis among the parts is proposed. A fuzzy evidential reasoning algorithm is developed to aggregate multi-criterion correlation degree among the parts and derives the comprehensive correlation degree among the parts. The mathematical models on the basis of the objectives, namely, high cohesion degree in a single module and low coupling degree among all the modules are established. To solve this model, a fuzzy non-dominated mechanism is introduced into the discrete multi-objective particle swarm algorithm and is then adopted to obtain the fuzzy non-dominated solutions. Technique for fuzzy order preference by similarity to an ideal solution is employed to select the optimal solution, which is the best modular design scheme. The modular design for a ram pump is taken as a case study to show the effectiveness and feasibility of the method we propose.

Key words: fuzzy correlation, fuzzy non-dominated solutions, particle swarm optimization algorithm, parts clustering