Multi-variation propagation prediction based on multi-agent system for complex mechanical product design

Abstract

The design change propagation in a complex product is characteristic of nonlinear, dynamic, and uncertain; its impact analysis becomes challenging. Furthermore, the multiple changes occurring concurrently during the whole design process increase the difficulty of the change propagation analysis. In this article, a new change propagation prediction method based on multi-agent system, especially for multi-variations, is proposed to improve the validity of the change routing paths. As the foundation of the proposed prediction method, a hierarchical network composited of the specific design properties is used to represent the design change analysis model. Based on the discussion of concurrent propagation patterns in the context of multi-variations, the quantitative definition of change propagation impact focuses on two factors: change impact and propagation likelihood. To reduce the subjective measurements estimated through the designer’s experience, the change impact and the propagation likelihood are quantified by evaluating the design specification and mining the design change records, respectively. Based on the quantified change propagation impact, the multi-variation propagation simulation is parameterized and implemented in the platform of NetLogo. To demonstrate the effectiveness of the proposed method, experiments are conducted with one typical scenario. Compared with the classical change propagation methods, the proposed method proves to be valid for analyzing the concurrent change propagation, and the analysis reliability is guaranteed.