A novel fault diagnosis approach of a mechanical system based on meta-action unit

Abstract

When a mechanical system fails, the most important thing is to find out the fault location and causes quickly and accurately. Based on the meta-action unit, this article proposes a comprehensive, qualitative, and quantitative fault diagnosis method composed of fault tree analysis and evidential network to analyze faults of the mechanical system. First, we expand the function–motion–action decomposition method, revise the definitions of meta-action and meta-action unit derived from the function–motion–action decomposition of the complex mechanical system, use the fault tree analysis approach to analyze the faults of the meta-action unit instead of the system directly, and establish the meta-action unit fault tree analysis (MU-FTA) model with simple levels. These activities can tremendously simplify the qualitative diagnostic model. Second, the MU-FTA model is mapped to the evidential network model according to their similarities. Finally, we fuse the after-sales data and expert experience information by Dempster’s combination rule to calculate the basic probability assignments of the parent nodes in the bottom layer of the evidential network model. Thus, the basic probability assignments of meta-action units are obtained by network transmission and the diagnostic conclusion is acquired. The fault diagnosis of a certain type of numerical control turntable is conducted as an example to demonstrate and verify the feasibility and effectiveness of the proposed approach.