A Fault Diagnosis Method Based on Belief Rule Base With Multielement Knowledge Transfer for Complex Electromechanical System

This article proposes a fault diagnosis method based on information transfer and a belief rule base (BRB) to address the challenges in the diagnosis of complex electromechanical system faults under parallel test conditions. These challenges include extremely imbalanced samples, individual parallel test differences, and knowledge scarcity concerning the target individuals. This article also proposes a transfer evolution coefficient calculation method. A new transfer learning-based BRB (T-LBRB) model which considers the transfer evolution coefficient is established, and the transfer of multiple elements such as knowledge, rules, and optimization parameters is achieved. To analyze the impact of multielement transfer on the resulting diagnostic accuracy quantitatively, a sensitivity analysis of the transfer coefficient was conducted on the basis of the modeling transparency and traceability of the T-LBRB model. An experimental study is conducted on a simulation platform for a certain type of rocket subsystem to verify the effectiveness of the proposed method.