A diagnosis method for PS-FF based on KELM with IBKA optimization

Abstract The Phase-Shift Flip-Flop (PS-FF) is the key component or module of power sources, and its fault leads to abnormal operation or accidents. However, there are few investigations on full fault analysis and there are fewer effective fault methods. In this paper, the typical PS-FFs used in the rod-controlled power source of a nuclear action station is taken as an example, the fault effects and modes are analysed, and then a diagnosis method based on the Kernel based Extreme Learning Machine (KELM) model and the Improved Black-winged Kite Optimization Algorithm (IBKA) is proposed. The vulnerable components of PS-FF and its failure effects are first analysed, and then the fault mode and external characteristics are analysed by dividing PS-FF into the phase-shift trigger module and the pulse drive module according to the circuit function of PS-FF. By comparing the external characteristics, new features and extraction methods from the time-domain and frequency domains are proposed for PS-FF diagnosis and the separability of faults is verified with circuit simulation analysis. Then, by combining the KELM and IBKA methods and using the online collection of phase-shifted voltage signals, the IBKA-KELM model diagnosis method is proposed to detect and classify faults of PS-FF online. The simulation and experimental results show that this proposed method can diagnose PS-FF faults and has a high accuracy rate.