Early-Stage Fault Diagnosis for Rotating Element Bearing Using Improved Harmony Search Algorithm With Different Fitness Functions

Abstract

Bearings are the backbone of every rotary machine that ranges from civilian to military applications. A single fault in a bearing can shut down the whole machine that causing personal damage and economic loss. To prevent the sudden shut down of rotating machines, signal processing techniques are used to find out the early-stage faults. In this paper, a new early-stage fault detection algorithm is proposed that uses the Harmony Search (HS) algorithm to determine the optimum fault frequency spectrum. For that, a bandpass filter is applied to the vibration signal, and the parameters of the bandpass filter such as center frequency, bandwidth, and order of the filter are dynamic and depend on the type of faults and the fault frequency resonance band. To estimate the dynamic parameters of the bandpass filter, different fitness functions are used based on kurtosis and spectrum kurtosis. The fitness functions of the optimum fault frequency spectrum have the highest value compared to the healthy frequency spectrum. The proposed method is fully data-driven, and HS algorithm is employed to optimize the parameters of the bandpass filter. The results of the proposed method have compared with the fast kurtogram, and it has concluded that the bandpass filter designed using the HS algorithm has better performance. To validate the proposed method, two datasets are employed and the simulation results are obtained using the MATLAB environment.