Autocorrelation energy and aquila optimizer for MED filtering of sound signal to detect bearing defect in Francis turbine

This paper presents a method to detect the bearing defects in Francis turbine by minimal entropy deconvolution (MED) filter making use of a sound signal. As the outputs of MED are mainly influenced by the filter length hence its appropriate selection is very necessary to recover a single random pulse in case of a weak faulty signal. The optimal filter length selection is done by Aquila optimizer adaptively which uses the autocorrelation energy as its fitness function. Experimentation done on defective bearings of Francis turbine suggested that the proposed method exposes periodic impulses effectively in case of a weak faulty signal or when the fault signal is embedded within the noise or interferences from other parts of Francis turbine. The proposed fault identification method has been compared with other models of MED such as particle swarm optimization -MED and maximum correlated kurtosis deconvolution. Results obtained reveals that the proposed method is superior in identifying the faulty signal embedded with heavy noise.

Automated summary

This paper presents a method to detect bearing defects in a Francis turbine by using a sound signal and a minimal entropy deconvolution filter. The method effectively exposes periodic impulses in weak faulty signals or when the fault signal is embedded within noise or interferences. Comparison with other models of minimal entropy deconvolution indicates that the proposed method is superior in identifying faulty signals embedded with heavy noise.