An amended grey wolf optimization with mutation strategy to diagnose bucket defects in Pelton wheel

A deep learning-based defect identification scheme for the Pelton wheel has been developed. Initially, the raw vibration signal is passed through a time-varying filter based empirical mode decomposition (TVF-EMD). Filter parameters of TVF-EMD are optimized by a newly developed optimization algorithm i.e., amended grey wolf optimization (AGWO) with Kernel estimate for mutual information (KEMI) as its fitness function. The prominent IMF obtained is used to construct scalogram and prepare dataset. The training dataset trains the convolutional neural network (CNN) model whose accuracy was evaluated by the test dataset and founds to be 100%. The proposed AGWO algorithm was evaluated on twenty-three classical benchmark functions and the Wilcoxon test. Results obtained at benchmark functions and the Wilcoxon test validate the efficiency and superiority of the proposed method as compared to other techniques. A CNN classifier is compared with other learning models which suggested that CNN outperforms all learning models.