A copula entropy enhanced mRMR feature selection method for fault diagnosis of hydropower units

Abstract

To address the challenges of conventional feature extraction methods in capturing nonlinear dependencies within fault signals and reducing feature redundancy during hydropower units fault diagnosis, this paper proposes a feature selection framework integrating the minimum redundancy maximum relevance (mRMR) criterion with copula entropy (mRMR-CE). This framework utilizes CE to capture both linear and nonlinear dependencies in vibration signals. Combined with the mRMR criterion to suppress feature redundancy, it achieves stable selection of highly discriminative features. In noisy environments and with various classifiers, the method shows strong performance and stability. To validate the effectiveness of the proposed frame, seven feature selection approaches—CE, mRMR, mRMR-CE, Pearson, Principal Component Analysis, Hibert-Schmidt independence criterion—Lasso, and concrete autoencoder —were applied to the training samples during the feature selection stage. Lastly, the chosen features were input into four different types of classifiers for training and testing: support vector machine, Random Forest, multi-layer perceptron, and extreme gradient boosting. Experimental results demonstrate that the proposed method exhibits outstanding performance on both the Case Western Reserve University (CWRU) bearing fault dataset and the Unit 3 dataset from a hydropower plant. On the CWRU dataset, it achieved an average precision of 99.76% and an average F1 score of 96.57%, while on the Unit 3 dataset, it attained a 98.96% average accuracy and an average F1 score of 98.97%. These results significantly outperform traditional feature selection methods while demonstrating high stability and robustness.