LightGBM-Based Fault Diagnosis of Rotating Machinery Under Changing Working Conditions Using Modified Recursive Feature Elimination

Abstract

This article presents an intelligent and accurate framework for fault diagnosis of induction motors using light gradient boosting machine (LightGBM). The proposed framework offers promising generalization ability when the testing data contains new unseen operating conditions unavailable during the training process. After the acquisition of vibration signals and feature extraction in multiple domains, we perform an iterative feature selection (FS) approach by utilizing a modified version of recursive feature elimination (RFE) and the features' importance scores obtained by LightGBM. To prevent overfitting and subsequent selection bias, an outer resampling loop encompasses the whole process of our RFE-LightGBM algorithm. Moreover, instead of the conventional resampling methods based on K-fold cross-validation (CV) or leave-one-out CV (LOOCV), we use a new scheme called leave-one-loading-out CV (LOLO-CV). Leveraging LOLO-CV, the proposed FS method identifies the optimal feature subset, making the fault diagnosis robust under changing operating conditions. Then, the final classification is performed with optimal feature subset by training a new LightGBM model with adjusted hyperparameters employing Bayesian optimization. Experimental results from two real case studies show that our proposed fault diagnosis framework achieves accuracies between 98.55% and 100% for various testing scenarios. For example, for the worst-case testing scenario in the bearing dataset of Case Western Reserve University where the no-load data (0hp) is absent during the training process and is only used for testing, the testing accuracy of LightGBM classifier before and after applying the proposed RFE-LightGBM-FS method is 88.04% to 97.23%, respectively. Using the Bayesian hyperparameter optimization further improves the accuracy to 98.55%.

Induction motor bearing fault classification using deep neural network with particle swarm optimization‐extreme gradient boosting

“…Furthermore, our proposed fault diagnosis method, which relies on the XGBoost model optimised using PSO, can obtain higher accuracy with faster training time than the nonoptimised XGBoost-based model. In contrast, DTL-based methods detailed in [16,43] have exhibited significantly higher accuracy and fast training time. However, the proposed method is still accurate and robust, demonstrating a lesser tendency of overfitting in the context of motor fault diagnosis.…”

“…These features are then fused and used as input to an XGBoost classifier for bearing fault classification. A recent classification model called LightGBM (Light Gradient Boosting Machine) has been utilised in the study by A.N.Saberi et al [16]. A LightGBM-based motor fault diagnosis model and Modified Recursive Feature Elimination have been proposed for rotating machinery under changing working conditions.…”

Knowledge features enhanced intelligent fault detection with progressive adaptive sparse attention learning for high-power diesel engine

“…(2) XGBoost [15] method: It utilizes predefined mechanistic features as inputs and employs XGBoost as the main structure for fault classification and diagnosis. (3) LightGBM [16] (6) Attention+ [30] method: It builds upon the M-1DCNN-Bi-LSTM method by incorporating attention mechanisms to achieve sensitive feature selection for efficient diagnosis. (7) Multi-level fusion+ [43] method: It extends the M-1DCNN-Bi-LSTM method by introducing a multilevel feature fusion approach for feature expansion and enhancement.…”

“…To more clearly demonstrate the fault diagnosis accuracy, as well as the overall accuracy and macro-avg-f1 across working conditions, we selected one experiment for each of the above 8 methods, which had diagnostic performance close to the average. The confusion matrix graphs for these experiments are shown in figures [13][14][15][16].…”

“…Guo et al [15] proposed a prediction method based on information fusion and XGBoost (Extreme Gradient Boosting) algorithm, which achieved feature selection and established a bearing remaining useful life (RUL) prediction model. LightGBM [16,17], an improved version of XGBoost, has also been widely applied in the field of fault diagnosis. These knowledge feature-driven methods have clear theoretical foundations and good interpretability.…”