Abstract
In the realm of manufacturing, industrial gearboxes play a vital function in systems and machinery, and any unforeseen breakdown or malfunction can result in costly downtime and diminished productivity. This paper proposes a Multi-Sensor Feature Fusion Framework for Predictive Maintenance of Industrial Mechanical Gearboxes via Vibration Signal Analysis based on machine learning. An overlapping sliding-window approach was used to segment the vibration data obtained from four accelerometer sensor signals, both in the healthy and broken tooth operating conditions. From each sensor, statistical and frequency domain features were extracted and feature-level fusion was performed to provide a single feature representation. Three supervised machine learning algorithms namely, Random Forest (RF), Support Vector Machine (SVM) and Extreme Gradient Boosting (XGBoost) were trained and evaluated under the same experimental conditions. The following accuracy, precision, recall, F1-score, receiver operating characteristic (ROC) analysis, confusion matrix, and training time are used to evaluate the performance of models. The results of the experiments have shown that the proposed framework can well differentiate the healthy condition and faulty conditions of the gearbox.