Hybrid CNN-LSTM model for fault diagnosis of rolling element bearings with operational defects

The operational reliability of rolling element bearings is crucial for the consistency of mechanical systems in various industries. Early detection of defects significantly enhances machine reliability and prevents downtime. This research presents a hybrid CNN-LSTM model for classifying faults in rolling element bearings with gradually developed wear defects using vibration signals. The study begins with a prolonged experiment in which bearings are run for 2000 h on a test rig to develop wear defects and vibrational signals are recorded at various intervals during the operation. The obtained signals are processed using Empirical Mode Decomposition to enhance signal quality and determine the optimum Intrinsic Mode Function for further analysis using the maximum energy ratio method. Then a hybrid CNN-LSTM model is designed and implemented to effectively classify various stages of bearing faults. The model achieved 99% accuracy in fault classification. These results demonstrate that the proposed model has a reliable diagnostic tool to improves the predictability and accuracy of fault identification in rolling bearings.