An intelligent self-adaptive bearing fault diagnosis approach based on improved local mean decomposition

Bearings are one of the crucial elements in rotating machinery and their malfunctioning is the major reason of machine failure. The identification and diagnosing of bearing performance deterioration is critical for the smooth and reliable operation of rotating equipment's. This paper proposes an intelligent vibration-based condition monitoring and fault diagnosis methodology for detecting the bearing defects. Experimental vibration data acquired for different bearing and operating conditions are analysed to establish a framework for identification and diagnosis of bearing faults to assess the machine health. Fault diagnosis is carried out using Improved Local Mean Decomposition (ILMD) for decomposition of the vibration signal. The vibration features extracted from the obtained pre-processed signal were selected using Principal Component Analysis (PCA) to remove the redundant features. Subsequently, these relevant features were provided as input to the machine learning methods, namely Random Forest (RF), Party Kit (PK) and Support Vector Machines (SVM) for the detection and classification of the different bearing defects. Experimental outcomes demonstrate that the proposed methodology has an enormous potential to avoid the unplanned breakdowns, which are caused by bearing failure in rotary machinery.