Real-Time Fault Diagnosis of Motor Bearing via Improved Cyclostationary Analysis Implemented onto Edge Computing System

Rolling bearing is a key component inside a motor, and its health status directly affects the operational reliability of the motor. Therefore, it is absolutely necessary to conduct research on bearing fault diagnosis. Most of the recent bearing fault diagnosis algorithms are implemented on a desktop/server, which might not satisfy the real-time diagnosis requirement in industrial fields. In this study, an improved cyclostationary analysis algorithm is proposed and implemented onto an edge computing system to diagnose motor fault in real time. Considering the traditional contact test methods based on vibration signal are difficult to carry out under complex working environments, sound signal is collected as the analysis object. Guided by cyclostationary theory, an improved cyclic feature enhancement algorithm is proposed and applied on the acquired signal to extract the distinct features related to the bearing faults. Simulation signal is constructed and analyzed firstly to verify the superiority of improved algorithm. Subsequently, experimental fault data is further analyzed to demonstrate the advantages of the proposed method. Furthermore, the proposed algorithm is deployed onto an edge computing system based on a micro controller unit. The online diagnosis result can be directly observed through an external display. The edge computing system with the embedded algorithm shows greatly potentials in motor real time fault diagnosis and intelligent maintenance.