AI Detect: A Machine Learning Based Approach for Fault Identification in Gear Bearing System using Low-Frequency Data

Rotary machines are essential components of any industrial operation. A fault in these machines and their components can result in a series of damage causing a fatal failure of the system. This work aims to develop a low-cost technique for effectively diagnosing various faults in a gear bearing system. This research presents the design of a monitoring rig that uses an accelerometer to detect vibration profiles of different gears and bearings in a system using low-frequency data. Furthermore, statistical features are extracted from the said profiles using MATLAB. A machine learning model based i.e. a bagged tree algorithm is trained on these features and validated. The highest validation accuracy of 98.6% is achieved with the proposed method using a train-test split of 80% and 20%. To show the effectiveness of the designed algorithm, comparison against SVM and KNN method is also demonstrated.