Fault Diagnosis of Shaft Misalignment for Rotating Machinery using Artificial Neural Network

Experimental investigations utilized a rotor dynamic testing system to anticipate the vibration spectrum caused by shaft misalignment. These studies used a custom-designed, simple three-pin flexible coupling. A dual-channel vibration analyzer was used to evaluate rotor shaft accelerations in the presence of misalignment circumstances. Nevertheless, the 2X misalignment does not adequately align with any of the system’s natural frequencies to notably excite the system. If the 2X shaft operating speed coincides with or approaches one of the system’s natural frequencies, the impact of misalignment may be exacerbated and observable in the frequency spectrum. This study employs a Neural Network model for classification to facilitate fault detection and diagnosis in bearings using vibration data. To build features and train models, the machine learning toolbox in MATLAB is used, as is the diagnostic feature. The results of the confusion matrix also indicate that the model has a relatively high overall accuracy of 0.93 and a good balance of performance for different classes. Major Findings: The faulty machine vibration data is collected using a vibrometer instrument for the span of one month and it is diagnosed using Mat lab for better accuracy to increase the life of the motor is suggested in the present work.