Misalignment Detection of a Rotating Machine Shaft Using a Support Vector Machine Learning Algorithm

Abstract

Fault diagnosis for rotating machinery is important for reliability and safety, and shaft misalignment is one of the most common causes of mechanical vibration or shaft failure. In this study, we detected defects caused by misalignment of the shaft axis in rotating machinery using an artificial intelligence machine learning technique for fault recognition. Unlike other methods that focus on the time domain, we changed vibration data in the time domain to the power spectrum component value in the frequency domain. Then, the power spectrum values​ were classified using principle component analysis (PCA). Based on the results, defects are determined using a machine learning technique that uses a support vector machine (SVM). The results show that defects can be quickly determined from only vibration data about normal and abnormal states. By using the proposed pre-processing method, the machine learning framework based on vibration data can be effectively applied to diagnosis not only shaft defects but also faults of other rotating machines such as motors and engines