Improving the accuracy of estimating abnormality in gear device based on the degree of abnormality in vibration data

Abstract

In this study, we propose a method of determining the degree of separation between normal and abnormal vibration level distributions as a means of extracting characteristic information that is easy for users to understand from the vibration data obtained and improving the accuracy of abnormality diagnosis for a problem for which no evidence is generally provided during abnormality diagnosis by machine learning. We propose a method of determining the degree of separation between normal and abnormal vibration level distributions. We then apply this method to the unbalanced load and abnormal gear conditions of actual gear device, and show that it is possible to improve the accuracy of abnormality estimation by limiting the vibration data to frequency components and measurement positions where the degree of abnormality is significant.