Abstract

Roller element bearings are a common component and crucial to most rotating machinery; their failure makes up around half of the total machine failures, each with the potential to cause extreme damage, injury and downtime. Fault detection through condition monitoring is of significant importance. This paper demonstrates bearing fault detection using widely accessible consumer audio tools. Audio measurements from a smartphone and a standard USB microphone, and vibration measurements from an accelerometer are collected during tests on an electrical induction machine exhibiting a variety of mechanical bearing anomalies. A peak finding method along with use of trained Support Vector Machines (SVMs) classify the faults. It is shown that the classification rate from both the smartphone and the USB microphone was 95 and 100%, respectively, with the direct physically detected vibration results achieving only 75% classification accuracy. This work opens up the opportunity of using readily affordable and accessible acoustic diagnosis and prognosis for early mechanical anomalies on rotating machines.