Implementation of an On-Line Hammering Sound Inspection System Based on Fast Machine Learning Algorithm

Abstract

Although progress has been made on algorithms to automate sensory inspection based on machine learning, there are several problems yet to be overcome. We have developed an on-line hammering sound inspection system for an automobile product. Since an upper limit is set on the inspection time (per part), a support vector machine method, which is able to compute quickly, is applied to the on-line inspection. The experimental results show that the accuracy of inspection reaches 99.8% within the specified processing time. This paper reports on the implementation of the hammering sound sampling process, the inspection model training process, and the instrumentation of inspection system.