A Sound Monitoring System for Fault Detection of Machine and Machining States

This paper describes a sound monitoring system using a speech recognition technique. The system provides the untended manufacturing system with the function of overall monitoring capability. It identifies various operational sounds emitted by the machine, and check if they are proper ones which are expected from the control information prior to the operation. We adopted the speaker-dependent isolated word recognition technique based on the short time spectrum analysis for sound recognition. The experimental system was developed using the speech recognition board equipped with the special LSI. The recognition tests of the operational sound of a vertical machining center showed the ability of the system to identify the sorts of operations including the cutting and non-cutting operations. It can also discriminate the cutting sound with a broken tool from the normal cutting sound. The ability to distinguish the variations of cutting parameters was also studied. The system distinguished the variation of cutting speed and depth of cut fairly well. The monitoring experiments were carried out for a series of machining operations of a test part. The results demonstrated the effectiveness of the system as an overall monitoring method.