OPERATION MONITORING OF UNTENDED MANUFACTURING SYSTEMS BY MEANS OF SOUND RECOGNITION.

This paper describes a sound monitoring system which provides an untended manufacturing system with the function of overall monitoring to detect a wide range of abnormal occurrences. The system recognizes various operational sounds, including stationary sounds and shock sounds, using a speech recognition technique, and it then compares them with the expected operational sounds. To detect the emission of shock sounds which must be taken into the recognition system upon their emission, an algorithm using an autoregressive model is proposed. The experimental system was developed using a speech recognition printed circuit board equipped with special LSI hardware. The monitoring experiments are performed for the operation of a vertical machining center. The results show a considerably high rate of recognition for the sound of machine movements and the sounds of machining.