A Study on Cutting State Observation using Mahalanobis Distance.

In automated manufacturing processes, it is indispensable to have the ability to monitor and diagnose automatically the faults occurring in machining processes. To realize such ability, it is important to develop a pattern recognition from the measured value of the state to give correct judgment of the normality. This paper presents a new recognition process based on Mahalanobis distance. The distance is calculated with the so-called Mahalanobis Space that is constructed by characteristics derived from only the normal states' wave patterns of cutting vibration. The "differential and integral" characteristics were picked up from the wave patterns. Experimental result shows that an on-line monitoring with high sensitivity can be realized. And, a method to progress the sensitivity keeping the high speed processing was examined using Genetic Algorithm (GA).