Intelligent Real-Time Predictive Diagnostics for Cutting Tools and Supervisory Control of Machining Operations

Abstract

Machining economics may be improved by automating the replacement of cutting tools. In-process diagnosis of the cutting tool using multiple sensors is essential for such automation. In this study, an intelligent real-time diagnostic system is developed and applied towards that objective. A generalized Machining Influence Diagram (MID) is formulated for modeling different modes of failure in conventional metal cutting processes. A faster algorithm for this model is developed to solve the diagnostic problem in real-time applications. A formal methodology is outlined to tune the knowledge base during training with a reduction in training time. Finally, the system is implemented on a drilling machine and evaluated on-line. The on-line response is well within the desired response time of actual production lines. The instance and the accuracy of diagnosis are quite promising. In cases where drill wear is not diagnosed in a timely manner, the system predicts wear induced failure and vice versa. By diagnosing at least one of the two failure modes, the system is able to prevent any abrupt failure of the drill during machining.