Real time expert system for predictive diagnostics and control of drilling operation

The suitability and applicability of a real-time expert system for integrating multiple sensors for predictive diagnostics, monitoring, and supervisory control of a drilling operation in an automated manufacturing environment were investigated. The real-time IDES (influence diagram based expert system) performs probabilistic inference and expected-value decision-making in integrating dynamic but noisy sensory data and subjective expertise in symbolic and numerical data structures designed for real-time performance. An application using spindle motor current, feed motor current, and spindle-mounted strain-gauge sensors on a numerically controlled drilling machine is described. In this example, with relatively simple signal processing, IDES achieves effective prediction about the state of the drill and optimally controls the performance of the drilling machine. The real-time expert system performance is demonstrated over a wide range of machining conditions: two workpiece materials, two drill sizes, six speeds, and seven feed rates. With an MS-DOS personal computer, the system was able to predict tool failure in 1.7 to 2.1 ms, well within the desired response time of an industrial production line operation