An application of Fuzzy ARTMAP neural network to real-time learning and prediction of time-variant machine tool error maps

The problem of real-time learning of thermal error maps in machine tools is investigated. This problem is treated as an incremental approximation of a functional mapping between thermal sensor readings and the associated positional errors at each location of the cutting tool. The Fuzzy ARTMAP is used as a tool to achieve this approximation in real-time. Experimental measurements of the positional errors for a turning center were performed using a laser ball-bar over two separate thermal duty cycles. The Fuzzy ARTMAP was trained online using the data collected during the first duty cycle. Data from a new duty cycle is used to test the performance of the trained network. Results show that the Fuzzy ARTMAP is not only able to learn thermal errors in real-time but can also make accurate predictions of the test data