A Fuzzy Decision System for Fault Classification Under High Levels of Uncertainty

Abstract

The problem of high levels of uncertainty existing in machine diagnosis is addressed by an approach based on fuzzy logic. In this approach, multiple sensors/channels are used, and the uncertainty is treated by membership functions in different stages of the signal processing. The concepts of fuzziness, fuzzy set, and fuzzy inference are described, particularly for the development of a practical procedure for machine diagnosis. The membership functions are established through a learning process based on test data, rather than being selected a priori. The information-gain weighting functions are also introduced in order to improve the robustness and reliability of this method. As a result, a framework of a Fuzzy Decision System (FDS) is proposed and applied to a machining process. Experiment verification with an optimistic success rate of 97.5 percent was achieved.