METHOD AND APPARATUS FOR DETECTING TROUBLE USING GENERAL REGRESSION NEURAL NETWORK

摘要

PURPOSE: A method and an apparatus for detecting trouble using a general regression neural network are provided to detect trouble of processing devices automatically controlled in real time. CONSTITUTION: A trouble reference setting is comprised of the steps of storing reference data of a transient state in which load changes and a normal state of initial operation, and of calculating a requested predicting value inputting the stored reference data and real measuring value into a general regression neural network. The trouble reference setting is further comprised of a step of setting an optimal smooth parameter and a trouble allowance reference. A trouble detecting is comprised of the steps of calculating a requested predicting value using the smooth parameter and the reference data, of calculating difference value between the predicting value and the real measuring value, and displaying whether or not trouble state is detected by comparing the difference value with the trouble allowance range value and discriminating the compared value.