Mesterséges intelligencián alapuló módszer elektromechanikus aktuátor hibadiagnosztikájára

Abstract

This paper introduces a novel method to detect, identify and separate different faults may occur in complex mechatronic systems e.g. drive chains and actuators. The method is based on the feedforward artificial neural networks, which is known being one of the most popular family of artificial intelligence. In the course of introduction of the model-based method the development and selection of the used models, structure of applied networks, influence of used models on the results will be detailed. A complex evaluation strategy is introduced to assert the efficiency of the method and active training algorithm was used to train the separator neural network structure.