APPLICATION OF THE DISCRETE WAVELET TRANSFORM AND PROBABILISTIC NEURAL NETWORKS IN SI ENGINE VALVE FAULT DIAGNOSTICS

摘要

This article presents the attempt to detect the valve faults in the engine by using the vibroacoustic signal. The object of research was a four-cylinder 1.3 dm~3 capacity combustion engine. The vibration acceleration signals registered on the engine block were assumed the source of information on the engine condition. In case of diagnosing combustion engines by vibration methods, the presence of numerous sources of vibration cannot be neglected, which are the reason for reciprocal interference of symptoms of fault. Owing to the necessity of analyzing non-stationary and impulse signals, a discrete wavelet transform (DWT) has been applied in this study. Based on the signals' decomposition performed by means of the transform, the value of energy and entropy was determined, which served as a basis in the construction of the states of engine operation intended for teaching neural networks. As results from the research, there is a possibility of using probabilistic artificial neural networks to diagnose fatigue crack of an exhaust valve.