سال انتشار: ۱۳۸۴
محل انتشار: اولین کنفرانس بین المللی و هفتمین کنفرانس ملی مهندسی ساخت و تولید
تعداد صفحات: ۹
J Rafiee – M.Sc. student in Manufacturing Engineering
F Arvani – B.Sc. graduate in Manufacturing Engineering
A Harifi – Ph.D. student in Control Engineering
M.H Sadeghi – Assistant Prof. in Mechanical Engineering
This paper presents an optimized gear fault identification system using Genetic Algorithm (GA) to investigate the type of gear failure of a gearbox system using Artificial Neural Networks (ANN) with a well-designed structure suited for practical implementations due to its short training duration and high accuracy. Slight-worn, medium-worn, and broken-teeth of gears are categorized as gear faults. Wavelet analysis which is implemented for non-stationary signals, is capable of providing both time-domain and frequency-domain information simultaneously and therefore recognized in this research as the most reliable signal analysis method to extract a feature vector to train ANN using normalized wavelet packet energy rate index of the vibration signal. GA was exploited to settle on an optimized system by determination of best values for wavelet function type, decomposition level and number of neurons of hidden layer leading to a high-speed, meticulous two-layer ANN with a particularly small size.