نوع مقاله : مقاله پژوهشی
نویسندگان
1 دانشکده مهندسی مکانیک، دانشگاه گیلان، گیلان، ایران
2 دانشکده مهندسی مکانیک، دانشگاه یزد، یزد، ایران
3 دانشکده مهندسی برق، دانشگاه یزد، یزد، ایران
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
Vibration analysis is one of the most practical methods for condition monitoring and fault detection of rotating equipment. In this research, a method for condition monitoring and fault detection of locomotive blowers is presented which using vibration analysis and SVM neural network. For this point, after the data collection process of vibrating blowers, the received signals were classified into four groups: healthy blowers and with unbalance defects, base looseness and blade warping.
The received signals were also processed and 11 frequency properties and 5 time properties were extracted. The ratio of the extracted properties defined as new properties which can help to fault detection process. These features are then given as input to the SVM neural network. Too many features often confuse the neural network, which is why a T-test filter is placed inside the neural network to help select the right features. The results show that this filter can increase the accuracy of the neural network to distinguish healthy samples from defective ones from 84.9% to 97.9%. Finally, a two-class SVM neural network was implemented, with and without a T-test filter, and we show that the network accuracy increases with a T-test filter in all cases.
کلیدواژهها [English]