چكيده به لاتين
Transportation, especially rail transportation, has a great contribution to the movement of people as well as needed goods due to its convenience, speed, reliability, as well as safety and the fact that it is closely related to the transportation of passengers and goods.Therefore, the demand for this type of travel is increasing; Therefore, measures should be taken to maintain the equipment and prevent irreparable damage.
In this research, maintenance using online monitoring has been used for the traction motor of the locomotive, which is a series dc motor, so that the first signs of the fault can be detected as soon as they occur and before the failure occurs. In this research, 2 methods for fault detection are presented. In the first method, due to the effect of the defect on the frequency harmonics of the motor current signal, for 47 motors, some of which are healthy and some are defective, wavelet transformation has been applied to all current signals and the frequency response of all current signals has been obtained. Then the values of energy characteristics, average, maximum value (peak), curvature (kurtusis), standard deviation (standard deviation) and... were calculated for the high frequency part of the wavelet transformation for each current signal. According to studies, the values obtained for defective engines and healthy engines should show a significant difference; But with the implementation of this method, the desired result was not obtained and the second method was proposed. In the second method, one of the healthy engines was considered as the reference engine. Then, the same input was applied to this engine and 46 other engines that must be determined to be healthy or defective, and the output difference of all these engines was calculated with the output of the reference engine, and a residual signal was generated for each engine. Then the time characteristics of energy, peak, mean, root mean square, mean square error (mse) and steady state value of the signal were calculated for each remaining signal. Due to the fact that the characteristic values in defective signals are higher than in healthy signals, an attempt was made to determine a threshold value for each of the characteristics in order to create a clear boundary between the values related to healthy and defective signals. But there were several examples of contradictions in each feature individually. Having reached this conclusion, it was found that a solution should be used in which, by comparing all 5 features, the contradictions cover each other; Therefore, the values of these 5 features were applied in the form of a matrix with 5 rows and 46 columns to the input of a neural network with 3 layers and the number of 12 neurons in the second layer, and the neural network was able to detect about 90% of engines being healthy or defective. It should be noted that in the second method, all actions were performed in the MATLAB simulation environment using the simulation of the series dc motor model and the application of the parameter estimation algorithm to obtain the parameters of the motors.
