| Abstract [eng] |
Aim of research: Create neural network capable of analyzing and classifying vibration data of brushless permanent magnets DC motors in linear conveyor sistems. Thesis starts with review about, how brushless DC motors are in need for diagnostics and monitoring systems, and how the faults occur in them. While doing the research, in order to simulate system failure, there were created most common faults occurring in brushless DC motors. In the first stage of the research, the fault frequencies were analized using motor electrical current signature analysis method, or so called MCSA method. Vibrations occuring from faulty motor were collected using accelerometer. For vibration data analysis, neural classification network motel was created. The data collected from vibration spectrum allowed to classify data in quite high accuracy, average classification accuracy was 83.375 %. Together with classification network, there was another network, called pattern recognition, used to analyze vibration data. Pattern recognition network had mean square error value equal to 0.1, and overall fault type identifying network accuracy was 78.1 %. |
