Fault diagnosis of hub motors based on artificial intelligence techniques

Electrical motors are a commonly used indispensable part of human life. Hub motors (in-wheel BLDC motors) are the members of BLDC (Brushless Direct Current) motors family, located at the intersection point of transportation area and electrical energy. They are also used in electrical vehicles and expected to be used more frequently in time. Hub motors are suitable for electrical vehicles structurally. Transportation is very important because of direct and indirect relation with human life. Therefore, hub motors must be more reliable and must be operating with less downtimes. The aim of fault diagnosis studies for hub motors is to make the hub motors more reliable and efficient. Hence, less downtime position for hub motors can be achieved. In this thesis, input variables of artificial intelligence techniques were determined firstly for detecting the differences of various faults by detecting the differences of input signals. Test set was designed for acquiring the determined data of torque, speed, source current, coil currents and source voltage as input variables for fault diagnosis of hub motor. Loading the hub motor mechanically is also possible with this test set. Feed-forward backpropagation neural network, cascade feed-forward neural network, Elman neural network, layer recurrent neural network and fuzzy logic based systems were designed and used for fault diagnosis of hub motor. Success percentages for fault diagnosis of all artificial intelligence techniques were tested and compared with eachother to choose the best performance technique for designing a real-time fault diagnosis system. Feed-forward backpropagation neural network was detected as the most successful artificial intelligence technique and used in the designed real time fault diagnosis system. 14 situations as 13 faults and normal situation, were successfully diagnosed. This study supports hub motors about safety and efficiency, with diagnosis of faults at beginning phase, with decreasing maintenance-mending costs, and with diagnosis of faults which reduce efficiency.