Research And Implementation Of BLDC In Intelligent Control System

The feature of Brushless DC Motor (BLDCM) is that there is no machinerycontacting converter structure formed by brush, while it is replaced by electroniccommutator which is composed of electronic logic circuit and power switching circuit.Electronic commutator transfers DC power to alternating currents in the motorwinding, so it realized the function of DC power supply. BLDCM has the advantagesof AC motor and DC motor, such as simply structure, reliable operating andconvenient maintenance of AC motor, and high speed control property, high operatingefficiency and no excitation loss of DC Motor and so on. It is widely applied in servocontrol, NC machinery, robot control and other fields. With the development ofmodern industry, the performance requirement of BLDCM control system is rising. Sothe research on BLDCM system with fast response, strong regulation and high controlaccuracy is vital.Firstly, mainly focused on control methods of sensorless BLDCM, this thesisintroduces the research status and development trend of the BLDCM control system.Secondly, this paper analyzes its design features and principle of work, and sets up amathematical model of the BLDCM system. Then we put forward a new BLDCMcontrol scheme based on the Wavelet Neural Network (WNN). The relation betweenrotor position and phase voltage is derived by analyzing detection principle of motorrotor position. According to the principle of position detection and structural featuresof microcontroller, this method makes up a multi-input and single-output WNN withan input of three phase voltages and an output of bridge coded signal. The initialnumber of WNN hidden layer nodes is zero. In the training process, it adds orremoves hidden nodes continuously in accordance with the adaptive algorithm to forma simple and compact WNN structure. The network makes offline training and onlinetraining in gradient descent algorithm. Offline training initializes parameter andascertains the number of WNN hidden layer nodes, while the weights of WNN areupdated by online learning with filter and logical processed network output signal asteacher. So that the phase voltage directly maps the commutating signal and takesplace of the traditional position sensor. Thirdly, the method of direct voltage controlbased on WNN for BLDCM without position sensor is simulated. The output signal ofour method tracks the sample signal very well, which verifies the superiority of our scheme.Lastly, we design and manufacture a practical sensorless BLDCM control system.Use dsPIC30F3010as the core and C language as the programming language. Weperform the scheme of this paper and give controller parameters and experimentaldata. Experimental results show that our method can provide accurate motorcommutation signal and then to achieve control without position sensor with highcontrol accuracy and fast dynamic response.