Research On Fault Diagnosis For Pneumatic Actuator System

Actuator, as one of the foundation equipment in Industrial Automation Process Control, is a final control element in the Automatic Control System. The development of the actuator affects heavily on the safety and reliability of the process of production. Because most of the actuators working under the formidable circumstance with high temperature, high pressure, and having corrosion, they are easy to make faults. It is necessary to increase the research of the pneumatic actuators. Therefore the development of pneumatic actuators should depend on advancing technique level, more importantly, it depends on the research of modeling. The research on actuator system's modeling not only forms the basic for the intelligent valve positioner algorithm design and the actuator's fault diagnosis, but also is the key for transforming these theories to actual product and technique.Firstly, this paper describes the components and working principle internal of the pneumatic actuators. It consists of three sections: the electro pneumatic valve positioner, the pneumatic diaphragm actuator and the control valve. The electro pneumatic valve positioner is composed of the controller, the torque motor, the pneumatic relay and the valve position feedback mechanism. The pneumatic diaphragm actuator consists of the pneumatic film and the spring mechanism. The open loop model can be obtained through analyzing the input and output data of the pneumatic actuator. Characteristic area method, rapid rising cure method and least square method are used to determine the system's model. By comparative analysis and the mechanical limit of the Apparatus itself, a kind of identification method with partial data is given. This method is proven to be effective both from theoretical analysis and experiment. Secondly, the common fault of the valve positioner, actuator and control valve is analyzed and diagnosed, combined with the use of BP neural network algorithm. The actuator system network structure is learned and trained using BP neural network algorithm with the experimental parameters. With 3-layer BP neural network to generalize of the actuator failures, the input data is the test parameters of the DABlib modules of MATLAB/Simulink different from the normal work and fault conditions, and the output data are different fault types of coding. Then the common fault of the valve clogging, valve plug or valve seat sedimentation, medium evaporation or critical flow, rod displacement sensor failure, positioner feedback fault, unexpected pressure change across the valve are diagnosed. Finally, in the MATLAB/Simulink environment, the test samples are trained by the actuator system network, and the simulation results shows that the algorithm is correct and feasible. In this paper, based on the principle of the actuator and system identification, the model of the pneumatic actuator is obtained and a method of system identification with partial data is given. Besides, the common fault of the valve positioner, control valve and pneumatic servomotor is diagnosed Using BP neural network algorithm. At last, the validity of the model and the BP neural network algorithm are proven under the MATLAB/Simulink circumstance.