Research On Control System Performance Assessment And Diagnosis

Control system performance assessment is important to ensure effective operation of control systems and save the cost of production. As an important part after control performance assessment, cause diagnosis of the poor performance has received more and more attention. Based on related references, the thesis focuses on the diagnosis methods to differ causes deteriorating control systems, from fault of control valves, model-plant mismatch, poor controller tuning and unmeasured disturbance aspects. The main contents and research results of the thesis are as follows:1. The causes of poor control system performance were analyzed, and the methods of loop oscillations detection were summarized. Based on existing researches of the control system performance diagnosis methods, the oscillations detection based control system performance diagnosis framework was established.2. In the diagnosis of control valve faults, an overview of the control valve nonlinearity detection methods was given. Comparative analysis and simulation examples tested the application conditions of the detection methods.3. In the detection of model-plant mismatch with unmeasured disturbances, existing diagnostic methods need to be taken to invasive test or join a certain stimulus. A time-frequency domain analysis diagnosis method was proposed to discriminate the causes. This method does not require external excitation conditions and the process model knowledge, and a multivariate system simulation proved the effectiveness. Meanwhile, cross-correlation analysis for the model-plant mismatch detection can be applied.4. For further discriminating between model-plant mismatch and controller tuning problem, the paper also proposed a sensitivity function diagnostic approach which is based on fast sampling identification technology. Confirm the existence of oscillations in the control loop, and the exclusion of control valve nonlinear factors cases, this method can distinguish the model-plant mismatch, the controller tuning problem and unmeasured periodic disturbance. Finally simulations proved its effectiveness.