Fault Detection And Fault Tolerant Control Design Of Batch Reactor

With complexity of modern control systems increasing and the requirement of higher performance of the system, reliability, maintainability and fault-tolerant capacity have attracted more and more attention of the world. For two decade, fault detection and tolerant control have made great improvement, and many remarkable results have been obtained.Starting with temperature control of the Vessel proceeding Batch Reactor, this paper makes a further studying on fault detection and tolerant control in condition of using cascade control to the reactor temperature as the master sensor fails to work, the main research work on it is done as follows:1) In this paper, for the use of artificial neural network to build the reactor control circuit fault diagnosis system. In order to ensure a complete network of training data, this article start from striking a system fault tree minimal cut , through the system failure mode to obtain training samples. In addition, taking into account the complicated and large-scale neural network will be faced with hard data collated teachers, low learning precision and low learning efficiency. In this paper, the composite fuzzy neural network technology was used to build the reactor's fault diagnosis system.2) An identification algorithm based on auxiliary models and least squares is developed and used to model the Batch Reactor. According to the characteristic of the reactor, the paper determines the model structure, and using auxiliary models to estimate Jacket temperature, cooling water temperature and the cooling water flow impact on the reactor temperature, and then, using least squares principle to produce the parameters. This method overcome the difficulties that standard least square does not work when the inner variables are unmeasurable. The simulated result shows that the algorithms proposed are effective and provides the basis for fault tolerant control.3) Using signal reconstruction to achieve a cascade control as sensor fails to work in the main control with proceeds mathematical mode. In the stimulation from fault detection unit, the feedback controller, with the inputs from the mathematical model estimates instead of the inputs from the master sensor, can make the systems in the normal operation. Finally, the simulation results show that, when the measurement of certain components in the system fails to work, leading to the lost of control signal, the assumptions to realize the system Fault Tolerant Control through the use of results calculated by the system model is totally feasible.