Fault Diagnosis Research on Three-Tank System
|School||Harbin Institute of Technology|
|Course||Control Science and Engineering|
|Keywords||Three-Tank System Fault Diagnosis T-S Fuzzy Model Unknown Input Observer|
This thesis aims at doing some fault diagnosis research on the DTS200 three-tank system, which has the ability to simulate various failures in practical situations, such as valve leakage, valve clogging, sensor failure, pump unworking, etc. The thesis mainly consists such contents as mentioned below:First of all, current research dynamics on three-tank system at home and abroad are introduced systematically, subsequently, general concept and commonly used methods on fault diagnosis are reviewed, the common results of current research and developing trends in the field of fault diagnosis are summarized here, too.Secondly, we construct the DTS200 three-tank system’s mathematical models, involving the failure system model, build the SIMULINK model and then establish the linear model respectively based on Taylor expansion and fuzzy T-S theory. After this, level control algorithms are studied. To control single input and single output system, a sub-level control method is given, comparison with general PID and fuzzy adaptive control method shows the superiority of the sub-level algorithm. Because higher order systems always have the characteristic of coupling, we apply state feedback decoupling algorithm on the double input and double output system to obtain two independent and decoupled subsystems, both simulation and experiment results demonstrates the effectiveness of the decoupling method.Finally, some fault diagnosis research is done based on double input and double output system. At first, we sum up different mathematical models of the system under various fault circumstances, describe their characteristics and give out the general form of the failure model. Furthermore, we study the Unknown Input Observer(UIO)’s application on the double input and double output system, the simulation result demonstrates that the method can detect the failure on time and estimate the fault precisely.