In high-level automation industrial processes where maintenance or repair can not be carried out immediately, it is important to design autonomous controllers capable of maintaining the performance, reliability and safety of plants operating under sensor, actuator faults and failures, equipment fouling, feedstock variation. Advanced control strategies such as Active Fault Tolerant Control (AFTC) have been used to accommodate system failures automatically. This research presents an AFTC methodology using model predictive control (MPC) combined with a bank of Kalman Filters. This hybrid fault tolerant control system are testing in a linearized 14-order boiler-turbine unit to deal with sensor faults and actuator faults. When sensor fault occurs, the virtual sensor techniques, which uses both a bank of Kalman Filter and a reconfigured Kalman Filter is applied to estimate the plant state and corrupted sensor value. The reconfigured MPC controller, which has naturally ability in dealing with output and actuator constraints, is equipped with some advanced capabilities such as online parameter tuning mechanism, the stability improvement techniques, the feasibility improvement techniques and reference management technique to handle the plant actuator faults. In case of some specific actuator faults, the MPC controller is restructured to deal with the faults better. The proposed fault tolerant control successfully recovers the system performance in the sensor fault cases and some of the actuator cases. In other fault cases, where the system performance recovery is impossible due to faults, the fault tolerant control degrades the system performance to keep the system stable. / text
| Identifer | oai:union.ndltd.org:UTEXAS/oai:repositories.lib.utexas.edu:2152/19469 |
| Date | 15 February 2013 |
| Creators | Do, Khoa Dang |
| Source Sets | University of Texas |
| Language | en_US |
| Detected Language | English |
| Format | application/pdf |
Page generated in 0.0014 seconds