الفهرس Only 14 pages are availabe for public view
 |  | from | 209 |  |  |
| | | from | 209 | | |
Abstract
ABSTRACT
In this thesis, an integrated sensor failure diagnosis and accommodation scheme is proposed. This scheme is applied to a boiler- turbine sensory system (strain gauge steam pressure, optical fiber current, capacitance probe water level, Zirconia substrate oxygen level sensors). The proposed scheme includes four stages; namely: sensor failures detection (SFD), sensor failures identification (SFI), sensor failure classification (SFC) and sensor failures accommodation (SFA). In SFD stage, Adaptive neuro-fuzzy inference system (ANFIS) is used to model each of the boiler-turbine sensors and characterized signals called ”residuals” are generated to declare the occurrence of a failure. The SFI stage receives residual signals and produces characterized signals that identify faulty sensors.
To implement sensor failures identifications (SFI), three techniques are proposed: adaptive neuro-fuzzy inference system (ANFIS), hierarchical structure of multi-layer feed-forward neural network (HSMFNN) and hierarchical structure of fuzzy neural network (HSFNN). SFI proposed techniques are able to diagnose either single or double abrupt or incipient faults. Sensor failure classification (SFC) is implemented using HSMFNN or HSFNN, SFA stage uses a group of adaptive neuro-fuzzy inference systems (ANFIS) to substitute failed sensor and continue the operation in a smooth manner.
Simulation of the proposed scheme proves robustness, high accuracy and less mathematical burdens.