الفهرس يوجد فقط 14 صفحة متاحة للعرض العام
 |  | from | 118 |  |  |
| | | from | 118 | | |
المستخلص
Fault diagnosis represents an important contemporary research field, due to the ever¬increasing need for safety, maintainability and reliability of industrial plants, communication systems and complex systems in general. The research in this field influeDCC8 important areas of our day-to-day life by increasing security when using safety¬critical devices, extending the lifetime of many expensive devices, and improving efficieDcy of manufacturing lines, which leads to smaller production expenses and lower prices for the end user.
’I’be main problems raised by the processes taking place within modem industrial
plants are their high nonlinearity, noisy signals, and uncertainty. Soft computing techniques - neural networks, fuzzy logic, genetic algorithms, etc. - are the answer to these
problems.
Fault identification represents the determination of the size and time-variant behavior
of faults. It facilitates the other fault diagnosis tasks which include fault detection and isolation. Therefore, fault identification is the most important of all the fault diagnosis tasks and it could be considered the main ingredient of the Fault Tolerant Control Systems
(FI’CSs).
This thesis proposes a new Fault Identification System using Sensors’ Measurements
(FISSM). The proposed FISSM is implemented using soft computing techniques. More specifically, the FISSM is implemented by two proposed genetically optimized models, the first ODe is Genetically Optimized Adaptive-Network-based Fuzzy Inference System (GO¬ANFIS) and the second is Genetically Optimized Feed Forward Neural Network (GO¬FFNN). For the FISSM implemented using a GO-ANFIS, subtractive clustering algorithm is used to find the data clusters in the input-output data space and hence find the fuzzy rules. OA was exploited to find the optimal values of the cluster radius and the number of delays used for each input or output sensor measurements. The proposed FISSM could be used to estimate the size of additive and multiplicative faults or functional faults in general.
The proposed FISSM could be used to make the control systems fault tolerant. Fault tolerant control is concerned with making the controlled systems able to maintain control objectives, despite the occurrence of faults. Identifying and compensating the faults that affect the output sensors permits the control system to keep a plant working sufficiently well until the necessary maintenance may be performed.
The performance of the proposed FISSM is tested using a nonlinear mathematical model of two Continuous Stirred Tank Reactors (2CSTRs) in series and an actuator model. the actuator model parameters ~ere tuned using experimental data achieved during the laboratory tests as well as data from a real actuator suited on the water inflow into a boiler drum in a polish sugar factory which is a benchmark problem studied within the project of Development and Application of Methods for Actuator Diagnosis in Industrial Control Systems (DAMADICS).
The obtained results show that the proposed FISSM could extract faults information
from sensors’ measurements and hence it could be used to detect, isolate and identify the faults affect the output sensors. In addition, it could be used to detect and isolate the faults 8ft’ect the input sensors or process components.
A proposed FTCS is tested using both 2CSTRs in series and DAMADICS actuator
models.