الفهرس Only 14 pages are availabe for public view
 |  | from | 110 |  |  |
| | | from | 110 | | |
Abstract
ABSTRACT
The present research involves the development of independent intelligent tools to diagnose and locate different faults in variable voltage variable frequency induction motor drive systems. The variable voltage variable frequency drive considered in this work is commonly used in many industrial applications. However, with the advances in solid state¬devices and variable frequency power converters, different approaches to induction motor drive systems have emerged.
Artificial neural network are utilized to diagnostic process after being trained based on indices derived at different healthy and faulty operating conditions.
Performance characteristics of the drive system are obtained under normal operation using the circuit arrangement, which is represented by the lumped-elements modeled under Matlap Simulink. The model is modified to accommodate different types of electric faults in the system in order to obtain faulty performance.
Healthy and faulty performance are compared to select one or more characteristic waveforms to be monitored in the time domain, and recording the average value of the rectifier output voltage, root mean square values of the inverter output voltages and Fourier transform magnitude of the fundamental component of motor input terminal voltage, with all values taken in per unit. Different characteristics waveforms and the recorded output values in per unit are considered the diagnostic indices to detect different faults.
The research focuses on electric faults of the drive system including open-circuit faults, short-circuit faults across the rectifier and the inverter electronic switches in addition to some types of motor faults.
The output waveforms and extracted data, which are used to perform diagnostic algorithm and train neural system, contain diagnostic indices extracted from simulated characteristics covering the majority of faults that are expected in the drive system. However, the detection of faults is achieved by extracted IF-THEN rules performing fault detection algorithms but the location of faults is incomplete. So that the conclusion of the logic program output can be treated by neural network depending on, the pattern recognition to classify and categorize the faults pattern compared with the healthy patterns in order to extract the decision by recalling the data of the e:xact fault.
The diagnostic artificial neural network classifier is tested within the training range and the results show acceptable performance of the developed neural system in diagnosing and locating the considered faults.
Experimental measurements are taken in the laboratory under most faults in order to verify the simulated results. Good agreement between simulated and measured values and waveforms is achieved.
In order to take into consideration the errors due to recording and transmission of the faulty system data, the ANN program has been tested under noisy conditions. The program response is correct up to a noise level 30 db.
IJ1