Fault Detect And Diagnosis For AC Variable Frequency Drives

The fault detection and diagnosis for AC variable frequency drives is studied systematically in this dissertation. With the development of AC variable frequency drives and its widely application in industry,the reliability of the variable-speed drive system has been considered much more than ever. A great deal of research effort has been and being spent on the fault detection and diagnosis of the drive system. Based on existing research results,this dissertation first presents a new fault detect and diagnosis method for AC variable frequency drives based on signal processing. The main results and achievements in this dissertation are given as follows:1. Based on state-space method to study the inverter fed permanent magnet synchronous machine,the sorts of the switch fault are given and the causes for the faults are analyzed. It investigates the currents and torque of the drive system fed by six-step inverter and PWM inverter under the different fault condition,and the valuable results are got2. Aiming at the unsymmetrical three phase voltages from inverter after a failure,the dissertation get the relationship among the positive sequence ( PS ),negative sequence (NS ) and the instantaneous symmetrical component of the PS,this build the foundation to detect the fault and to get the fault feature.3. The dissertation proposes the extended Kalman filter based to estimate the fundamental symmetrical component of the PS and the frequency while the frequency is unknown. The concept of the Basic Spectral Residual is presented,a detection decision function is constructed and a simple strategy for fault detection is provided. This provide a valid way to detect the faults occurred in the variable frequency drives.4. Considering the long-term characteristics of fault detect or system monitoring,the paper propose the extended Kalman filter based on singular value decomposition to improve the numerical stability,and it accomplish fault detect or system monitoring based on the method.5. Based on the conventional Fourier transform and windowed Fourier transform,the dissertation analysis the spectral of the inverter output signal. It put forwards the spectral residual concept,the fault feature are obtained. A fault decision logic is presented,the fault state of the switch and the fault bridge of the inverter is fixed on.6. Considering the complex-value signals processing or parameter estimation,the dissertation study the complex-value least square method. The paper put forward complex recursive least square method either weighted data or using forgetting factor to avoid data saturation and increase the observed data reliability. The paper propose the complex value U-D factorization based least square method to improve the numerical stability and the fault feature are obtained.7. Based on analyzing the multilayer perceptron neural network,it induces the error backpropagation algorithm while taking the hyperboloid tangential function as nonlinear activation function. The dissertation put forwards the relative spectral residue concept,and get the relative spectral residue results when the inverter is under fault. The paper proposes the faults isolation method based on spectral estimation and neural network..