Sanna Pöyhönen

Numerical magnetic field analysis and signal processing for fault diagnostics of electrical machines

Numerical magnetic field analysis is used for predicting the performance of an induction motor and a slip‐ring generator having different faults implemented in their structure. Virtual measurement data provided by the numerical magnetic field analysis are analysed using modern signal processing techniques to get a reliable indication of the fault. Support vector machine based classification is applied to fault diagnostics. The stator line current, circulating currents between parallel stator branches and forces between the stator and rotor are compared as media of fault detection.

Fault diagnostics of an electrical machine with multiple support vector classifiers

Support vector machine (SVM) based classification is applied to fault diagnostics of an electrical machine. Numerical magnetic field analysis is used to provide virtual measurement data from healthy and faulty operations of an electric machine. Power spectra estimates of a stator line current of the motor are calculated with Welchs method, and SVMs are applied to distinguish the healthy spectrum from faulty spectra. Multiple SVMs are combined with a majority voting approach to reconstruct the final classification decision.

Signal processing of vibrations for condition monitoring of an induction motor

Vibration monitoring is studied for fault diagnostics of an induction motor. Several features of vibration signals are compared as indicators of broken rotor bar of a 35 kW induction motor. Regular fast Fourier transform (FFT) based power spectrum density (PSD) estimation is compared to signal processing with higher order spectra (HOS), cepstrum analysis and signal description with autoregressive (AR) modelling. The fault detection routine and feature comparison is carried out with support vector machine (SVM) based classification. The best method for feature extraction seems to be the application of AR coefficients. The result is found out with real measurement data from several motor conditions and load situations.

Raising the Dead: Extending Evolutionary Algorithms with a Case-Based Memory

In dynamically changing environments, the performance of a standard evolutionary algorithm deteriorates. This is due to the fact that the population, which is considered to contain the history of the evolutionary process, does not contain enough information to allow the algorithm to react adequately to changes in the fitness landscape. Therefore, we added a simple, global case-based memory to the process to keep track of interesting historical events. Through the introduction of this memory and a storing and replacement scheme we were able to improve the reaction capabilities of an evolutionary algorithm with a periodically changing fitness function.

Insulation defect localization through partial discharge measurements and numerical classification

Partial discharge (PD) analysis is a fundamental tool to guide decision making in electrical insulation diagnosis for condition based maintenance. In this paper, PD signals are analyzed to localize defects in insulation systems. The task of automatic defect localization with respect to electrodes has a wide range of industrial applications. In fact, depending on the apparatus type, risk assessment is remarkably affected by defect location with respect to the electrodes. In this study, various parameters are first extracted from PD distributions, and statistical analysis is performed to select the most significant parameters concerning localization. Then, the localization process is carried out through numerical classification. Three different classification methods are compared to find the best approach for this application. Comparing a k-nearest neighbor classifier, a probabilistic neural network and a support vector machine (SVM) based classifier, the best results are gained with SVM. although the former two are simpler to implement and easier to tune. SVM based classification has not been applied in PD analysis before this research.

Support vector classification for fault diagnostics of an electrical machine

Support vector classification (SVC) is applied to fault diagnostics of an electrical machine. Numerical magnetic field analysis is used to provide virtual measurement data from healthy and faulty operation of an electrical machine. Power spectra estimates of the stator current of the motor are calculated with Welchs method, and SVC is applied to distinguish healthy spectrum from faulty spectra. Results are promising. Most of the faults can be classified correctly.

Coupling Pairwise Support Vector Machines for Fault Classification

Different coupling strategies to reconstruct a multi-class classifier from pairwise support vector machine (SVM)-based classifiers are compared with application to fault diagnostics of a cage induction motor. Power spectrum density estimates of circulating currents in parallel branches of the motor are calculated with Welchs method, and SVMs are trained to distinguish a healthy spectrum from faulty spectra and faulty spectra from each other. Majority voting, a mixture matrix and a multi-layer perceptron network are compared in reconstructing the global classification decision. The comparison is done with simulations and the best method is validated with experimental data.