Fault diagnostics of electrical AC machines
In the VAIVI project, soft sensor methods are studied for fault diagnostics and condition monitoring of electrical AC machines.
A Support Vector Machine (SVM) is a modern machine learning method based on Vapnik's statistical learning theory. An SVM based classifier is claimed to have better generalisation properties than neural network based classifiers. In addition to this, the efficiency of an SVM based classifier does not depend on the number of features of classified entities. This property is very useful in fault diagnostics because the number of features to be chosen to be the base of fault classification does not have to be limited. An SVM has been successfully applied to numerous different classification and pattern recognition problems like text categorization, image recognition and bioinformatics.
In the project, for example SVMs are used for fault classification of electrical machines. Power spectrum estimates of motor currents differ from each other in different fault situations. An SVM based classification structure is trained to distinguish the faults in the machine based on the spectrum.
Also fuzzy logic based inference and model-based fault diagnostics methods (parameter estimation) are studied.
This project is part of SOSE.
Our research partner is the Laboratory of Electromechanics.
- Keywords: Fault diagnosis, support vector machine
- Duration: 2001-2004
- Research area: Power Electronics & Electric Machines, Process Control, Operations & Maintenance
- Related projects: TILLIKKA, SOSE
- Publications (total: 13)
- Heikki Koivo - person in charge
- Sanna Pöyhönen - project manager
- Jarmo Lehtonen
- Sanna Pöyhönen: Support vector machine based classification in condition monitoring of induction motors (2004)
- Sanna Pöyhönen: Support vector machines in fault diagnostics of electrical motors (2002)
- Jarmo Lehtonen: Neural network model-based fault diagnosis for induction motors (2004)