An improved PSO-SVM approach for multi-faults diagnosis of satellite reaction wheel

Authors:
Di Hu;Yunfeng Dong;Ali Sarosh
Affiliations:
School of Astronautics, BeiHang University, Beijing, China;School of Astronautics, BeiHang University, Beijing, China;School of Astronautics, BeiHang University, Beijing, China
Venue:
AICI'10 Proceedings of the 2010 international conference on Artificial intelligence and computational intelligence: Part II
Year:
2010

Citing 5
Cited 0

The nature of statistical learning theory

The nature of statistical learning theory
Pairwise classification and support vector machines

Advances in kernel methods
Least Squares Support Vector Machine Classifiers

Neural Processing Letters
Reliability Modeling for Momentum Wheel Based on Data Mining of Failure-Physics

WKDD '10 Proceedings of the 2010 Third International Conference on Knowledge Discovery and Data Mining
Short communication: Multi-fault classification based on support vector machine trained by chaos particle swarm optimization

Knowledge-Based Systems

Quantified Score

Hi-index	0.00

Visualization

Abstract

Diagnosis of reaction wheel faults is very significant to ensure longterm stable satellite attitude control system operation. Support vector machine (SVM) is a new machine learning method based on statistical learning theory, which can solve the classification problem of small sampling, non-linearity and high dimensionality. However, it is difficult to select suitable parameters of SVM. Particle Swarm Optimization (PSO) is a new optimization method, which is motivated by social behavior of bird flocking. The optimization method not only has strong global search capability, but is also very simple to apply. However, PSO algorithms are still not mature enough for handling some of the more complicated problems as the one posed by SVM. Therefore an improved PSO algorithm is proposed and applied in parameter optimization of support vector machine as IPSO-SVM. The characteristics of satellite dynamic control process include three typical reaction wheel failures. Here an IPSO-SVM is used in fault diagnosis and compared with neural network-based diagnostic methods. Simulation results show that the improved PSO can effectively avoid the premature phenomenon; it can also optimize the SVM parameters, and achieve higher diagnostic accuracy than artificial neural network-based diagnostic methods.