In fault diagnosis practice of recent years, the neural networks obtain many harvests, but it has a lot of questions in network structure selecting and network training. In the paper, the power spectrum of fault signals are decomposed by wavelet analysis, which predigests choosing method of fault eigenvectors, and a fault diagnosis model based on least squares support vector machine (LSSVM) is presented. Via structural risk minimization principle to enhance extensive ability, the model preferably solves many practical problems, such as small sample, non-linear, high dimension number and local minimum points. In the model, the non-sensitive loss function is replaced by quadratic loss function and the inequality constraints are replaced by equality constraints. The parameter of kernel function is chosen on dynamic to enhance the preciseness rate of diagnosis. The simulation results show the validity of the LSSVM model
[1]
Corinna Cortes,et al.
Support-Vector Networks
,
1995,
Machine Learning.
[2]
Alexander J. Smola,et al.
Support Vector Method for Function Approximation, Regression Estimation and Signal Processing
,
1996,
NIPS.
[3]
Liu Hao.
Method and application of fault diagnosis based on probabilistic neural network
,
2004
.
[4]
Michel Verleysen,et al.
Fast bootstrap for least-square support vector machines
,
2004,
ESANN.
[5]
Shie Mannor,et al.
The kernel recursive least-squares algorithm
,
2004,
IEEE Transactions on Signal Processing.
[6]
Ye Zhi.
PROBABILISTIC NEURAL NETWORKS BASED ENGINE FAULT DIAGNOSIS
,
2002
.
[7]
Chih-Jen Lin,et al.
A comparison of methods for multiclass support vector machines
,
2002,
IEEE Trans. Neural Networks.