원문정보
초록
영어
In the traditional model of fault diagnosis, neural network classification requires high demand for the number and completeness of samples with a problem cannot be overcome- -"the curse of dimensionality". While the actual bearing failure is a typical case of small sample with few samples and the number of different types of samples is asymmetrical and even not complete. And the pattern classification effects of the support rector machine in case of small sample are better. Therefore, according to the above comparative analysis, combined with the character of small samples of actual bearing failure mode, this paper selects to build classification model based on the support vector machines, and after researching, the model proved to be feasible.
목차
1. A Non-Linear Support Vector Machine
1.1. Constructions of Nonlinear SVM
1.2. Kernel Function Selection of SVM
1.3. Classification SVM
2. Typical Optimization Algorithm Analysis
2.1. Simulated Annealing Algorithm
2.2. Genetic Algorithms
2.3. Ant Colony Algorithm
3. Seeker Optimization Algorithm
3.1. Basic Behavior
3.2. Calculations on Search Step Length and Search Direction
3.3. Simulated Annealing Algorithm
4. Analysis on the Performance of SOA
4.1. Simulation Analysis of Sphere Function
4.2. Simulation Analysis of Schaffer Function
4.3. Simulation Analysis of Rastrigin Function
5. Parameters Optimization Process of Multi-Class SVM Based On SOA
6. Conclusions
References