基于数据分布的故障分类器核函数参数优化

Data Distribution Based Kernel Parameter Optimization for Fault Classifier

针对基于支持向量机的故障分类器的参数优化问题,研究了基于数据分布的支持向量机核函数参数优化的基本原理,提出了一种简化算法。依据该简化算法,实现了故障分类器的参数优化,并应用于汽轮发电机组的故障分类。测试结果表明,该参数优化方法可以提高故障分类器的分类能力,具有算法简单、优化效率高等优点,但其通用性有待进一步提高。

Aimed at parameter optimization of fault classifier, the principle of kernel parameter optimization based on data distribution was discussed. Then a simplified algorithm was proposed in order to achieve parameter optimization of fault classifier for turbo-generator sets. Testing results show that the classification capability of the fault classifier can be improved by using the simplified algorithm with high efficiency. But the adaptability of the simplified algorithm needs to be improved.