论文部分内容阅读
针对高炉故障诊断系统快速性和准确性的要求,提出基于全局优化最小二乘支持向量机的策略.首先,采用变尺度离散粒子群对最小二乘支持向量机的参数和故障特征的选取进行优化;然后,利用核主元分析法对选取的特征向量进行压缩整理;最后,构造了以Fisher线性判别率为标准的启发式纠错输出编码.仿真结果表明,通过对故障训练样本有意义地分割重组,用较少的最小二乘支持向量机分类器,得到较高的故障判断准确率且增强了整个系统的实时性.
Aiming at the requirements of the fastness and accuracy of the blast furnace fault diagnosis system, a strategy based on globally optimized least square support vector machines is proposed.Firstly, the selection of parameters and fault features of least squares support vector machines is optimized by using variable size discrete particle swarm optimization Then, the selected eigenvectors are compressed by using the kernel principal component analysis method. Finally, a heuristic error correction output coding based on the Fisher linear discriminant rate is constructed. The simulation results show that by dividing the fault training samples into meaningful ones Reorganization, with fewer least square support vector machine classifier, get a higher fault diagnosis accuracy and enhance the real-time performance of the entire system.