一种基于动态剪枝二叉树SVMs的高炉故障诊断新方法

Novel blast furnace fault diagnosis method based on dynamic pruned binary tree SVMs

高炉故障诊断是一个多类分类问题,且各个故障类别间具有一定的关系,在识别其中某一类故障时,并不需要区分全部故障的类别,为此提出了基于剪枝二叉树的支持向量机改进算法,每次识别时都去除相对没有价值的支持向量,根据类间相似度重新构造二叉树,剪掉没有价值的枝节,减少支持向量机个数,加速识别过程。通过对高炉故障模型的仿真实验,比较不同多类分类算法的性能,证明该方法能够在保证识别准确率的情况下提高故障诊断速度。

Blast furnace fault diagnosis is a multi-class classification problem, and the fault sorts have special relation among each other. It is not necessary to distinguish all the fault sorts when identifying one of them. In this paper, a novel algorithm based on pruned binary tree SVMs is proposed. In order to improve classification efficiency, we take out the relatively flimsy support vectors in identification process; construct a new binary tree without flimsy branches by defining similarities between every two sorts. Compared with different multi-class classification algorithms, the simulation results of blast furnace fault diagnosis show that this algorithm can improve the efficiency and speed of blast furnace fault diagnosis while insuring the identification accuracy.