基于粒子群算法优化支持向量机汽车故障诊断研究

汽车故障检测和诊断技术一直是国内外研究热点问题。支持向量机用于汽车故障诊断时,其多分类组合决策对分类正确率及诊断时间有很大影响,为了有效提高汽车系统故障诊断的效率和精度,提出了一种基于粒子群算法优化层次支持向量机汽车故障诊断检测方法。针对分解支持向量机具有测试时间短、结构难以确定的特点,利用粒子群算法,依据最大间隔距离原则优化层次支持向量机模型,使每个节点的支持向量机具有最大分类间隔,减少了误差积累,从而优化了多级二叉树结构的SVM,实现故障的分级诊断。仿真实验结果表明,提出的算法在所有参比模型中精度最高,能高效地对汽车系统的故障进行检测与定位,具有较强的泛化能力,同时缩短了故障诊断时间。

Automotive fault diagnosis based on SVM and particle swarm algorithm

Automobile fault detection and diagnosis technology has been a research hotspot. Support vector machine used in automobile fault diagnosis, the classification decision on the rate of correct classification and diagnosis time have great influence. In order to effectively improve the automobile fault diagnosis efficiency and accuracy, this paper proposed a method based on particle swarm optimization algorithm for hierarchical support vector machine fault diagnosis detection method. According to the decomposition support vector machine has short test time, is difficult to confirm the structure characteristics, this paper used the particle swarm algorithm, based on the maximum distance principle optimization of hierarchical support vector machine model, so that each node of the support vector machine had the maximal margin classification, reduced the error accumulation, thus optimized the multilevel binary tree structure of SVM, to realize fault hierarchical diagnosis. The simulation results show that the proposed algorithm, all the reference model of the highest accuracy, can be efficient for automobile system fault detection and location. This algorithm has strong generalization ability, and can shorten the time of fault diagnosis.