基于改进混合蛙跳算法优化SVM的道岔故障诊断
Tumout Fault Diagnosis Based on Improved Shuffled Frog Leaping Algorithm to Optimize SVM Model
孙波 1孟庆虎 1何晖2
作者信息
- 1. 山东科技大学电子信息工程学院,山东青岛 266590
- 2. 湖南华慧特自动化科技有限公司,湖南长沙 410002
- 折叠
摘要
针对道岔故障难以模拟导致的故障样本少、故障诊断困难等问题,提出一种改进混合蛙跳算法优化的支持向量机模型,基于小样本数据进行道岔故障诊断.支持向量机需要对参数择优选择,否则会造成过拟合或者欠拟合现象.将差分进化算法及模拟退火算法与混合蛙跳算法相融合,解决了混合蛙跳算法易陷入局部最优的问题,并将其用于优化支持向量机参数,提高支持向量机模型的故障诊断能力.通过对实测数据进行试验,测试结果表明:在相同条件下,本文提出的模型比支持向量机模型与混合蛙跳算法优化的支持向量机模型的平均故障诊断准确率提高了 34.28％,比仅融合差分进化算法的混合蛙跳算法优化的支持向量机模型的平均故障诊断准确率提高了 5.71％.故障诊断结果表明,本文提出的方法对基于小样本数据的道岔故障诊断更加有效.
Abstract
In response to the problems of limited fault samples and difficult fault diagnosis caused by the difficulty in simulating turnout faults,this paper presented a support vector machine(SVM)model optimized by improved shuffled frog leaping algorithm(SFLA)to diagnose turnout faults based on small sample data.SVM requires optimal parameter selection to avoid overfitting or underfitting.In this paper,when the integration of differential evolution algorithm,simu-lated annealing algorithm and the SFLA solved the problem of the SFLA easily falling into local optimum,it was used to optimize the parameters of SVM to improve the fault diagnosis ability of the SVM model.The measured data of experi-ment show that under the same conditions,the proposed model improves the average fault diagnosis accuracy by 34.28％compared with the SVM model and the SVM model optimized with SFLA and improves the average fault diagnosis accura-cy by 5.71％,compared with the SVM model optimized by the SFLA that only integrates the differential evolution algo-rithm.The method proposed in this paper based on small sample data is more effective for turnout fault diagnosis.
关键词
道岔/故障诊断/支持向量机/混合蛙跳算法/模拟退火算法/差分进化算法Key words
turnout/fault diagnosis/support vector machine/shuffled frog leaping algorithm/simulated annealing algo-rithm/differential evolution algorithm引用本文复制引用
基金项目
国家自然科学基金(62073024)
北京市自然科学基金(L201006)
出版年
2024
NETL
NSTL
万方数据