摘要
近年来,电子产品的故障诊断与故障预测主要采用健康管理与故障预测(PHM)技术,但要准确预测其健康状态还是很难。以此为出发点,构建交换模块的状态预测模型,首先对训练数据进行预处理和统计分析,通过相关性分析初步得到影响交换模块状态的特征参数,通过特征选择进一步确定特征,然后通过算法比较选择机器学习算法中的K最近邻分类算法,通过参数优化最终得到交换模块状态预测模型。采用该方法进行了应用验证,获得交换模块状态预测准确度为99.8%,达到了较好的预测效果和精度。
In recent years,Prognostic and Health Management(PHM)is mainly used for fault diagnosis and fault prediction of electronic products,but it is still difficult to accurately predict the health status thereof.Based on this,a commutating module statuses prediction model is constructed.This prediction model first performs pre-process and statistical analysis on training data,and then preliminarily obtain feature parameters which affect the status of commutating modules by correlation analysis,determine features by the selection of the features.After these,using algorithm comparison to choose the K-Nearest Neighbor classification algorithm in machine learning algorithms,finally obtaining a commutating module statuses prediction model through parameter optimization.This method is used to perform application verification,obtaining the prediction accuracy of the commutating module statuses is 99.8%.Therefore,it achieves fairly good prediction effect and precision.
出处
《信息技术与标准化》
2022年第11期25-29,39,共6页
Information Technology & Standardization
基金
国防科工局“十三五”技术基础科研项目,项目编号:JSZL2018210C003。
关键词
机器学习
交换模块
K
最近邻分类
相关性分析
预测准确度
machine learning
commutating modules
K-Nearest Neighbor Classification(KNNC)
correlation analysis
prediction accuracy
