|
基于GRP降维方法的电机故障诊断多源信号SVM预测
张镭 2026/5/21 20:18:11
山东兖矿技师学院，山东邹城，273500
摘要：针对故障信号支持向量机（SVM）降维问题，传统主成分分析（PCA）方法存在准确率偏低的问题，难以适应多源信号的需求。为此设计了一种基于高斯随机映射（GRP）降维的机床电机故障诊断多源信号SVM预测方法。利用随机映射算法来实现故障并达到对多源信息降维的效果。实验研究结果表明：采用GPR方法运算速度获得极大提升，识别精度达到99%以上。采用全局数据融合方案可大幅提高整体系统预测精度，使最终准确率提高到99.36%，取得了明显的效果。
关键词：电机；随机映射；支持向量机；故障诊断；多源信息融合
中图分类号：TH165
SVM Prediction of Multi-source Signals for CNC Machine Tool Fault Diagnosis Based on GRP Dimensionality Reduction Method
Zhang Lei
Shandong Yankuang Technician College, Zoucheng, Shandong 273500
Abstract: Regarding the dimensionality reduction problem of support vector machine (SVM) for fault signals, the traditional Principal Component analysis (PCA) method has the problem of low accuracy and is difficult to meet the requirements of multi-source signals. For this purpose, a multi-source signal SVM prediction method for CNC machine tool fault diagnosis based on Gaussian Random mapping (GRP) dimensionality reduction is designed. The random mapping algorithm is utilized to implement faults and achieve the effect of dimensionality reduction for multi-source information. The experimental research results show that the operation speed is greatly improved by using the GPR method, and the recognition accuracy reaches over 99%. The adoption of the global data fusion scheme can significantly enhance the overall system prediction accuracy, raising the final accuracy rate to 99.36% and achieving remarkable results.
Key words: CNC machine tools; Random mapping Support Vector Machine; Fault diagnosis; Multi-source information merge
为了确保机床电机能够实现安全稳定运行，需要对其初期出现的问题进行精确辨识，以防止可能出现的突发状况造成经济损失。目前，多源信息融合法由于综合了多种信息的数据分析过程，可同时具备高效与高精度数据处理能力，已在制造业中得到广泛运用[1]。液压系统故障通常具有突变性、不规律性的特点，且故障状态通常不易被察觉到，有些故障甚至还会出现多个并存或交叉的情况，再加上外部因素的综合作用，极大增加故障诊断难度[2]。
前期学者通过实验测试表明，采用随机投影(RP)的方式可以构建低维空间，从而实现各种智能工具的类似主成分分析功能，而且相较传统降维方法可以获得更快计算速度并降低算法复杂度[3]。李涛等[4]提出基于自适应观测器的故障诊断与容错控制方法，将电机故障与不确定性估计结果引入到积分终端滑模控制中，能够确保伺服电机准确跟踪指令信号。王正建等[5]提出基于智能诊断系统的故障维修技（未完，下一页）
附件下载：基于GRP降维方法的电机故障诊断多源信号SVM预测
|