摘要
针对齿轮箱在复杂工况下运行时振动信号非平稳、传统信号处理方法难以准确识别故障的问题,提出了一种基于多传感器振动信号特征提取与随机森林(RF)、支持向量机(SVM)相结合的齿轮箱故障诊断方法。通过在齿轮箱不同位置布设4个加速度传感器,采集5种典型工况下的振动信号,采用等间隔分段方法提取脉冲因子、峭度因子及能量值等时域特征参数,构建多维特征集,并输入随机森林与支持向量机模型进行训练与分类。实验结果表明,该方法能够有效区分正常状态与多种故障状态,整体识别准确率达到96.54%,为工业齿轮箱的状态智能监测与故障早期诊断提供了可行的技术路线与实践依据。
Addressing the challenges of non-stationary vibration signals and inaccurate fault identification using traditional signal processing methods during gearbox operation under complex conditions,this study proposes a gearbox fault diagnosis method combining multi-sensor vibration signal feature extraction with Random Forest(RF)and Support Vector Machine(SVM).By deploying four accelerometers at different positions within the gearbox,vibration signals were collected under five typical operating conditions.Time-domain feature parameters-including pulse factor,kurtosis factor,and energy values-were extracted using an equi-interval segmentation method.A multidimensional feature set was constructed and fed into RF and SVM models for training and classification.Experimental results demonstrate that this method effectively distinguishes between normal operation and multiple fault conditions,achieving an overall recognition accuracy of 96.54%.This provides a feasible technical approach and practical basis for intelligent condition monitoring and early fault diagnosis of industrial gearboxes.
作者
吴宏彬
孙博宇
唐耀玲
韩炬
Wu Hongbin;Sun Boyu;Tang Yaoling;Han Ju(North China University of Technology,Tangshan Hebei 063210,China)
出处
《机械管理开发》
2026年第1期78-80,85,共4页
Mechanical Management and Development
基金
唐山市科技计划项目(22130219G)
中央引导地方科技发展资金项目(246Z1912G)。
关键词
齿轮箱
故障诊断
振动信号
随机森林模型
支持向量机
gearbox
fault diagnosis
vibration signal
random forest model
support vector machine
