基于多尺度特征提取的风机音频信号故障诊断方法研究

针对风机传动链部件音频信号成分复杂,依靠单一特征提取方法难以实现故障识别的问题,提出了一种基于多尺度特征提取的风机主轴音频信号故障诊断方法。首先,采集风机传动链部件的音频文件,将其转换为数字信号后对数据进行了预处理;然后,对音频信号进行了多尺度特征提取,从时域、频域和倒谱域3个维度提取了5大特征,组成了多维复合特征矩阵,并对提取的特征进行了分析及降维;最后,利用支持向量机(SVM)分类预估器对多维复合特征矩阵进行了有监督学习,并使用粒子群算法(PSO)对SVM参数选择过程进行了优化,并通过多组对比实验,对PSO-SVM分类预估器在音频信号模式分类上的性能进行了验证。研究结果表明:所提取的多尺度特征能很好地表征音频信号的信息,具备一定的鲁棒性;使用经PSO优化后的SVM分类模型,对风机传动链部件音频信号模式识别的准确率可达98%以上,并具备良好的泛化能力。

Combined-Type Continuous Variable Transmission with Quadric Crank Chains and One-Way Clutches

The conventional Continuously Variable Transmission （CVT） has been classified as belt-type or toroidal CVTs. Each CVT is fundamentally composed of several components to transmit the frictional force or torque generated by the driving force. Since the conventional CVTs use friction force, their energy （force or torque） transfer efficiency might be inferior due to slippage and pressure between the transmission components. Consequently, we propose a new type of structural CVT. The CVT we proposed in this paper means a combined-type transmission with quadric crank chains and one-way clutches.