考虑声-振模态结合的抽水蓄能机组轴承故障诊断

为解决抽水蓄能机组轴承磨损故障难以监测和识别的问题,提出一种结合声振数据的双模态神经网络机组轴承诊断模型。首先分析了抽水蓄能机组声振特性,融合相似软阈值对奇异值分解去噪方法进行改进,有效消除非接触式传感器固有噪声干扰。其次提出逆巴克频谱变换方法,并结合巴克频谱变换和格拉姆角和场变换等特征工程技术,提取机组轴承的声纹和振动特征图。通过融合相对位置编码的自注意力机制和深度可分离卷积,建立特征图传递网络。同时运用多头自注意力机制和双向长短期记忆网络搭建了时序数据传递网络,并以平行网格架构构建了机组轴承故障诊断模型。实验对比分析表明,所提方法具有较高的故障识别准确率,为抽水蓄能电站机组轴承监测问题提供了有效的解决途径。

Prediction and Reduction of StructureBorne Noise in Vehicle

The structure and the acoustic medium of a passenger vehicle are modeled using the finite element method(FEM), and the interior noise is studied the help of the modal synthesis method (MSM). Sound pressure level (Lp) of the noise is calculated in several conditions of the models, and has good agreements with its test results. The MSM am be consequently used for predicting the vehicle interior noise in dssign stage so that the structure may be optimized for the Purpose of the most reduction of noise.