基于WT时频分析和最大类间阈值法图像分割的机械故障诊断

针对目前大多数机械故障诊断技术存在的建模困难、样本训练难度大等问题，提出一种基于连续小波变换时频分析和最大类间阈值法图像分割的机械故障诊断技术。即通过基于连续小波变换的时频分析技术，将测量获得的时域振动信号转换到时频域空间，然后通过可视化方法以振动热力图形式表征，通过图像强化对区域时频特征进行预处理，实现各频率成分分布的面积计算，以特征面积比的方式对振动信号进行分析。实验结果表明，该方法具有处理速度快、过程透明性好、适应性强等特点。

An Empirical Study on Mechanical Fault Diagnosis Technology Based on WT Time-Frequency Analysis and Maximum Inter-Class Threshold Image Segmentation

In view of the flaws of modeling difficulty and sample training difficulty found in most current mechanical fault diagnosis technologies, a mechanical fault diagnosis technology, which is based on continuous wavelet transform time-frequency analysis and maximum inter-class threshold image segmentation, has thus been proposed. By adopting a time-frequency analysis technology based on continuous wavelet transform, the measured time-domain vibration signal is transformed to the frequency-domain space, and then represented in the form of vibration thermal diagram by using the visualization method. The area calculation of each frequency component distribution can be realized by preprocessing the regional time-frequency features after an image enhancement, followed by an analysis of the vibration signal in the way of characteristic area ratio. The experimental results show that the proposed method is characterized with a fast processing speed, an improved process transparency, and a good adaptability.