基于图像分解原理的旋转机械复合故障特征分离方法研究

The accurate separation of rotating machinery compound fault feature has great significance to enhance the accuracy of rotating machinery compound fault diagnosis. The rotating machinery compound fault feature manifests as the coupling of several single fault features like the fusion of image features and is hard to be separated, the ability of the image decomposition that can extract the image features effectively and separate different image features well is utilized in this project, and the separation method of rotating machinery compound fault feature based on the principle of image decomposition is proposed. The research is specified as the following three aspects. First, the dynamical models of the rotating machinery typical single fault and compound fault are built, and the dynamic behavior and response characteristics are analyzed, and the feature and evolution rule of the matrix data of single fault and compound fault are studied, and the extraction and representation method of basic feature and compound feature is formed. Second, when the basic features constitute the compound feature, the evolution rule of the basic feature and the overlapping and occlusion between basic features and it's evolution rule are analyzed, the compound characteristics of the compound fault is revealed. Third, basic feature matching method is developed, the search method of basic feature in compound fault matrix data is proposed, the decomposition method of compound feature is investigated, and the mapping relationship between decomposed basic feature and single fault is established. The project aims to solve the problem that the rotating machinery compound fault feature is hard to be separated and provide theoretical and technological support for rotating machinery compound fault diagnosis.

旋转机械复合故障特征的准确分离对提高旋转机械复合故障诊断的准确率具有重要意义。旋转机械复合故障特征表现为多个单故障特征如图像特征融合般耦合在一起而难以分离，项目利用图像分解对图像特征能有效提取和对不同图像特征能良好分离的能力，提出基于图像分解原理的旋转机械复合故障特征分离方法，具体进行以下三个方面的研究：(1)建立旋转机械典型单故障和复合故障动力学模型，分析其动力学行为和响应特性，研究单故障和复合故障矩阵数据的特征及其演变规律，形成基特征和复合特征提取和表征方法；(2)研究基特征构成复合特征时基特征的演变规律和基特征间的交叠和遮挡现象及其演变规律，揭示复合故障的复合特性；(3)构建基特征匹配方法，提出复合故障矩阵数据中基特征搜索方法，研究复合特征分解方法，建立分解基特征与单故障间的映射关系。项目旨在解决旋转机械复合故障特征难以分离的难题，为旋转机械复合故障诊断提供理论和技术支持。

旋转机械复合故障特征的准确分离对提高旋转机械复合故障诊断的准确率具有重要意义。旋转机械复合故障特征表现为多个单故障特征如图像特征融合般耦合在一起而难以分离，项目利用图像分解对图像特征能有效提取和对不同图像特征能良好分离的能力，提出基于图像分解原理的旋转机械复合故障特征分离方法。.(1) 旋转机械滚动轴承复合故障振动机理。基于接触弹性变形和油膜润滑理论，将滚动体、油膜、滚道三者之间的接触细化，构建了滚动体与滚道接触等效刚度与等效阻尼理论计算方法。针对滚动轴承内外圈等薄壁件的模态参数的需求，提出了单点激光连续扫描测振的薄壁件模态测试方法。引入单个故障源，将损伤点分别加载到外圈、内圈、滚动体上建立单故障的振动理论模型，采用四阶龙格库塔计算方法对模型求解获得单故障滚动轴承内圈的振动数据。分别在不同部件上引入两种故障复合、三种故障复合，建立了复合故障振动理论模型，对模型求解获得复合故障滚动轴承内圈的振动数据。最后对复合故障的滚动轴承振动特性进行了分析。将多种单故障同时加载到滚动轴承不同部件上形成的复合故障，滚动轴承内圈振动频谱特征信号并不是简单的单故障频谱特征信号的线性叠加，故障与故障之间复合存在振动频谱特征信号的非线性叠加，复合故障频谱信号包络幅值的大小与各个单故障的相位角差以及故障种类相关，外圈故障与内圈故障相复合时则会引发虚假滚动体故障特征频谱信号。.(2) 基于振动图像的旋转机械滚动轴承复合故障特征分离方法研究。基于S变换的构建滚动轴承故障振动图像，针对滚动轴承产生内外圈故障时振动表现为复杂的调幅振动的问题，采用Hilbert包络进行解调，采用S变换生成滚动轴承故障振动图像，获得的振动图像能够清晰表示各个故障频率段。研究了基于形态小波分解的图像融合方法，通过对异类的故障图像的融合研究获得复合故障形成机理。研究了基于灰度共生矩阵、灰度矩的振动图像故障特征表征与提取方法，通过灰度共生矩阵的特征参数优化，不仅能够很好提取纹理特征参数，还可以减少计算的工作量。研究了基于图像分解运算的复合故障特征分离方法，采用图像运算相关算子实现了图像灰度分布特征的分解，而图像上的灰度分布反映振动的特性。.项目研究对丰富机械故障诊断理论与技术，提高旋转机械复合故障诊出率有着重要的研究意义和应用价值，成果有着广泛的应用前景。