风力齿轮箱轴承故障的AE信号特征提取与诊断方法研究
发布时间：2019-07-03 19:30
【摘要】：由于能源危机和环境污染问题日益严重,风能作为无污染可再生能源已受到世界各国的高度重视。随着风电装机容量的不断增加,齿轮箱故障发生率也不断升高,严重影响了风电的利用率。滚动轴承是风力齿轮箱中故障率较高的部件,轴承故障严重时会导致重大的事故。由于滚动轴承在故障形成初期及发展阶段都会产生声发射信号,所以采用声发射技术对其进行早期故障诊断,对避免重大事故的发生和降低运行维护成本都具有重要的意义。本论文研究目的是探索风力齿轮箱轴承声发射信号特征提取和故障诊断的新方法,以解决目前传统方法存在的抗噪声干扰能力差、参数选择复杂和模糊样本难识别等问题,以期提高特征提取准确性和故障诊断的正确率。本论文主要研究内容如下:首先,针对风力齿轮箱轴承声发射信号在采集时,单通道中存在着多故障源信号复合问题,提出了一种基于集成经验模态分解和改进的快速独立分量分析算法的单通道盲源分离方法。该方法将一维的单通道复合故障声发射信号通过集成经验模态分解算法分解成多维的本征模态函数分量,然后根据估计的源信号数目建立相同数目的输入信号,最后输入信号通过改进的快速独立分量分析算法进行分离。该方法解决了单通道信号盲源分离的欠定问题,克服了原快速独立分量分析算法对初值敏感的不足,对复合故障(损伤和断裂)声发射信号进行了有效的分离。其次,针对风力齿轮箱轴承声发射信号具有非平稳性和不确定性的特点,提出了一种基于完整集成经验模态分解和云模型特征熵的特征提取方法。该方法用完整集成经验模态分解算法将信号分解成多维的本征模态函数分量,由相关系数法选取出的敏感本征模态函数分量重构信号,再利用逆向云发生器计算重构信号的云模型特征熵作为信号的特征参数。通过实验测试与分析,该方法不仅有效的提取了声发射信号的特征,还克服了传统熵方法存在的参数选择复杂和阈值取值敏感等缺点。再次,为解决在风力齿轮箱轴承声发射信号进行特征提取时存在的强噪声干扰问题,提出了一种基于改进的集成经验模态分解算法和多尺度排列熵偏均值的特征提取方法。该方法首先通过云相似度法选取敏感本征模态函数分量,然后由敏感本征模态函数分量重构信号,最后计算重构信号的多尺度排列熵偏均值作为信号的特征参数。该方法克服了传统方法在选取敏感本征模态函数分量时存在的误判缺点,降低了噪声干扰,从而提高了特征提取的准确性。最后,为解决具有不确定因素的样本影响风力齿轮箱轴承故障诊断正确率的问题,提出了一种基于多维云模型确定度的模糊支持向量机故障诊断方法。该方法采用多维云模型确定度作为模糊支持向量机算法的隶属度,克服了传统模糊支持向量机算法不能将噪声或野值样本从有效样本集中区分出来的缺点。利用轴承故障声发射数据进行验证,结果表明该方法可以有效地抑制不确定信息(噪声或野值样本)的干扰,具有较高的故障诊断性能。
[Abstract]:As the energy crisis and environmental pollution are becoming more and more serious, wind energy, as a pollution-free renewable energy, has been highly valued by the countries of the world. With the increasing of the installed capacity of wind power, the fault rate of the gear box is increasing, and the utilization rate of wind power is seriously affected. The rolling bearing is a component with higher failure rate in the wind-driven gear box, and the bearing failure can lead to a major accident. As the acoustic emission signal is generated at the initial stage and the development stage of the rolling bearing, it is of great significance to use the acoustic emission technology to diagnose the early fault and to avoid the occurrence of major accidents and to reduce the operation and maintenance cost. The purpose of this paper is to explore the new method of the feature extraction and fault diagnosis of the bearing acoustic emission signal of the wind-driven gear box, to solve the problems of the current traditional method, such as the difference of the anti-noise interference ability, the complex parameter selection and the difficult identification of the fuzzy samples, In ord to improve that accuracy of feature extraction and the correct rate of fault diagnosis. The main contents of this thesis are as follows: First, when the sound emission signal of the bearing of the wind-driven gear box is collected, the multi-fault source signal composite problem exists in the single channel. A single-channel blind source separation method based on integrated empirical mode decomposition and improved fast independent component analysis algorithm is proposed. The method comprises the following steps of: decomposing a one-dimensional single-channel composite fault sound emission signal into a multi-dimensional intrinsic mode function component through an integrated empirical mode decomposition algorithm, and then establishing the same number of input signals according to the estimated source signal number, The final input signal is separated by an improved fast independent component analysis algorithm. The method solves the problem of the problem of the single-channel signal blind source separation, overcomes the defect that the original fast independent component analysis algorithm is sensitive to the initial value, and effectively separates the composite fault (damage and fracture) acoustic emission signal. Secondly, aiming at the characteristics of non-stationarity and uncertainty of the acoustic emission signal of the wind-driven gearbox, a feature extraction method based on the complete integrated empirical mode decomposition and the characteristic entropy of the cloud model is proposed. The method comprises the following steps of: decomposing a signal into a multi-dimensional intrinsic mode function component by a complete integrated empirical mode decomposition algorithm, and reconstructing a signal by a sensitive intrinsic mode function component selected by a correlation coefficient method, And then using a reverse cloud generator to calculate the characteristic entropy of the cloud model of the reconstructed signal as the characteristic parameter of the signal. Through the experimental test and analysis, the method not only effectively extracts the characteristics of the acoustic emission signal, but also overcomes the defects of the traditional entropy method that the parameter selection is complex and the threshold value is sensitive and the like. Thirdly, in order to solve the problem of strong noise interference in the feature extraction of the acoustic emission signal of the wind-driven gear box, a feature extraction method based on the improved integrated empirical mode decomposition algorithm and the multi-scale arrangement entropy bias means is proposed. The method comprises the following steps of: firstly, selecting a sensitive intrinsic mode function component by a cloud similarity function, and then reconstructing a signal from a sensitive intrinsic mode function component, and finally calculating a multi-scale arrangement entropy partial average value of the reconstructed signal as a characteristic parameter of the signal. The method overcomes the misjudgment of the traditional method when the sensitive intrinsic mode function component is selected, reduces the noise interference, and improves the accuracy of the feature extraction. Finally, a fuzzy support vector machine fault diagnosis method based on multi-dimensional cloud model determination is proposed in order to solve the problem that a sample with uncertain factors influences the fault diagnosis rate of the bearing of the wind-driven gear box. The method adopts the multi-dimensional cloud model to determine the degree of membership of the fuzzy support vector machine algorithm, and overcomes the defect that the traditional fuzzy support vector machine algorithm cannot distinguish the noise or the field value sample from the effective sample set. The results show that the method can effectively suppress the interference of uncertain information (noise or field value samples) and has higher fault diagnosis performance.
【学位授予单位】：哈尔滨工业大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TH133.3
本文编号：2509615
[Abstract]:As the energy crisis and environmental pollution are becoming more and more serious, wind energy, as a pollution-free renewable energy, has been highly valued by the countries of the world. With the increasing of the installed capacity of wind power, the fault rate of the gear box is increasing, and the utilization rate of wind power is seriously affected. The rolling bearing is a component with higher failure rate in the wind-driven gear box, and the bearing failure can lead to a major accident. As the acoustic emission signal is generated at the initial stage and the development stage of the rolling bearing, it is of great significance to use the acoustic emission technology to diagnose the early fault and to avoid the occurrence of major accidents and to reduce the operation and maintenance cost. The purpose of this paper is to explore the new method of the feature extraction and fault diagnosis of the bearing acoustic emission signal of the wind-driven gear box, to solve the problems of the current traditional method, such as the difference of the anti-noise interference ability, the complex parameter selection and the difficult identification of the fuzzy samples, In ord to improve that accuracy of feature extraction and the correct rate of fault diagnosis. The main contents of this thesis are as follows: First, when the sound emission signal of the bearing of the wind-driven gear box is collected, the multi-fault source signal composite problem exists in the single channel. A single-channel blind source separation method based on integrated empirical mode decomposition and improved fast independent component analysis algorithm is proposed. The method comprises the following steps of: decomposing a one-dimensional single-channel composite fault sound emission signal into a multi-dimensional intrinsic mode function component through an integrated empirical mode decomposition algorithm, and then establishing the same number of input signals according to the estimated source signal number, The final input signal is separated by an improved fast independent component analysis algorithm. The method solves the problem of the problem of the single-channel signal blind source separation, overcomes the defect that the original fast independent component analysis algorithm is sensitive to the initial value, and effectively separates the composite fault (damage and fracture) acoustic emission signal. Secondly, aiming at the characteristics of non-stationarity and uncertainty of the acoustic emission signal of the wind-driven gearbox, a feature extraction method based on the complete integrated empirical mode decomposition and the characteristic entropy of the cloud model is proposed. The method comprises the following steps of: decomposing a signal into a multi-dimensional intrinsic mode function component by a complete integrated empirical mode decomposition algorithm, and reconstructing a signal by a sensitive intrinsic mode function component selected by a correlation coefficient method, And then using a reverse cloud generator to calculate the characteristic entropy of the cloud model of the reconstructed signal as the characteristic parameter of the signal. Through the experimental test and analysis, the method not only effectively extracts the characteristics of the acoustic emission signal, but also overcomes the defects of the traditional entropy method that the parameter selection is complex and the threshold value is sensitive and the like. Thirdly, in order to solve the problem of strong noise interference in the feature extraction of the acoustic emission signal of the wind-driven gear box, a feature extraction method based on the improved integrated empirical mode decomposition algorithm and the multi-scale arrangement entropy bias means is proposed. The method comprises the following steps of: firstly, selecting a sensitive intrinsic mode function component by a cloud similarity function, and then reconstructing a signal from a sensitive intrinsic mode function component, and finally calculating a multi-scale arrangement entropy partial average value of the reconstructed signal as a characteristic parameter of the signal. The method overcomes the misjudgment of the traditional method when the sensitive intrinsic mode function component is selected, reduces the noise interference, and improves the accuracy of the feature extraction. Finally, a fuzzy support vector machine fault diagnosis method based on multi-dimensional cloud model determination is proposed in order to solve the problem that a sample with uncertain factors influences the fault diagnosis rate of the bearing of the wind-driven gear box. The method adopts the multi-dimensional cloud model to determine the degree of membership of the fuzzy support vector machine algorithm, and overcomes the defect that the traditional fuzzy support vector machine algorithm cannot distinguish the noise or the field value sample from the effective sample set. The results show that the method can effectively suppress the interference of uncertain information (noise or field value samples) and has higher fault diagnosis performance.
【学位授予单位】：哈尔滨工业大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TH133.3
【引证文献】
相关硕士学位论文 前4条
1 马雯萍;基于VMD的天然气管道泄漏信号特征提取与检测技术研究[D];东北石油大学;2018年
2 李红贤;齿轮信号干扰下风电齿轮箱轴承早期故障诊断方法研究[D];重庆大学;2018年
3 吴文轩;基于变分模态分解的齿轮箱复合故障提取研究[D];中北大学;2018年
4 陈炳光;基于EMD和SVM煤矿通风机轴承故障诊断的研究[D];中国矿业大学;2018年
，本文编号：2509615
本文链接：https://www.wllwen.com/jixiegongchenglunwen/2509615.html
