基于信息融合技术的高速冲床故障诊断研究

发布时间：2019-06-26 17:49

【摘要】：在故障诊断的实践中人们发现：(1)基于不同位置传感器的诊断结论有时会冲突；(2)基于不同的特征域的诊断结论有时会冲突；(3)基于不同的诊断推理方法的诊断结论有时会冲突。这些都是由于大型设备结构复杂和运行条件多样等所导致故障诊断过程中不确定大量引入,致使诊断的可靠性和准确性下降,难以满足日益大型化复杂化设备的故障诊断需求。为此,论证信息融合技术在高速冲床故障诊断中的应用,降低故障诊断的不确定性,提高设备的诊断精度显得尤为必要。本文主要是从理论上和实践中探索了信息融合技术在高速冲床振动故障诊断系统中的应用,将多个传感器信号、设备多个方面故障特征信息和多种故障诊断推理方法综合合理融合利用,最大限度降低诊断的不确定性,实现对设备全面与准确的诊断。其主要工作如下： (1)通过对信息融合技术和对故障诊断过程中的不确定性进行分析,采用信息融合技术在故障诊断中的理论框架,确立并采用了信息融合诊断组建方法,保证故障诊断过程中存在的不确定性经达融合后能够最大限度相互削弱,从而从理论上降低融合诊断的不确定性,达到精确诊断的目的。 (2)主元分析能够有效处理线性问题,核函数理论具有将低维非线性问题转化为高维线性问题的特性,将主元分析和核函数理论相结合,构成了核主元分析方法,使其对非线性问题具备非常强的处理能力。将其应用于机械设备故障特征压缩提取,经实验证实效果很好,从而成功解决多源信息融合诊断中信息量大且冗余的难题。 (3)归纳总结出神经网络在故障诊断中的具体应用方法,并通过实验分析发现,核主元分析与神经网络相结合能有效简化网络结构、缓减诊断推理的复杂度,从而提高了故障诊断的准确率。 (4)将证据理论与加权思想相结合,形成了加权证据理论。它通过对各证据进行加权组合,客观体现了不同来源的证据对识别框架中各真子集的识别具有不同的可靠性和权威性这一普遍事实,弥补了证据理论在应用中的缺陷,为证据理论在融合故障诊断中的应用打下基础。 (5)为了将多个特征域的局部诊断结果进行有效的决策融合,本文依据加权证据理论,通过构建加权证据理论在故障融合诊断中的具体实施框架,并遵循第二章确立的融合诊断组建方法,验证了基于加权证据理论的融合故障诊断方法。最后,对沈阳造币有限公司1号高速冲床进行实验分析,先分别从频域、时域和轴心轨迹三个特征域进行局部诊断,再将三个局部诊断的结果进行决策融合。实验结果表明：多故障特征信息融合后的诊断结果可信度明显增大,不确定性明显减小,故障诊断的准确率显著提高,充分验证了本文所采用的融合诊断方法的效性,并且该方法富有开放性、易实现,具有很强的工程实际应用价值。
[Abstract]:In the practice of fault diagnosis, it is found that: (1) the diagnosis conclusions based on different position sensors sometimes conflict; (2) the diagnosis conclusions based on different feature domains sometimes conflict; (3) the diagnosis conclusions based on different diagnostic reasoning methods sometimes conflict. These are caused by the complexity of large equipment structure and various operating conditions, resulting in a large number of uncertainties in the process of fault diagnosis, resulting in the decline of reliability and accuracy of diagnosis, and it is difficult to meet the fault diagnosis requirements of increasingly large and complex equipment. Therefore, it is particularly necessary to demonstrate the application of information fusion technology in fault diagnosis of high speed punch, to reduce the uncertainty of fault diagnosis and to improve the diagnosis accuracy of equipment. In this paper, the application of information fusion technology in high speed punch vibration fault diagnosis system is explored in theory and practice. Multiple sensor signals, equipment fault characteristic information and various fault diagnosis reasoning methods are integrated and utilized reasonably, so as to reduce the uncertainty of diagnosis to the greatest extent and realize the comprehensive and accurate diagnosis of equipment. The main work is as follows: (1) through the analysis of information fusion technology and the uncertainty in the process of fault diagnosis, the theoretical framework of information fusion technology in fault diagnosis is adopted, and the construction method of information fusion diagnosis is established and adopted to ensure that the uncertainties existing in the process of fault diagnosis can weaken each other to the maximum extent after fusion, so as to reduce the uncertainty of fusion diagnosis in theory. To achieve the purpose of accurate diagnosis. (2) Principal component analysis can effectively deal with linear problems. Kernel function theory has the characteristic of transforming low-dimensional nonlinear problems into high-dimensional linear problems. Combining principal component analysis with kernel function theory, a kernel principal component analysis method is formed, which makes it have a very strong ability to deal with nonlinear problems. It is applied to the fault feature compression extraction of mechanical equipment, and the experimental results show that the effect is very good, thus successfully solving the problem of large amount of information and redundancy in multi-source information fusion diagnosis. (3) the application method of neural network in fault diagnosis is summarized, and through experimental analysis, it is found that the combination of kernel principal component analysis and neural network can effectively simplify the network structure and reduce the complexity of diagnosis reasoning, thus improving the accuracy of fault diagnosis. (4) the weighted evidence theory is formed by combining the evidence theory with the weighted thought. Through the weighted combination of each evidence, it objectively reflects the general fact that the evidence from different sources has different reliability and authority for the recognition of each true subset in the recognition framework, makes up for the defects of the evidence theory in the application, and lays the foundation for the application of the evidence theory in the fusion fault diagnosis. (5) in order to fuse the local diagnosis results of multiple feature domains effectively, according to the weighted evidence theory, this paper verifies the fusion fault diagnosis method based on weighted evidence theory by constructing the concrete implementation framework of weighted evidence theory in fault fusion diagnosis, and following the fusion diagnosis construction method established in Chapter 2. Finally, the No. 1 high speed punch of Shenyang Mint making Co., Ltd. is analyzed experimentally. the local diagnosis is carried out from three characteristic domains: frequency domain, time domain and axis trajectory, and then the results of the three local diagnosis are combined. The experimental results show that the reliability of the diagnosis results after multi-fault feature information fusion is obviously increased, the uncertainty is obviously reduced, and the accuracy of fault diagnosis is significantly improved, which fully verifies the effectiveness of the fusion diagnosis method used in this paper, and the method is open, easy to implement, and has a strong practical engineering application value.
【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TG385.1;TH165.3

【参考文献】