复合无量纲免疫检测器在机组故障诊断技术的应用研究

发布时间：2018-02-08 17:44

本文关键词： 阴性选择算法旋转机械无量纲指标遗传编程免疫检测器集成诊断　出处：《太原理工大学》2013年硕士论文　论文类型：学位论文

【摘要】：人工免疫系统是对生物免疫系统的模拟,具有强大的信息处理能力,通过学习外界物质的自然防御机理的学习技术,提供噪声忍耐、自学习、自组织、记忆等进化学习机理,结合分类器、神经网络和机器推理等系统的一些优点。受免疫系统“自己”与“非己”识别机理启发得来的阴性选择算法为故障诊断的研究提供了新思想和新方法。本文研究一种基于阴性选择算法和无量纲指标的旋转机械故障诊断方法。主要工作如下： (1)针对现有的无量纲指标只对某些故障种类较为敏感,这导致了对其它一些故障种类分类效果可能不好。另外,随着旋转机械向集成化、精密化、复杂化的快速发展,旋转机械出现的故障种类越来越多,必要的无量纲指标也应该越来越多,而目前可供使用的无量纲指标数目有限,因此有必要针对旋转机械构造出一些新的无量纲指标,来克服传统无量纲指标诊断能力上的不足以及数量上的不足。本文利用遗传编程方法对已有的五种无量纲指标(波形指标、峰值指标、裕度指标、脉冲指标、峭度指标)进行重新组合和优化,构建出对于旋转机械设备较为常见的轴系、轴承座故障足够敏感的新的复合无量纲指标。结果表明,复合指标对常见的轴系、轴承座故障具有很好的分类效果。 (2)针对经典阴性选择算法产生检测器存在计算量大、盲目性强等问题,本文利用一种变异搜索方法来产生检测器,结果表明,该方法可以高效地产生检测器。 (3)针对构建无量纲指标免疫检测器过程中因进行约简及聚类等分类处理导致了其中一部分有用故障特征信息丢失的问题,本文研究一种简单、快速诊断的集成诊断算法来弥补,以提高机械故障的诊断准确率。最后在试验机组进行验证,结果表明,该方法有效提高了诊断准确率。
[Abstract]:Artificial immune system (AIS) is a simulation of biological immune system. It has powerful information processing ability. It provides evolutionary learning mechanisms such as noise tolerance, self-learning, self-organization, memory and so on by learning the natural defense mechanism of external substances. Combining the advantages of classifier, neural network and machine reasoning, the negative selection algorithm inspired by the recognition mechanism of immune system "self" and "non-self" provides a new idea and method for fault diagnosis. In this paper, a fault diagnosis method for rotating machinery based on negative selection algorithm and dimensionless index is studied. The main work is as follows:. 1) in view of the fact that the existing dimensionless indexes are more sensitive to some kinds of faults than others, the classification of other types of faults may not be effective. In addition, with the rapid development of integrated, precise and complicated rotating machinery, There are more and more kinds of faults in rotating machinery, and the necessary dimensionless indexes should be more and more. However, the number of dimensionless indexes available for use at present is limited, so it is necessary to construct some new dimensionless indexes for rotating machinery. In order to overcome the deficiency of traditional dimensionless index diagnosis ability and quantity, this paper uses genetic programming method to analyze five kinds of dimensionless indexes (waveform index, peak value index, margin index, pulse index). The kurtosis index) is recombined and optimized to construct a new composite dimensionless index which is sensitive enough to the shaft system which is more common for rotating machinery and equipment. The results show that the composite index is suitable for common shafting. Bearing bearing failure has good classification effect. 2) aiming at the problems of large computation and blindness in the generation detector of the classical negative selection algorithm, a mutation search method is used to generate the detector. The results show that this method can efficiently generate the detector. In order to solve the problem that some useful fault feature information is lost due to the reduction and clustering in the process of constructing a dimensionless index immune detector, a simple and fast integrated diagnosis algorithm is studied in this paper to make up for the loss of some useful fault feature information. In order to improve the accuracy of mechanical fault diagnosis, the test results show that the method can effectively improve the accuracy of diagnosis.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TH165.3

【参考文献】