基于声发射的球轴承疲劳演化特征提取研究

发布时间：2018-09-10 20:17

【摘要】：滚动轴承是工业领域关键元件之一,剥落是滚动轴承重要的失效形式,通常经滚动体与滚道表面的交变应力作用而在表面以下薄弱点处形成裂纹核心,随后裂纹扩展至材料表面而形成点蚀和剥落。及时获取滚动轴承性能状态直接关系到企业能否合理制定设备维护、维修、备品、备件计划,以应付突发状况,避免经济损失、重大事故甚至人员伤亡等情况的发生,并可合理利用滚动轴承的使用寿命,充分挖掘其工作潜力而避免浪费。因此,掌握滚动轴承的疲劳演化进程将有助于提高企业经济效益,也有助于对滚动轴承损伤机理的科学研究。此外,准确的疲劳演化数据是滚动轴承性能退化评估、寿命预测等理论的重要数据来源,同时也是其计算结果的精度保障。但滚动轴承的寿命离散度大,采用基于概率理论、断裂力学与损伤建模等手段均不能较好的获取滚动轴承个体疲劳信息。而经典的基于振动等状态监测与故障诊断技术仅能获取表面已损伤的状态,对早期疲劳探测有一定局限性。材料受力变形或裂纹扩展将产生声发射信号(Acoustic Emission,AE)。二十世纪五十年代初期,由德国学者Kaiser开拓性的研究工作推动了现代声发射技术的诞生与发展,并逐渐成为获取滚动轴承疲劳早期损伤信息的有力工具。声发射技术可高效、可靠的检测材料表面下的活性缺陷。自然失效的轴承AE数据虽然可较为准确的反应其疲劳演化进程,但时间成本高,而现有的疲劳试验方案传递路径复杂,信号衰减大。且AE信号中存在难以避免的噪声,常规手段难以应对,尤其损伤初期,信号微弱而易被湮没。现有的获取滚动轴承疲劳进程的方法多采用3~5个传统AE指标的趋势分析,存在采用人工确定固定门槛电压而易引入主观干扰等诸多不足,且采用少数特征不足以全面反映滚动轴承疲劳状态,特征与损伤关系的研究还不够深入。而采用多数特征又会增加计算负担,特征间的冗余与不相关会干扰对疲劳进程信息的获取。加之各特征敏感性不同,演化信息不均匀地分散于各特征中,需要一定的专业知识与从业经验。逐一分析不仅耗时,而且工作量大,这将对演化进程的判定造成不便。因此,准确有效地提取滚动轴承疲劳演化特征的方法将成为解决上述问题的关键。本学位论文在国家自然科学基金项目"基于声发射和数值模型的球轴承点蚀疲劳寿命估计研究"(项目编号:51465022),国家自然科学基金"水下运动目标时变噪声场欠定盲提取模型及其算法研究"(项目编号:51265018)等的资助下,以声发射技术为基础,围绕滚动轴承自无损伤至剥落失效的全寿命阶段中疲劳演化特征信息的有效提取问题,采用理论研究、试验验证相结合的研究路线,初步建立了基于声发射的滚动轴承接触疲劳演化信息的提取方法框架,主要包括如下内容:(1)结合滚动轴承状态监测理论和工程需求,回顾了滚动轴承疲劳的相关研究方法、声发射理论和声发射监测技术、降噪技术、特征评估及特征提取技术的国内外研究现状,并对其优缺点进行了评述,确定了本文研究内容与技术路线。(2)分析了传统声发射监测指标存在的不足,给出了一种基于浮动阈值与平均化的AE监测指标,开展了多指标趋势分析。分析表明,改进的AE指标能更好的给出疲劳演化信息,随后筛选出了多个对球轴承疲劳较为敏感的指标,并对球轴承的一般演化规律进行了归纳与总结。设计了一种新型滚动轴承疲劳试验台,为后续的理论研究提供了试验验证平台。(3)针对采集的声发射信号中易混入噪声的问题,深入研究了声发射信号中的噪声来源与特性并给出了常规处理办法。对常规处理难以解决的噪声,从声发射信号中噪声与引入的噪声在自相关形式上的相似性角度,提出了一种基于二次相关理论的加权阈值小波包降噪算法,并进行了仿真和实测推力球轴承声发射信号的试验研究。研究表明,该方法能够较好的抑制采集声发射信号的噪声,信噪比提升稳定,降噪后的声发射信号冲击明显。(4)分析了现有距离评估方法的不足。针对多参数中可能存在的不相关与冗余性而干扰滚动轴承损伤程度信息准确识别的问题,以传统距离评估方法为理论基础,从算法对提升目标识别的准确性与稳定性角度,提出了一种基于位置补偿系数的距离评估方法。试验研究表明,通过该算法筛选出的敏感特征集具有较高的损伤识别精度,相对其他算法,该算法的稳定性更好,分类正确率更高。(5)在综合上述算法的基础上,以最少特征成分涵盖最多有效信息为目标,提出了一种基于改进粒子群优化核熵成分分析的疲劳演化信息提取算法。随后,根据熵的计算法则,又提出了一种二次特征融合算法,将已融合特征进一步融合,以更为高效的提取滚动轴承演化信息。经实测AE数据分析表明,改进的核熵成分分析的主要核熵得分数据能有效识别滚动轴承疲劳演化阶段,二次融合特征提取算法可极大汇聚来自各维特征的疲劳演化信息,仅采用单一的二次融合特征即可便捷、有效地表征滚动轴承的疲劳进程。
[Abstract]:Rolling bearing is one of the key components in the industrial field. Peeling is an important failure form of rolling bearing. It usually forms a crack core at the weak point below the surface by alternating stress between the roller and the raceway surface, and then the crack expands to the material surface to form pitting corrosion and peeling. Whether the enterprise can reasonably formulate equipment maintenance, maintenance, spare parts, spare parts plan to cope with unexpected situations, avoid economic losses, major accidents and even casualties, etc., and can reasonably use the service life of rolling bearings, fully tap their working potential and avoid waste. Therefore, mastering the fatigue evolution process of rolling bearings will have In addition, the accurate fatigue evolution data is an important data source for the theory of performance degradation assessment and life prediction of rolling bearings, and it is also an accurate guarantee for the calculation results. The classical state monitoring and fault diagnosis technology based on vibration can only obtain the damaged state of the surface, which has certain limitations for early fatigue detection. In the early 1950s, the pioneering work of Kaiser, a German scholar, promoted the birth and development of modern acoustic emission technology, and gradually became a powerful tool to obtain early fatigue damage information of rolling bearings. Although AE data of bearing can accurately reflect its fatigue evolution process, the time cost is high, and the existing fatigue test schemes have complex transmission paths and large signal attenuation. The process method usually adopts trend analysis of 3-5 traditional AE indexes, which has many shortcomings, such as manual determination of fixed threshold voltage and easy introduction of subjective disturbance. A few features are not enough to fully reflect the fatigue state of rolling bearings, and the research on the relationship between characteristics and damage is not deep enough. The redundancy and irrelevance between features will interfere with the acquisition of fatigue process information. Moreover, different characteristics have different sensitivities. Evolutionary information is unevenly distributed among the features, requiring a certain amount of professional knowledge and experience. Effective extraction of fatigue evolution characteristics of rolling bearings is the key to solve the above problems. This dissertation is based on the National Natural Science Foundation project "Research on pitting fatigue life estimation of ball bearings based on acoustic emission and numerical model" (item number: 51465022), and the National Natural Science Foundation of China (NSFC), "Undetermined time-varying noise field of underwater moving targets" Blind Extraction Model and Its Algorithms Research (Project Number: 51265018) and other funding, based on acoustic emission technology, around the problem of effective extraction of fatigue evolution characteristic information in the whole life stage of rolling bearings from non-destructive to peeling failure, a research route combining theoretical research with experimental verification was established preliminarily based on acoustic emission technology. The extraction method framework of rolling bearing contact fatigue evolution information mainly includes the following contents: (1) Combining with rolling bearing condition monitoring theory and engineering requirements, this paper reviews the related research methods of rolling bearing fatigue, acoustic emission theory and acoustic emission monitoring technology, noise reduction technology, feature evaluation and feature extraction technology at home and abroad. (2) The shortcomings of traditional AE monitoring indexes are analyzed, and an AE monitoring index based on floating threshold and averaging is given. The trend analysis of multiple indexes is carried out. The analysis shows that the improved AE index can give better fatigue evolution information. A new type of rolling bearing fatigue test bench is designed, which provides a test and verification platform for subsequent theoretical research. (3) Aiming at the problem of easy mixing noise in the acoustic emission signals collected, in-depth study is carried out. The noise sources and characteristics of AE signals are analyzed and the conventional processing methods are given. For the noise which is difficult to be solved by conventional processing, a weighted threshold wavelet packet denoising algorithm based on quadratic correlation theory is proposed from the angle of the similarity of the autocorrelation form between the noise in AE signals and the introduced noise, and the simulation and implementation are carried out. Experimental study on acoustic emission signals of thrust ball bearings shows that the method can suppress the noise of acoustic emission signals, improve the signal-to-noise ratio and stabilize the acoustic emission signals. The impact of acoustic emission signals after noise reduction is obvious. (4) The shortcomings of existing distance evaluation methods are analyzed. On the basis of traditional distance evaluation method, a distance evaluation method based on position compensation coefficient is proposed to improve the accuracy and stability of target recognition. Experimental results show that the sensitive feature set selected by this method has higher damage identification. Fifthly, based on the above algorithm, a fatigue evolution information extraction algorithm based on Improved Particle Swarm Optimization (PSO) kernel entropy component analysis is proposed, which covers the least feature components and the most effective information. A quadratic feature fusion algorithm is proposed to fuse the fused features to extract the evolution information of rolling bearings more efficiently. The results of AE data analysis show that the improved kernel entropy analysis can effectively identify the fatigue evolution stage of rolling bearings and the quadratic fusion feature extraction algorithm can converge greatly. The fatigue evolution information from each dimension feature is gathered, and the fatigue process of rolling bearings can be easily and effectively characterized by a single secondary fusion feature.
【学位授予单位】：昆明理工大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TH133.33

【相似文献】