基于声发射信号分析的2D及3D纺织结构复合材料损伤机制研究

发布时间：2018-01-17 18:21

本文关键词：基于声发射信号分析的2D及3D纺织结构复合材料损伤机制研究　出处：《东华大学》2015年博士论文　论文类型：学位论文

【摘要】：随着纤维增强复合材料更加广泛地应用到航空航天、汽车和建筑工程等领域，损伤破坏等安全性问题是其生产和使用的前提和关键。声发射技术是一种实时无损检测技术，比其它无损检测方法能更早地预报出复合材料及结构的损伤。由于复合材料在加载过程中可产生大量声发射信号，且声发射信号包含了关于损伤源许多有用的信息，对这些信息进行解析可为我们了解和掌握复合材料损伤破坏机理提供有效的依据。复合材料声发射技术研究的目的是识别声发射信号包含的损伤源信息，所以对大量声发射信号的分类以及与损伤机制之间建立联系成为研究的重点，也是难点。近年来，聚类分析是一种被广泛接受的用于区别大量声发射信号的一种无监督的模式识别技术。声发射技术用于纤维增强复合材料的损伤研究已有不少报道，但主要是针对单向和正交热固性复合材料，而在2D及3D纺织复合材料等复杂增强体热固性复合材料以及热塑性复合材料方面的研究和结果比较少。本论文提出了表征声发射信号的Laplacian分值结合相关分析及主成分分析的特征选择方法，以复杂结构纤维增强复合材料拉伸过程产生的声发射信号为基础，对六种聚类算法进行了对比讨论；采用Silhouette值及Davies-Bouldin指标进行聚类有效性判定，并基于Matlab软件设计开发了简单实用的聚类分析工具，便于对声发射信号数据进行聚类分析和处理。本论文采用声发射信号聚类分析技术对热固性复杂结构和热塑性复合材料进行了全面的研究。首先探讨了2D及3D玻璃纤维和碳纤维机织增强热固性复合材料的拉伸声发射信号特征与聚类分析，其次对PE/PE梯度热塑性复合材料的力学性能和声发射信号特征进行了分析，然后分别采用基于Weibull理论的纤维束模型、数字高速摄影技术裂纹计数法、扫描电子显微镜技术等，建立了不同聚类与损伤机制之间的联系，并结合损伤阈值方法讨论了纤维增强复合材料拉伸过程中损伤的产生与发展过程，最后采用小波分析方法讨论了热固性和热塑性复合材料声发射特征的异同。论文研究的详细内容如下： 1.考虑到聚类分析过程中特征选择的重要性，针对声发射信号研究了特征选择方法。提出了采用Laplacian分值和相关分析的特征选择方法，以消除特征冗余，选择出具有较高分类能力及表征声发射信号的特征参数。在进行聚类分析前，对选取的九个声发射信号进行特征选择，选择出四个特征：峰值振幅、峰值频率、RA值及质心频率为具有分类能力的特征，根据主成分分析法可以找出有效表征聚类声发射信号的参数，便于对聚类结果的分析和讨论。 2.对比研究了k-means、 k-means++、 fuzzy c-means (FCM)、SOM+k-means、SOM+FCM和自适应仿射传播AAP算法用于聚类复合材料AE信号数据的优缺点。k-means++和AAP算法的聚类效果较好，但是AAP算法的计算时间较长，计算成本较大，且必须保证选取的AE信号特征参数使信号之间分离度较好时才易于收敛，否则将陷入不断迭代过程，而k-means++算法对于初始聚类中心优化选择，易得到好的聚类结果，且算法简便、计算成本低、计算速度快，更适用于复合材料AE信号数据的聚类分析。基于MatlabR2013a软件设计开发了聚类分析工具，，可对不同材料加载产生的未知声发射信号进行全面的分析处理。 3.研究了2D及3D玻璃纤维机织热固性复合材料的声发射信号特征。采用聚类分析技术对热固性2D平纹及3D正交玻璃纤维增强环氧树脂复合材料的拉伸过程产生的声发射信号数据进行了分析和处理。特征选择方法选择出的四个声发射特征可将AE事件可分为四类。峰值振幅与峰值频率是可以表征AE信号的最重要的声发射特征参数，所有试样的聚类分布形状和聚类边界范围相似。 4.研究了2D及3D碳纤维机织热固性复合材料的声发射信号特征。对声发射信号进行了聚类分析并与相同结构的玻璃纤维/环氧树脂复合材料的聚类结果进行了对比与讨论，结果表明，相同结构不同材料的声发射信号的聚类分布及聚类参数范围相似，而碳纤维的聚类边界值比玻璃纤维高。2D及3D复合材料不同加载方向产生的声发射信号可归纳为三类：低振幅低频率类，高振幅低频率类，以及宽振幅高频率类。 5.研究了热固性复合材料声发射信号聚类结果与损伤机制之间的联系。基于Weibull理论纤维束模型模拟了纤维断裂数目，考察了高频聚类中的声发射信号数目的对比，结果显示出一定程度上的一致性，为高频声发射事件与纤维断裂相关提供了额外的依据。根据2D玻璃纤维/环氧树脂复合材料在透射光下的特点，采用数字高速摄影技术对可见裂纹进行了计数，将不同聚类中的声发射事件与原位观察到的裂纹数目和发展趋势进行了对比，并建立了机织热固性复合材料声发射信号聚类与典型损伤机制之间的联系。 6.研究了热塑性纤维增强复合材料的力学性能，并采用声发射聚类分析技术和损伤阈值方法探讨了拉伸损伤机制。首先，分别制备和研究了单向及正交梯度PE/PE复合材料的力学性能，并分析了正交梯度和非梯度材料的声发射信号能量分布及特征。其次，对PE/PE正交梯度复合材料的拉伸声发射信号进行了分析和聚类，不同试样的声发射信号分为5类时最佳，峰值振幅与质心频率作为主要模式特征对聚类结果进行了讨论。然后，结合文献中模型试样的AE特征及频谱分析结果，建立了5个聚类与5种典型损伤机制之间的联系；并采用小波分析法比较了热固性和热塑性复合材料声发射信号特征，热固性复合材料AE信号的相应频率成分的持续时间比热塑性复合材料长，衰减较慢，且热塑性复合材料的低频AE信号的频率分布较为集中。
[Abstract]:In recent years , cluster analysis is a kind of non - supervised pattern recognition technology which is widely accepted to distinguish a large number of acoustic emission signals . Based on the analysis of the Laplacian value of the acoustic emission signal and the feature selection method of the principal component analysis , this paper makes a comparative discussion on the six kinds of clustering algorithms based on the acoustic emission signal generated by the stretching process of the fiber reinforced composite material . 1 . Considering the importance of feature selection in cluster analysis , the feature selection method is studied for the acoustic emission signal . The feature selection method of Laplacian value and correlation analysis is proposed to eliminate the feature redundancy . Four characteristics are selected : the peak amplitude , the peak frequency , the RA value and the centroid frequency are the characteristics of the classification ability . According to the principal component analysis method , we can find out the parameters which can effectively characterize the acoustic emission signal and facilitate the analysis and discussion of clustering results . 2 . Comparing the advantages and disadvantages of k - means , k - means ++ , fuzzy c - means ( FCM ) , SOM + k - means , SOM + FCM and adaptive affine propagation , the clustering results of AE signal data of cluster composite materials are better . 3 . The acoustic emission signal characteristics of 2D and 3D glass fiber woven thermosetting composites were studied . The acoustic emission signal data generated by the stretching process of the thermosetting 2D plain film and the 3D orthogonal glass fiber reinforced epoxy resin composite material were analyzed and processed by cluster analysis technique . The four acoustic emission characteristics selected by the feature selection method can be divided into four categories . The peak amplitude and the peak frequency are the most important acoustic emission characteristic parameters that can characterize the AE signal , and the cluster distribution shapes and cluster boundary ranges of all samples are similar . 4 . The acoustic emission signal characteristics of 2D and 3D carbon fiber woven thermosetting composites were studied . The results show that the cluster distribution of the acoustic emission signals and the cluster parameter ranges are similar to that of the glass fiber / epoxy resin composites with the same structure . The results show that the acoustic emission signals generated by different loading directions of the 2D and 3D composites are classified into three types : low - amplitude low - frequency class , high - amplitude low - frequency class , and wide - amplitude high - frequency class . 5 . The relation between the clustering result and the damage mechanism of the acoustic emission signals of thermosetting composites is studied . Based on the Weibull theory , the number of fiber breaks is simulated , the number of acoustic emission signals in the high frequency cluster is compared , the results show that some degree of consistency is provided . According to the characteristics of the 2D glass fiber / epoxy resin composite material under transmitted light , the visible cracks are counted by the digital high - speed photography technology . The relationship between the acoustic emission events in different clusters and the number of cracks observed in situ is compared , and the relationship between the acoustic emission signal cluster and the typical damage mechanism of the woven thermosetting composite material is established . 6 . The mechanical properties of thermoplastic fiber reinforced composites were studied . The mechanical properties of unidirectional and orthogonal gradient PE / PE composites were investigated by acoustic emission cluster analysis technique and damage threshold method .

【学位授予单位】：东华大学
【学位级别】：博士
【学位授予年份】：2015
【分类号】：TQ327;TB33

【参考文献】