【作者】 李林升；

【导师】 曾理；

【作者基本信息】 重庆大学 ， 仪器科学与技术， 2011， 博士

【摘要】 随着航空航天、铁路运输、公路运输等动力机械的速度、承载能力的提升,安全事故的破坏程度也呈上升趋势,其中应力及疲劳寿命导致的安全事故占很大一部分,当前迫切需要了解工件破坏的应力及疲劳寿命状况。工业CT技术可获得工件内部断面或三维图像,为内部缺陷检测和工件应力分析及疲劳寿命预测提供基础。本论文以工业CT (Computed Tomography,CT)为研究平台,以CT图像边缘提取获得的数据为基础,研究了扩展有限元应力分析、疲劳寿命预测,这对提高工件的安全性能具有重要指导意义。本论文的贡献与创新主要表现在以下几个方面:1、研究了维图像边缘提取的稳健(Robust)统计尺度区域拟合模型。图像中灰度不均的目标与背景以及局部细节特征很难提取。为了解决这个问题,前人提出了区域尺度拟合模型,该模型利用了可控尺度局部区域的灰度信息,但是,仅仅使用灰度信息可能导致低的收敛率和较弱的降噪能力。本论文充分利用稳健统计方法抑制图像噪声的优点以及尺度区域拟合模型提取图像局部信息的优点,在尺度区域拟合模型输入信息中添加局部四分位距与局部绝对离差信息,用于锐化输入图像目标边缘,以提高活动轮廓的定位精度和收敛速度;添加局部中值信息用于去除图像噪声。与尺度区域拟合模型相比,改进的尺度区域拟合模型收敛速度更快,抗噪声能力更强。2、研究了三维图像边缘提取的局部稳健统计C-V模型。由于C-V模型能够获得封闭而精确的目标轮廓,该模型已经广泛地应用于图像边缘提取并推广到三维。然而,伪轮廓和噪声的存在会导致活动轮廓停靠在不理想的边界。为了克服上述困难,通过引进稳健统计方法改进三维C-V模型。在改进模型中,把局部四分位距,局部绝对离差和局部灰度中值信息代替三维C-V模型的图像灰度信息,局部四分位距与局部绝对离差信息用来锐化图像目标轮廓,局部灰度中值信息用来抑制噪声。与三维C-V模型相比,改进的三维C-V模型图像边缘提取精度更高,抗噪声能力更强。3、研究了基于CT图像边缘提取的深埋缺陷应力分析方法。普通有限元软件ANSIS对于规则缺陷能够作出其矢量图形,进而分析其应力,但对于不规则缺陷,由于很难构造其矢量图形,因而不易获得其应力分布。此外,为了提高缺陷区域应力分析的精度,必须在该区域划分高密度网格。为了解决上述问题,本论文设计了基于CT图像边缘提取的铸件缺陷应力分析的流程,并对流程的关键步骤—扩展有限元分析进行了改进:用C-V方法的水平集函数构成扩展有限元的富集函数,以分析任意形状缺陷的应力。对于规则缺陷,与有限元软件ANSIS相比,改进的扩展有限元方法在应力分析精度大致相当的情况下需要较少的网格。对于不规则缺陷,基于CT图像边缘提取的缺陷应力分析流程已运用于铁道货车侧架缺陷的应力分析。4、研究了基于CT图像边缘提取的在役工件深埋裂纹扩展寿命预测方法。表面裂纹的扩展寿命已经能够被预测,而对于在役工件的深埋裂纹,由于对其检测的困难以及裂纹扩展后形状变化的复杂性,很难预测其扩展寿命。为此,本论文设计了基于CT图像边缘提取的在役工件深埋裂纹扩展寿命预测流程:在役工件经工业CT扫描,得到含深埋裂纹的CT图像,采用Canny算法提取其裂纹边缘轮廓。然后对长宽比比较大的裂纹,研究了逐步缩小椭圆短轴的方法以实现裂纹边缘椭圆拟合,最后把逐步缩小椭圆短轴的椭圆拟合算法与Paris疲劳裂纹扩展寿命预测算法有机结合,跟踪裂纹扩展后的形状变化,预测深埋裂纹扩展寿命。5、研究了基于CT图像边缘提取的在役车轮疲劳寿命预测方法。弯曲疲劳寿命是车轮重要的性能指标,目前其预测方法主要适用于车轮的设计模型。为此,本论文设计了基于CT图像边缘提取的在役车轮疲劳寿命预测流程:首先用工业CT对在役车轮进行扫描,得到其切片图像;在运用高斯滤波函数对切片图像预处理后,采用移动立方体方法提取三维表面;接着运用顶点删除法和Laplacian算法简化和平滑三维表面;将三维表面模型导入UG进行处理得到三维CAD模型;最后借助于有限元方法,通过模拟在役车轮工况,建立边界条件,得到在役车轮的应力分布,并运用名义应力法预测在役车轮的疲劳寿命。更多还原

【Abstract】 At present, with the improvement of speed and load of power machine such as aerospace, railway transport and highway transport, the extent of damage of safety accidents increasingly grows, where the safety accidents resulting from stress and fatigue life take a great part. Therefore, it is necessary to understand the stress and fatigue life of workpiece. Industrial computer tomography (CT) can get the image of workpiece section or three-dimension image of workpiece, which help for defect detection, stress analysis and fatigue life prediction. In this dissertation, we take industrial CT as research platform and CT image edges extraction as basis of research. A very close research on stress analysis with extended finite element method(XFEM), fatigue life prediction is made, which has significant meaning to improve the safety performance of workpiece.The contribution and novelty of this dissertation are as following:1. The two-dimensional image edges extraction method based on the local robust statistics and region-scalable fitting (RSF) model is studied. Intensity inhomogeneity often takes place in CT images, which may cause considerable difficulties in CT image edges extraction. In order to overcome the difficulties caused by intensity inhomogeneity, RSF model was put forward. This model draws upon intensity information in local regions at a controllable scale. But only using intensity information may lead to slow convergence speed and poor ability of noise reduction. In this dissertation, the intensity information of image in RSF model is replaced with local robust statistics which is the weighted combination of local inter-quartile range, local mean absolute deviation and local intensity median. Specifically, local inter-quartile range and local mean absolute deviation are introduced to sharpen target edges, and local intensity median is introduced to reduce image noise. Compared with the RSF model, the improved RSF model demonstrates the fast convergence rate and robustness to noise.2. The three-dimensional image edges extraction method based on the local robust statistics and three-dimensional C-V model is studied. Since the C-V model is capable of getting closed and accurate target contours, it has been widely applied to image edges extraction and extended to three-dimensional model. However, the existence of forged boundaries and noise may make active contours stop at the undesired boundaries. In order to overcome the difficulties caused by those effects, the three-dimensional C-V model is improved with robust statistics method. In this improved model, the intensity information of image in three-dimensional C-V model is replaced with local robust statistics information which is the weighted combination of local inter-quartile range, local mean absolute deviation and local intensity median. Here, local inter-quartile range and local mean absolute deviation are introduced to sharpen target edges, local intensity median to reduce image noise. Compared with the three-dimensional C-V model, the improved three-dimensional C-V model demonstrates high precision of image edges extraction and robustness to noise.3. A method of stress analysis of the embedded defect based on CT image edges extraction is studied. The software ANSIS of finite element method can draw vector graph of regular defect and analyze its stress. However, the stress distribution of irregular defect is not easy to be got by this software because the vector graph of the irregular defect is drawn with difficulty. In addition, in order to improve the accuracy of stress analysis of defect region, high dense meshes in this region are needed. In order to solve these problems, we design a flow for stress analysis of the casting defect based on CT image edges extraction. To analyze the stress of defect with arbitrary shape, we improve the XFEM used in stress analysis by using the level set function(LSF) in Chan-Vese(C-V) method to constitute enrichment function of XFEM. For regular defect, compared with the software ANSIS of finite element method, the flow for stress analysis of the casting defect in this dissertation can analyze embedded defect with arbitrary shape. The result of stress analysis with XFEM is as accurate as that with ANSIS, while high dense meshes in discontinuous region are not needed in XFEM. For irregular defect, this flow for stress analysis of the casting defect based on CT image edges extraction has been applied in analyzing the stress of defect in the side frame of railway freight car.4. A method for forecasting embedded crack propagation lifetime of work piece in service based on CT image edges extraction is studied. The propagation lifetime of surface cracks can be forecasted by the Paris method. Due to the difficulty of crack detection and the complexity of crack propagation, it is difficult to forecast the propagation lifetime of embedded cracks of work piece in service. In this dissertation , we present a flow for forecasting crack propagation based on CT image edges extraction, whereby the work piece containing the embedded crack is firstly scanned by industrial CT to obtain its images, extracting crack edges with the Canny operator. Secondly, in order to fit crack boundaries that have a large length-to-width ratio with an ellipse, a method of gradually shortening the length of ellipse’s minor axis is developed. Finally, in order to improve predictions of crack lifetime durations, we combine this method with the formula of planar crack propagation to track its shape change.5. A method of fatigue life prediction of wheel in service based on CT image edges extraction is studied. Bending fatigue life is the most important performance of the wheel. At present, the fatigue life prediction is mainly applied to design model of the wheel while is not applied to the wheel in service. Here, we design a flow for fatigue life prediction of wheel in service based on CT image edges extraction. Firstly, the wheel is scanned to get a sequence of industrial CT images. After the CT images are processed by Gaussian filter, the 3-D surface model is extracted from the CT images by using Marching Cubes (MC) algorithm, and then simplified and smoothed by Vertex Removing(VR) algorithm and Laplacian algorithm respectively. Then the 3-D surface model is processed by the software UG. Finally, based on finite element method, the stress distribution of the wheel is acquired by simulating its work condition and building up a boundary condition. Moreover, the fatigue life of the wheel is predicted with nominal stress approach. Through this flow, the 3-D CAD model of the wheel can be rebuilt quickly and accurately according to its industrial CT data. The dangerous parts of the wheel can be located and the fatigue life of the wheel is realized.更多还原