基于改进的支持向量回归机算法的磁记忆定量化缺陷反演

针对焊缝缺陷磁记忆检测中存在定量化反演难题,建立了基于改进的支持向量回归机定量反演模型.以预制不同尺寸未焊透和夹渣缺陷的Q235焊接试样为试验材料,进行磁记忆扫描检测发现:缺陷位置的磁记忆信号特征参数随尺寸变化而呈现一定的变化规律,但同时存在分散性和不确定性.鉴于磁记忆信号样本的有限性、分散性和非线性,首先将提取到的磁记忆特征参数进行归一化处理,引入支持向量回归机建立焊缝缺陷磁记忆定量反演模型,并进一步利用模拟退火算法对支持向量回归机参数进行优化,使目标函数达到全局最优而非局部最优.最后,考虑到由磁记忆信号逆向反推缺陷的三维尺寸,存在解的不确定性,为此在缺陷单维尺寸反演模型的基础上,通过构建多层结构的支持向量回归机进行多尺寸反演输出,建立了基于模拟退火支持向量回归机的焊缝缺陷磁记忆定量反演模型,结果表明:未焊透缺陷尺寸反演最大相对误差为7.96%,夹渣缺陷为4.97%,为焊缝缺陷的磁记忆反演与定量化评价提供一种新的思路.     

基于粒子群最大似然估计的焊缝早期隐性损伤磁记忆精确定位模型

针对由于焊接残余应力、磁场噪声等干扰,造成磁记忆检测在焊缝早期隐性损伤位置定量评价上的困难,提出基于粒子群算法优化的最大似然估计磁记忆梯度定量模型.通过对预制未焊透缺陷的Q235焊接试件进行焊缝疲劳拉伸实验,同步对比扫描电镜和X射线检测结果,发现磁记忆信号梯度对早期隐性损伤位置反应比较敏感,并获得了梯度随着与隐性损伤的距离增大而减小的衰减变化规律,构建隐性损伤位置参数与磁记忆梯度的非线性函数,考虑磁场噪声对隐性损伤定位结果的影响,引入最大似然估计建立目标函数,进一步考虑目标函数的非线性容易陷入局部极值而非全局极值的问题,采用具有全局搜索能力的粒子群算法对目标函数进行优化,建立基于粒子群最大似然估计的焊缝隐性损伤位置磁记忆定量模型,验证结果表明定位误差仅为3.48%,为实际工程中利用磁记忆技术及时发现早期隐性损伤并精确定位提供了新的思路.     

基于小波神经网络的管道腐蚀缺陷定量识别研究

漏磁检测是目前广泛采用的油气管道检测方法,通过漏磁检测仪获得腐蚀缺陷漏磁场切向分量的漏磁场信号,提取与腐蚀缺陷外形长、宽、深有关的漏磁场信号波形特征量,结合小波分析和神经网络的优势形成小波神经网络,分析设计了小波神经网络的结卡句,给出了网络训练算法,利用网络的非线性逼近性能,对腐蚀缺陷外形进行定量评价,给出预测评价曲线,试验验证方法有效可行。     

基于特征量和神经网络的钢管缺陷预测模型

分析了钢管缺陷几何大小与缺陷漏磁信号(MFL)特征量之间关系，建立了一组全方位的钢管缺陷信号特征量，并将人工神经网络理论和算法应用于钢管缺陷预测。通过实验取得样本，在对网络进行训练的基础上，建立了基于钢管缺陷漏磁信号特征量和神经网络的缺陷预测模型，继而根据漏磁信号对缺陷进行定量预测。给出了实验结果，结果表明采用这种方法能够较好地实现管道缺陷的定量识别。     

油气输送管道腐蚀缺陷检测与识别研究

应用漏磁检测仪进行埋地油气管道腐蚀检测和识别,叙述了检测仪的硬件采集和信号处理过程.在直径273 mm的管道表面上加工人工缺陷,采集缺陷漏磁场信号,进行处理和转化为漏磁场云图,从信号和云图两者来分析漏磁场,提出了缺陷漏磁场的干扰消除方法、影响因素补偿技术和缺陷漏磁场图像的截取,并分别获得较好的效果,分析了缺陷定性和定量识别的方法,研究了缺陷外形参数量化的方法,得出了缺陷外形量化的数学模型,试验验证缺陷识别的效果,证明整个过程和方法有效可行,大大提高了管道缺陷检测和量化的精度.     

漏磁法检测薄带钢内部缺陷的研究进展

漏磁法检测带钢内部缺陷是利用铁磁性材料在磁场中的导磁现象,若材料中有缺陷存在,就会在材料表面缺陷处形成漏磁场,通过漏磁场大小即可检测出缺陷大小。文章中给出了基于漏磁技术的薄带钢内部缺陷检测试验结果。在试验中,采用厚度为0.13mm的薄带钢为检测对象,并在其上打出直径为0.13mm的通孔和盲孔作为人工缺陷,将钢板磁化至深度饱和,能够较好地得到该缺陷的漏磁信号。目前,日本太平工业和Nireco两家公司开发出的漏磁检测设备已经进入工业化大生产阶段,二者的缺陷检测精度分别为0.0014mm3和0.000373mm3。分析认为可以根据现场检测条件和检测精度需求选择合适的磁敏传感器,并采用直流磁化方法,将带钢磁化至深度饱和来保证微小缺陷信号的检出。采集到漏磁信号后,用小波分析法进行信号预处理,然后以统计法或者神经元网络法来实现缺陷尺寸量化,往往能够得到较为精准的检测结果。对漏磁技术应用中的难点和技术发展趋势进行了展望。     

疲劳载荷下金属材料的磁记忆信号特征

设计拉伸疲劳试验,研究了疲劳载荷下磁信号暂时到达的稳定状态的特点以及疲劳过程中磁信号的分布、变化情况,并结合改进的J-A理论进行分析和讨论.结果显示:磁信号暂时到达的稳定状态在一个循环周期内沿环线变化,并且环线随循环载荷类型不同而变化,说明用改进的J-A理论描述磁记忆现象合理;疲劳过程中磁信号分三个阶段变化,这个特征由改进的J-A模型中描述磁畴壁运动所受阻碍的参数(ξ″)随疲劳损伤发展的变化规律解释;磁信号在缺陷附近呈\     

改进粒子群优化神经网络及其在产品质量建模中的应用

针对传统神经网络优化算法易陷入局部最优值的问题,在标准粒子群算法的基础上,对粒子速度与位置更新策略进行改进,提出一种基于改进粒子群优化算法的BP神经网络建模方法.使用sinc函数、波士顿住房数据及某钢厂带钢热镀锌生产的实际数据进行验证.结果表明,与标准的反向传播神经网络和支持向量机相比,基于改进粒子群优化的神经网络模型可以有效提高预测精度.     

基于高相关滤波算法的PSO-LSTM连铸坯质量预测模型

文章提出了一种将高相关滤波算法与粒子群优化长短期记忆神经网络相结合的连铸坯质量预测模型.首先,使用高相关滤波算法对影响连铸坯质量的多种因素进行降维处理,然后,利用粒子群优化长短期记忆神经网络进行连铸坯中心偏析陷等级的预测实验.实验结果验证了该模型在预测性能方面的优势.     

基于循环神经网络的板形模式识别模型

板形模式识别是板形控制关键一环。传统的板形模式识别模型存在识别精度差、抗干扰能力差等缺点。随着数据回归任务复杂程度的增加,以深度学习为基础的深度神经网络已经大量用于数据分类、图像处理、模式识别、特征提取等任务。深度学习可通过学习一种深层非线性网络结构,实现复杂函数逼近。基于这一背景提出了基于循环神经网络RNN的板形模式识别模型。结果表明,基于RNN的板形模式识别模型可以完成板形的大数据训练,模型的识别精度和泛化能力都很高,为进一步提高板形控制精度提供了新方法。     

Failure of large-diameter steel pipe with rolling scabs

Russian pipelines employ large-diameter pipe of straight-seam, two-seam, and spiral-seam type (diameter up to 1420 mm, API strength class up to K65). The newest developments in the production of large-diameter (1020, 1220, and 1420 mm) straight-seam welded pipe (strength classes K38–K65 and X42–X80, wall thickness up to 52 mm, length up to 18 mm, and working pressure up to 22.15 MPa) is stepwise press shaping (the JCOE process), proposed by SMS Meer (Germany). The SMS Meer technology is widely used at Russian pipe plants (AO Vyksunskii Metallurgicheskii Zavod, AO Izhorskii Trubnyi Zavod, PAO Chelyabinskii Truboprokatnyi Zavod) and also plants in Russia, China, and India. However, the accident statistics for Russian pipelines show that stress corrosion of the pipe wall mainly occurs in pipelines of large diameter (700–1420 mm). More than 80% of pipeline failures associated with stress corrosion occur in pipelines of diameter 1020–1420 mm. Corrosion cracking of pipe walls may be attributed to three main factors: (1) poor steel quality and pipe defects in manufacturing (such as high residual stress, microcracks and micropeeling of the metal after shaping of the pipe blank, corrugation, scratches, scabs from the rolling process, and imperfections of the weld seams); (2) the presence of a corrosive medium and its access to the metal surface; (3) multicyclic fatigue and failure of the metal on account of pulsation of the working pressure within the pipe and hydraulic shocks. In Russian oil pipelines, failures due to production defects and assembly and installation problems are twice as frequent as in the United States and Europe. Therefore, careful study of pipeline failure due to production flaws is of great importance. In the present work, a mathematical approach is proposed to determining the critical pressure in the pipe at which elastoplastic failure of the pipe will occur at rolling scabs accompanied by a scratch on the pipe’s outer surface. The results may be used in failure diagnostics of large-and medium-diameter steel pipe for major delivery pipelines and transfer pipelines.     

深井矿山膏体充填系统增阻减磨试验研究

针对深井矿山自流充填过程中的管道磨损问题,通过对浆体管输流动模式的分析,明确不满流是引起管道磨损的主要原因。从动量角度出发,研究了管道磨损机理及其影响规律。基于水力学理论推导,提出采用小管径增加系统阻力损失,减轻管道磨损的措施。会泽铅锌矿增阻减磨工业试验,将原Ф150mm管道部分更换为Ф85mm管道,管内压力值增加了6倍,满管率由7．4％提高至47％。研究结果表明,采用小管径管道是一种有效的减小磨损的措施,对于类似矿山具有借鉴意义。     

基于KPCA-IPSO-LSSVM的充填管道磨损风险预测

为克服传统预测模型存在的适用性差、预测精度不足和参数选取随意性强等缺陷,提出了一种将核主成分分析法(KPCA)、改进粒子群算法(IPSO)与最小二乘支持向量机(LSSVM)相结合的充填管道磨损风险预测新方法.通过KPCA对管道磨损影响因素进行特征提取,将提取结果作为LSSVM的输入,同时利用具有较强全局搜索能力的IPS...     

Cyclic Tests of Stress-corrosion Cracks in Steel Gas Pipelines with No Corrosive Fluid

Stress-corrosion defects in the output pipeline of a compressor station are subjected to metallographic analysis and mechanical tests. The pipeline consists of pipe imported in the 1980s. Diagnostic data obtained in nondestructive monitoring of such pipelines are analyzed, and the effectiveness of various diagnostic methods in detecting stress-corrosion defects in large-diameter pipe is assessed. The grade of steel in the pipeline is identified. The cracks are identified by type and morphology, and their development is determined. Nonmetallic sulfide inclusions have no influence on pipeline failure in the stress-corrosion conditions considered. The sulfur content in the corrosion products is no more than the sulfur content in the metallic sample. In some samples, the sulfur content may be decreased except for local sections with nonmetallic sulfide inclusions. Electron-microscope images show that such nonmetallic inclusions do not promote failure. The results of cyclic tests of cracked samples cut from the pipeline are presented. The test conditions are selected in accordance with recent operating conditions of the compressor station. According to the results, pipe with defects at the initial stage of development exhibits considerable durability in the tests. In fact, the cracked samples withstand (1.6–7.5) × 10⁶ cycles under transverse cyclic flexure in a single plane, in the absence of corrosive fluid. In practice, the number of such cycles is no more than 120–200 per year, in normal pipeline operation. Hence, a pipeline with stress-corrosion defects in their initial stage of development will last a long time if its metal wall is protected from the action of corrosive fluid.     

Buried Pipelines Subject to Subgouge Deformations

Engineers often model pipe/soil interaction events based on the concept of subgrade reactions originally proposed by Winkler. Engineering models often utilize beam and nonlinear/plastic spring elements to represent pipelines and the surrounding soil medium, respectively. The spring formulations, defining soil resistance to deformations in three-dimensional space, are usually assumed to be independent and the responses are discrete between adjacent soil zones. However, this idealization does not truly replicate a soil medium behavior. This study presents coupled numerical analyses of pipeline for the specific problem of subgouge deformations due to ice gouge events. Three dimensional continuum analyses of coupled pipe/soil/ice keel interaction using an explicit arbitrary Lagrangian finite- element approach were performed. The study compares the continuum finite-element results with Winkler-type analysis for the specific analyzed problem. A Lagrangian adaptive meshing technique was employed to model very large movement and achieves a steady-state condition; and reasonable ice/soil and soil/pipe interaction interfaces are employed. The numerical analysis shows the potential for continuum modeling of pipe/soil interaction events and develops a better understanding of ice gouging and pipe/soil/ice keel interaction.     

Automatic girth welding and performance evaluation of the joints of hot-induction-bend and line pipes with different wall thickness

During the process of laying long-distance oil and gas transmission pipelines,the hot-induction-bend method is extensively used when the direction has to be changed.By considering the pipeline's ongoing processing and loading states during service,the pipeline that is generally used exhibits thicker walls than those that are observed in the line pipe.As such,during pipeline construction,hot-induction-bend and line pipes with different wall thickness are girth-welded.The chemical composition of hot-induction-bend and line pipes differs,with the carbon content being particularly higher in the hot-induction-bend pipe;it also depicts a higher carbon equivalent,which makes it possible to modify the girth of the pipe.In this study,using Baosteel's standard X70M UOE hotinduction-bend and line pipes,solid-wire automatic gas-metal-arc girth welding was performed and the performance of the girth-welded joint was evaluated.Furthermore,the weldability of the pipeline girth and the microstructure of the girth-welded joint were analyzed.The results reveal that Baosteel's standard UOE hot-induction-bend and line pipes exhibit good girth weldability,and their technical quality can be guaranteed in case of consumer fieldconstruction applications.