【作者】 李培麟；

【导师】 陆皓；

【作者基本信息】 上海交通大学 ， 材料加工工程， 2012， 博士

【摘要】 多丝埋弧焊有着较高的熔敷率与生产效率，逐渐成为管线钢焊接与压力容器焊接的主要焊接工艺，然而多丝埋弧焊工艺参数确定仍需通过进行反复的试验才能确定。目前多丝埋弧焊热源模型参数确定方法主要为试算，由于研究人员的经验及时间限制，容易引入人为误差，很难保证热源模型的精度，同时在另一方面又增加了开发成本。为此，急需设计一种更为合理的热源模型以及获得其参数的方案，以降低在试算过程中引入的人为误差。反演计算作为一种“由果及因”的算法，能够在输入参数与结果参数之间的映射关系尚不明确的情况下根据结果参数确定对应的最适合的输入参数。将反演算法引入至焊接热源模型参数的确定工作，可以提高数值模拟的精度，减少工艺试验，节约开发成本。然而反演计算的结果基于试验结果，无法获得未经试验的工艺参数条件对应的热源模型参数，从而导致计算结果离散化。对于未经试验的焊接工艺参数，采用智能优化算法能够对反演结果进行扩展，并使计算结果连续化，从而能够获得在一定范围内的任意工艺参数组合所对应的热源模型参数。经过反演算法与智能优化算法的共同计算，能够直接获得不同焊接工艺参数所对应的热源模型参数，将焊接模拟的试算工作量大大降低，并大幅提升了模拟的精度。本文采用这一方案，将多丝埋弧焊的焊缝成形参数预测误差控制在5%以内。为进行反演算法与智能优化算法对热源模型参数的预测，首先进行了焊接工艺试验。设计了一种全新的热物理性能测量的方法，通过模式搜索法获得了母材以及埋弧焊剂随温度变化的的热导率及热容等热物理性能。对多丝埋弧焊的焊缝成形测量方法进行了设计，并采用调整焊接工艺参数的方式，获得了不同工艺条件下的多丝埋弧焊焊缝成形结果。焊缝成形结果采用熔宽、母材上表面2mm深度处熔宽、熔深以及余高四个参数进行描述。对送丝速度随焊接工艺调整后的变化规律进行了测量与总结，发现送丝速度与焊接电流呈较好的线性关系。在对目前已经提出的热源模型进行总结的基础上，对现有的热源模型进行了分析与改进。通过对双椭球模型的参数敏感性测试与参数回归，获得了通过焊缝成形结果获得双椭球热源模型参数的回归公式。在对热源模型进行进一步的分析后，提出了一个新的面-体热源复合模型。在这个新的复合热源模型中，面热源基于高斯热源，并根据双椭球热源的构想，对高斯面热源进行了改写。针对多丝埋弧焊，对复合热源进行了分段改写，使其能够在任意焊丝间距的条件下使用。此复合热源考虑了电弧倾角对热源的影响，以模拟在焊接过程中不同电弧倾角对焊缝成形结果的影响。基于上述工作，对于常用的多丝埋弧焊焊接工艺参数采用反演算法热源模型参数进行了研究。对于提出的面-体复合热源，讨论了其参数的敏感性，并基于敏感性分析结果，对热源模型进行了相应的简化。根据多丝埋弧焊热源模型的特点，采用模式搜索法作为反演算法进行由果及因的分析。针对热源模型参数数量级差异的问题，对模式搜索法进行了改进，并采用改进后的模式搜索法获得了不同焊接电流与电弧电压条件下的热源模型参数。同时，研究了在不同坡口、板厚、焊接速度、散热条件以及筒体纵缝焊条件下等因素对焊缝成形结果带来的影响。反演算法只能获得已经经过试验的工艺条件所对应的焊接热源模型参数，为将反演算法获得的热源模型参数加以推广，分别采用最小二乘法拟合、神经网络、支持向量回归机对反演算法获得的热源模型参数进行了函数拟合。发现最小二乘法拟合对单丝焊的预测结果较为准确，而双丝焊的预测结果则出现了较大的偏差。神经网络的预测结果与验证试验的结果普遍较为接近，是一种较为优秀的参数推广方案。支持向量回归机由于其算法中未知参数较多，对热源模型参数预测的结果误差较大。利用三丝埋弧焊验证试验，将参数推广获得的热源模型参数加以验证，发现参数推广方案能够很好的获得多丝埋弧焊的热源模型参数，并通过有限单元法计算获得相应的焊缝成形结果。为分析在多丝埋弧焊过程中焊剂的影响以及焊缝成形与熔池内流动状态，根据流体力学的基本方程，构建了埋弧焊在受到重力模型、电磁力模型、能量传输模型、熔化-凝固模型以及表面张力模型等控制的熔滴过渡与熔池流动模型。采用熔池流动行为模型，对三丝埋弧焊的焊缝成形进行了模拟。讨论了表面张力大小、表面张力温度系数、接触角以及送丝速度对焊缝成形的影响。最后获得的三丝焊焊缝成形参数与实际焊接结果相比误差为2.76%。更多还原

【Abstract】 Because of the high deposition efficiency and production efficiency of multi-wirewelding, it becomes the most important welding process of the pipeline and pressure vesselwelding. However, the welding process parameters are always unavailable, and they areusually decided by experience. The most widely used heat source decision method is trial.Because of the experience limit and the time limit, it is easy to cause the artificial errer, and itis hard to ensure the accuracy of the heat source model. On the other hand, it will increase thedevelopment cost. It is emergent to develop a more reasonable heat source model and themethod to obtain its parameters according to the process parameters, which can reduce theartificial error. Inverse analysis is a from-effect-to-cause algorithm, which can determine themost accurate input parameters according to the effect parameters without the mapping ofthem. When the inverse analysis is introduced to the determination of the heat source modelparameters, it can increase the precision of the simulation, reduce the experiment and save thetrail cost. However, the reverse analysis is based on the result of the experiment, and it is notable to obtain the heat source parameters according to the unexperimented process parameters.The result is discrete. For the process parameters without experiment, the intelligentoptimization method which is based on the results of the inverse analysis can extent theparameters and make the results continuous. With the analyse of the inverse analysis and theintelligent optimization method, the heat source parameters according to the different weldingprocess parameters can be obtained directly. The method can reduce the cost of trail andincrease the precision of the simulation greatly. In this work, the error of the prediction modelwas controlled within5%.For the prediction of the heat source parameters by the reverse analysis and intelligentoptimization method, the welding experiment was applied. A new measuring method was designed to obtain the temperature-dependent thermal conductivity and the heat capacity ofthe base metal and the welding flux. The size of weld measurement was designed. Byadjusting the welding process parameters, the different sizes of weld of the correspondingprocess conditions were obtained. The size of weld is composed by the weld width, the weldwidth at2mm depth from the top surface, the penetration and the reinforcement. The weldfeed rate influenced by the welding process parameters was also measured and analyzed. Theresults showed that the weld feed rate was proportional to the welding current.Based on the summarization of the heat source models which were available nowadays,the heat source models were analyzed and improved. The parameter sensitivity of the doubleellipsoid heat source model was analyzed, and a regress function of the weld size influencedby the double ellipsoid heat source model parameters was obtained. A new surface-bodyhybrid heat source model was put forward. In the new hybrid heat source model, the surfaceheat source was based on Gaussian heat source model. Considering the character of thedouble ellipsoid model, the surface heat source model was rewritten. According to the requestof the multi-wire welding, the hybrid heat source model was rewritten to be a separated model,and it could be applied on the multi-wire welding of any distance of the wires. The deflectangle of the welding arc was also considered. It makes the heat source model can simulate thedifferent deflect angel of the welding wire.The inverse analysis was applied to research the relationship between the welding processparameters and the heat source model parameters. The parameter sensitivity of the hybrid heatsource model was analyzed. Based on the parameter sensitivity results, the hybrid heat sourcemodel was simplified. According to the character of the multi-wire submerged arc welding,the pattern search method was applied. The pattern search method was improved with theproblem of the magnitude of the different heat source parameters. The relationship betweenthe different welding process parameters and the heat source model parameters was obtained.The different groove, thickness of the base metal, welding speed, heat dissipation and basemetal shape were obtained. The sizes of weld were also analyzed.The inverse analysis can only obtain the heat source corresponding to the experimentedprocess parameters. In order to extend the results of the heat source model, the least squaremethod regression, the artificial neural network and the support vector machine was applied on the inverse analysis results. It showed that the least square method regression result of thesingle wire welding was accurate, and that of the tandem wire welding was in a high error.The artificial neural network results showed good agreement with the verification results. Thesupport vector machine results had a high error with the verification results since it had manyunknown parameters. The regression results were verified by the triple wire welding. The heatsource parameters predicted by the regression model was applied on the triple wire weldingmodel, and the size of weld was simulated. It showed that the regression result could predictthe heat source model parameters very well.In order to analyze the influence of the flux, the appearance of the weld and the flow in theweld pool, a hydrodynamic model was designed. The model was controlled by the basefunction of the hydrodynamics. The model considered the gravity model, the electromagneticforce model, the heat transfer model, the solidification model and the surface tension model.The appearance of the weld and the flow in the weld pool of triple wire welding weresimulated by the hydrodynamic model. The simulation results showed that the weld pool sizeof triple wire welding was in little difference with the experiment result, and the error was2.76%.更多还原