铝型材柔性三维拉弯回弹预测及工艺优化

发布时间：2019-04-02 21:18

【摘要】：随着车辆轻量化和工业发展,传统的铝型材二维拉弯成形构件己不能满足工业需求,铝型材三维拉弯成形构件越来越受到关注。本文提出了一种新型柔性三维拉弯成形工艺,实现了快速重构模具型面和加工多种截面型材。该成形工艺可进行铝型材三维弯曲的加工,且成形面可重构,能有效地缩短模具的设计、调试和生产的时间,从而提高生产效率。铝型材的三维拉弯成形是一个复杂的力学过程,型材成形形状难以精确控制,而且拉弯成形过程中构件易出现回弹、起皱和破裂等成形缺陷。为了得到成形质量较高的三维拉弯成形铝型材构件,需要对成形过程中的工艺参数进行严格控制和优化。本文对柔性三维拉弯成形工艺进行了系统地研究,提出了回弹预测方法和工艺优化方法。本文将拉弯工艺与多点成形思想进行有机地结合,提出了一种新型三维拉弯成形工艺,该工艺将三维拉弯成形分解为水平弯曲和垂直弯曲。首先介绍了该工艺的成形原理、成形过程和关键部件柔性模具单元体的结构,并建立了柔性三维拉弯成形的有限元模型。研究内容包括模型简化、模拟算法选择、材料本构模型、单元选择、接触和摩擦处理以及边界条件处理等问题,为后续的柔性三维拉弯成形工艺研究中有限元模拟部分提供理论依据。回弹是影响拉弯成形工艺的最大难点,针对三维拉弯成形工艺回弹的特点,本文提出了采用支持向量回归算法对三维拉弯成形回弹进行预测的模型。目前人工神经网络是常用的回弹预测算法,支持向量回归算法是一种与人工神经网络类似的机器学习方法,但是支持向量回归算法的泛化能力高于神经网络算法。首先分析了弹性模量、屈服应力、预拉量、补拉量、水平弯曲角和垂直弯曲角对回弹影响的规律。然后以这些参数作为输入参数,水平回弹角和垂直回弹角作为输出参数,分别采用支持向量回归算法和人工神经网络算法建立了柔性三维拉弯成形的回弹预测模型。采用相同的样本对两个模型进行训练和精度检验,证实了该模型预测精度高于人工神经网络模型。传统的拉弯工艺成形过程中拉力保持不变,本文提出了三维拉弯成形过程中采用变拉力的方法,并以减小回弹为目标建立了变拉力轨迹优化模型。首先以复杂T型材为例分析了三维拉弯成形不同阶段的应力应变状态,建立了力学分析模型。将拉力分为四个阶段,预拉阶段、水平拉弯阶段、垂直拉弯阶段和补拉阶段,将各阶段的拉力、预拉量以及补拉量作为设计变量,以回弹最小化为优化目标建立了变拉力优化模型。优化模型以型材最小减薄率和最大增厚率作为约束条件,用试验数据建立目标函数和约束条件的响应面代理模型,然后采用粒子群算法对不同阶段的拉力大小进行优化。通过试验对优化后的变拉力组合与未优化的拉力组合成形后回弹进行比较,证明优化后的拉力轨迹可以有效地减小回弹。本文对矩形型材的三维拉弯成形进行了研究,建立了其三维拉弯成形的多目标优化模型。由于矩形型材截面中空的特性,其三维拉弯成形过程中易出现截面凹陷的成形缺陷,本文提出了填充钢块抑制截面凹陷的方法。分析了预拉量和补拉量对截面凹陷影响的规律,发现增加预拉量和补拉量会引起截面凹陷增加,但增加预拉量和补拉量是减小回弹的主要方法。为了达到同时控制回弹和截面凹陷的目的,建立了矩形型材三维拉弯成形的多目标优化模型。该模型以预拉量和补拉量作为设计变量,以回弹最小化和截面凹陷最小化作为目标函数,以型材最小减薄率和最大增厚率作为约束条件。用多项式响应面法建立目标和约束的代理模型,用非支配遗传算法对其进行多目标优化,得到优化解集,对解集进行分析后选择最优解。优化后的预拉量和补拉量组合能够同时达到减小回弹和截面凹陷的目的。本文还提出了基于迭代回弹补偿优化柔性三维拉弯成形模具型面的方法。该方法以控制回弹为目的,根据回弹量大小对成形面进行迭代补偿,直到成形件的形状偏差满足要求。针对型材柔性三维拉弯成形特点,提出了分段补偿因子的概念,根据回弹大小对型材进行分段,根据分段型材的回弹量分别采用不同的回弹补偿因子。使用优化后的模具型面进行三维拉弯成形,可以极大地降低与目标形状的形状偏差。
[Abstract]:With the light weight and industrial development of the vehicle, the traditional two-dimension drawing-forming component of the aluminum profile has not met the industrial demand, and the three-dimensional stretch-bending forming component of the aluminum profile is more and more concerned. In this paper, a new flexible three-dimensional drawing process is proposed, which realizes the rapid reconstruction of the profile of the mould and the processing of a variety of cross-sectional profiles. The forming process can be used for processing the three-dimensional bending of the aluminum section, and the forming surface can be reconstructed, so that the design, the debugging and the production time of the die can be effectively shortened, and the production efficiency is improved. The three-dimensional bending and forming of the aluminum profile is a complex mechanical process. The shape of the profile is difficult to control precisely. In order to obtain the three-dimensional drawing-forming aluminum profile component with high forming quality, the process parameters in the forming process need to be strictly controlled and optimized. In this paper, the flexible three-dimensional bending forming process is studied systematically, and the method of springback prediction and the process optimization method are put forward. In this paper, the pull-bending process and the multi-point forming idea are organically combined, and a new three-dimensional drawing-bending forming process is proposed, and the three-dimensional bending forming is decomposed into horizontal bending and vertical bending. First, the forming principle, forming process of the process and the structure of the flexible die unit of key parts are introduced, and the finite element model of flexible three-dimensional drawing and bending is established. The research contents include model simplification, simulation algorithm selection, material constitutive model, cell selection, contact and friction treatment and boundary condition treatment. The springback is the most difficult point to influence the bending forming process. In the light of the characteristics of the springback of the three-dimensional pull-bending forming process, this paper presents a model for predicting the springback of three-dimensional pull-bending forming by using the support vector regression algorithm. The artificial neural network is a common rebound prediction algorithm, and the support vector regression algorithm is a kind of machine learning method similar to that of the artificial neural network, but the generalization ability of the support vector regression algorithm is higher than that of the neural network algorithm. First, the law of elastic modulus, yield stress, pre-pull, pull-up, horizontal bending angle and vertical bending angle on springback is analyzed. Then, using these parameters as the input parameters, the horizontal rebound angle and the vertical rebound angle as the output parameters, the rebound prediction model of the flexible three-dimensional pull-bending formation is established by using the support vector regression algorithm and the artificial neural network algorithm, respectively. The two models were trained and verified with the same sample, and the prediction accuracy of the model was higher than that of the artificial neural network model. In the traditional drawing-bending process, the tension remains the same. In this paper, the method of changing the tension in the process of three-dimensional drawing and bending is put forward, and a variable-tension trajectory optimization model is set up with the aim of reducing the springback. First, the stress-strain state of three-dimensional pull-and-bending forming is analyzed by taking the complex T-section as an example, and the mechanical analysis model is established. The pulling force is divided into four stages, a pre-drawing stage, a horizontal pull-bending stage, a vertical pull-bending stage and a pull-up stage, and the pulling force, the pre-pulling amount and the pull-up amount of each stage are taken as a design variable, and a variable-tension optimization model is established for the optimization target by the rebound minimization. In this paper, the response surface agent model of the objective function and the constraint condition is established by using the test data as a constraint condition based on the minimum thinning rate and the maximum thickening rate of the profile, and then the size of the pulling force at different stages is optimized by the particle swarm algorithm. The result shows that the optimized tension combination can reduce the springback effectively by comparing the optimized tension combination with the unoptimized tension combination. In this paper, the three-dimensional bending and forming of the rectangular profile is studied, and the multi-objective optimization model for the three-dimensional bending formation is established. Due to the hollow character of the cross section of the rectangular section, the forming defect of the cross-section depression is easy to occur in the three-dimensional curve forming process, and the method of filling the steel block to suppress the cross-section depression is proposed. The influence of the amount of pre-drawing and the amount of pull-up on the depression of the cross-section is analyzed, and it is found that the increase of the amount of pre-drawing and the amount of pull-up can cause the increase of the cross-section depression, but the increase of the amount of pre-drawing and the amount of pull-up is the main method to reduce the rebound. In order to achieve the purpose of simultaneously controlling the springback and the cross-section depression, a multi-objective optimization model for the three-dimensional drawing and bending of the rectangular profile is established. The model takes the pre-draw amount and the pull-up amount as the design variable to minimize the rebound and minimize the cross-section depression as the objective function, and the minimum thinning rate and the maximum thickening rate of the profile are taken as the constraint conditions. The objective and constrained agent model is established by the polynomial response surface method, and the multi-objective optimization is carried out by the non-dominant genetic algorithm to obtain the optimized solution set, and the optimal solution is selected after the solution set is analyzed. The optimized pre-pulling amount and the pull-up amount combination can achieve the purpose of reducing the rebound and the cross-section depression at the same time. The invention also provides a method for optimizing the profile of a flexible three-dimensional pull-bending forming die based on the iterative rebound compensation. The method is used for controlling the rebound as the purpose, and the forming surface is subjected to iterative compensation according to the size of the rebound quantity until the shape deviation of the forming part meets the requirement. According to the characteristics of the flexible three-dimensional drawing and bending of the profile, the concept of the section compensation factor is put forward, and the sectional material is segmented according to the rebound size, and different rebound compensation factors are respectively used according to the springback amount of the sectional profile. The shape deviation of the target shape can be greatly reduced by using the optimized mold profile for three-dimensional drawing and bending.
【学位授予单位】：大连理工大学
【学位级别】：博士
【学位授予年份】：2015
【分类号】：TG389

【相似文献】