多点成形中回弹量的测量与控制方法研究

针对多点成形中工件的回弹问题,提出了一种测量和控制回弹的新方法.回弹量的测量方面,采用自行研制的非接触式三维激光扫描装置测量工件表面的三维点云信息;对不在同一坐标系下的测量数据和目标模型数据,用两步配准方法进行全局优化配准,克服了配准过程中常见的局部极小问题;采用改进的Shepard插值函数为初始曲面模型,用基于高斯曲率的采样型值点拟合出NURBS曲面.通过上述步骤快速、准确地计算出回弹量,改变了传统的利用靠模测量回弹量的方法.回弹控制方面,用测量的回弹量试验数据建立BP神经网络回弹预测模型,根据预测量调整基本体群的工具曲面形状,经过多次预测和基本体曲面形状的修改,最终实现控制回弹的目的.通过实例表明,上述方法可行,回弹量可控制到允许的精度.

Springback measuring and controlling during multi-point forming process

A new method for measuring and controlling springback in multi-point forming(MPF) is proposed. To springback measuring problem, the 3d cloud points are acquired firstly by a self-developed non-contact laser scanning system we devised.Global optimal registration between the set of measuring data and the one of model data is realized through two step method, which can improve the usual problem of local minimum. An initial model of surface with the method of Modified Quadratic Shepard is created,then the fitting NURBS surface is calculated by the points based on Gaussian curvature. With the above step,the value of springback is calculated rapidly and exactly,which improve the traditional method which uses the mould to measure the value of springback in the industry. On the aspect of springback controlling, the BP artificial neural network based on the experimental data is built to predict the value of springback, and the working surface constructed by element matrix is modified according to the prediction value.Through repeating the process of predicting and modifying working surface, springback in MPF will be controlled under tolerance. The example with the above method is performed in which springback is decreased significantly.