Development of an Intelligent System for Sheet Metal Bending

该项目旨在开发一种智能金属成形系统，特别是普通板材弯曲成形系统，具有生产灵活性和高精度的前馈控制功能。利用最新研制的适合于位移控制的弯曲工装，在线测量板料变形，根据测量数据估计材料性能，并结合高精度的刀具位移进行加工。为了进一步提高控制精度，在控制系统中加入了神经网络。(1)在原型弯曲机上用新研制的圆形或椭圆形弯曲器代替传统的V形弯曲器进行板材弯曲。从而证实，灵活性得到了提高，精度也提高了一倍。这要归功于学习控制系统。(2)在上述刀具试验中，工件和刀具之间的接触点随着刀具的位移而移动。开发了一种智能传感器，可以测量法向和切向的力。将描述复杂现象的神经网络引入到系统中，建立了高精度的控制模型。将变形模型估计的材料性能输入神经网络，从而确定约束条件。因此，塑性模型和神经模型的结合有可能成为一种高效、高精度的控制模型。

This project is to develop an intelligent metal forming system, in particular a common sheet metal bending, that performs flexibility in production and high accuracy incorporating feed forward control. By using a newly developed bending tool set which is suitable for displacement control, deformation of sheet metal is measured on-line ; the material properties was estimated by the measured data and combining with high accuracy of the tool displacement was performed. Further, a neural network was incorporated in the control system for further improvement of accuracy.(1) Instead of conventional V-angled bending tool, the newly developed tools of circular or elliptical shape was used to bend sheet metals on a prototype bending machine. It was thus confirmed that flexibility was improved and the accuracy was also improved by one fold. This is due to the learning control system.(2) In the above tool est, the contact point between the workpiece and the tool shifts according to the tool displacement. An intelligent sensor that measures the forces, both normal and tangential directions, was developed. The contact point and thus bend angle can be measured.(3) A neural network, which describes complicated phenomena was incorporated in the system to develop a control model with high accuracy. The material properties estimated by the deformation model was input into the neural network and binding condition was thus determined. It was thus concluded that combination of plasticity model and neuro-model has its potential of an efficient control model with high accuracy.