基于神经网络的拉深力智能化预测系统
来源期刊：中国有色金属学报2000年第3期
论文作者：吕冬 丁柯 何丹农 张永清 阮雪榆 彭大暑 江勇
文章页码：420 - 425
关键词：拉深力； 塑性力学； 正交试验； 人工神经网络
Key words：drawing loads; mathematical theory of plasticity; orthogonal tests; artificial neural network
摘 要：结合塑性力学理论、 正交试验法及神经网络技术建立了精确计算杯形件拉深力的智能化预测系统。 根据Hill的各向异性理论导出了新的计算杯形件拉深过程中拉深力变化的理论公式， 并求出了最大拉深力。应用正交试验法分析了各工艺参数对最大拉深力的影响。针对在应用BP网络时遇到的两个关键问题进行了讨论并提出了解决方案。应用人工神经网络技术把理论公式与试验数据结合在一起建立了智能化预测系统。
Abstract: With the combination of the mathematical theory of plasticity, orthogonal test and the technology of artificial neuralnetwork, an intelligent prediction system was established to calculate the drawing load of cupdrawing precisely. According to Hill's anisotropic theory, a new theoretical formula was derived to compute the drawing load variations and the maximum drawing load. The technological parameters affecting the maximum drawing load were analyzed by applying orthogonal tests and the following conclusions are drawn: 1) At the notability level of one percent, the maximum drawing load is relevant to blankholder pressure, the radius of the die arc, and the type of lubricant; 2) The most notable factor affecting the maximum drawing load is the type of lubricant, the second is the radius of the die arc, then the blankholder pressure and the radius of the punch arc. By applying t, he artificial neural network, the theoretical formula and the experimental data were combined so that the model error of the theoretical formula was mended, which enhances the accuracy of prediction. Two key problems encountered in the application of BP network were discussed and the solutions were given. Then the intelligent prediction system was constructed, which is not only practically applicable in engineering, but also valuable for the better understanding of the cupdrawing behavior of sheet metal.
