模拟成形性的计算机预测

运用人工神经网络,对板料基本成形性(单向拉伸、平面应变等基本试验的参数)与模拟成形性(拉深、杯突、扩孔、福井等模拟试验指标)二者的相关性进行了研究.在大量试验数据和反向传播算法的基础上,建立了描述相关性的B-P网络模型.通过该模型,对已知基本成形性参数的板料的模拟成形性指标进行了计算机预测,预测结果与试验结果比较接近.本文的研究方法和结果表明,人工神经网络是研究板料的模拟成形性的一条有效途径.

Prediction of Simulant Formability by Computer

The relationship between the fundamental formability (tension, plane-strain, etc) and the simulant formability (deep-drawing, cupping, hole-expansion, fuki's cup) was studied by means of artificial neural network. Based on a lot of experimental data and the back-propagation algorithm, a B-P neural network model on the relationship was established. The variables of simulant formability of the sheet metal with known parameters of fundamental formability were predicted by computer. The predicted results are in good agreement with the experimental results. It is shown that artificial neural network is effective to study the simulant formability of the sheet metal.