【摘要】 采用Gleeble1500对高温合金GH648进行了不同温度、不同应变率下的压缩实验。结果表明,材料在950℃到1150℃,应变率低于1/s时体现出动态再结晶特性,在应变率为10/s时体现为动态回复特性。鉴于其在高温变形过程中体现出复杂的动态响应特征,根据所获得的实验数据,应用BP人工神经网络建立了合金的高温本构模型,同时提出了一种改进的学习方法,显著地减缓网络振荡,提高收敛速度。研究为该高温合金的有限元数值模拟提供了必要的前提,采用的人工神经网络可以推广应用于其它非线性关系的确立。更多还原

【Abstract】 Hot compression experiments on superalloy GH648 were conducted with Gleeble1500 thermal simulator at different temperatures and strain rates. It is shown that,at temperatures from 950℃to 1 150℃), superalloy GH648 shows characteris-tics of dynamic recrystallization at strain rate less 1 /s and dynamic recovery at strain rate 10/s. As for ite complicated dynam-ic response characters during deformation at elevated temperatures, artificial neural network with the error back propagation (BP) algorithm was used for establishing constitutive model of superalloy GH648 based on the experimental data. At the same time, an improved method was proposed to reduce network vibration and increase convergence velocity. The research provided a necessary precondition for finite element numerical simulation of superalloy GH648, and artificial neural network applied in this paper can be widely used to establish other nonlinear relations.更多还原