基于神经网络预测的TC23合金高温流动行为精确仿真

Accurate simulation of hot flow behavior of TC23 titanium alloy based on neural network prediction

为了研究TC23钛合金的本构关系,开展了一系列热压缩试验获得了其真应力-应变数据。真应力-应变数据的分析结果表明,TC23钛合金在高温变形过程中同时发生了加工硬化、以动态再结晶为主的动态软化和相变,导致流动应力与热力参数呈复杂的高度非线性关系。为了描述和预测TC23钛合金的真应力-应变关系,将热变形参数作为输入,将流动应力作为输出构建了人工神经网络。结果表明,所构建的神经网络能够精确地表征TC23钛合金的高温流动行为。通过将人工神经网络植入有限元软件并建立热压缩试验有限元模型,实现了TC23钛合金高温变形行为的精确仿真。

To study the constitutive relationship of TC23 titanium alloy, a series of hot compression tests were conducted to obtain its true stress-strain data. The analysis results of the true stress-strain data show that work hardening, dynamic recrystallization dominated dynamic softening and phase transformation occur simultaneously in TC23 titanium alloy during the hot deformation process. Thus, the flow stress represents complicated highly-nonlinear relationship with the thermodynamic parameters. To describe and predict the true stress-strain relationship of TC23 titanium alloy, an artificial neural network was constructed, which employs hot deformation parameters as the inputs and flow stress as the output. The results show that the constructed neural network can accurately characterize the hot flow behavior of TC23 titanium alloy. By implanting the artificial neural network into finite element software and constructing the finite element model of hot compression test, the accurate simulation of hot flow behavior of TC23 titanium alloy is achieved.