基于Deform软件二次开发和BP神经网络的TA15多向锻造微观组织预报

通过热压缩实验构建了TA15合金的热变形真应力-应变曲线，以此为基础分别建立了合金双相区及单相区温度区间的热变形本构方程；基于热压缩试样动态再结晶的统计数据建立了TA15合金的动态再结晶模型。借助Deform提供的二次开发功能实现相关数学模型的程序化，制定正交实验方案，实现了TA15合金多向锻造变形的微观组织仿真。通过正交实验分析得出各项因子的影响对象及强弱差异，提出了双相区及单相区温度区间内的多向锻造最佳因子组合。建立了TA15合金多向锻造变形微观组织的BP （Back Propagation）神经网络预测模型，将预测结果与有限元仿真结果进行比较，结果表明两种方法的预测结果基本一致，但神经网络具备有限元仿真难以实现的良好细节预测能力，能更为细致地实现对微观组织分布状态的划分。

Microstructure prediction of multi-directional forging of TA15 alloy based on secondary development of Deform and BP neural network

The true stress-strain curve of hot deformation of TA15 alloy was constructed through the hot compression test. The constitutive equations for hot deformation in the temperature range of the dual-phase region and single-phase region of the alloy were then established. A dynamic recrystallization model of TA15 alloy was established based on the statistical data of dynamic recrystallization of hot compressed samples. With the help of the secondary development function provided by Deform, programming of related mathematical models was realized, the experimental plan was formulated by the orthogonal method, and then simulation of the microstructure evolution of the multi-directional forging deformation of the dual-phase region and single-phase region of TA15 alloy was realized. Through analysis of the orthogonal experiment results, the objects under the influence of various factors and the influence degree of the factors were obtained, and the optimal combination of factors for multi-directional forging in the temperature range of the dual-phase region and the single-phase region was proposed. A Back Propagation (BP) neural network prediction model for the multi-directional forging deformation microstructure of TA15 alloy was established. The prediction results were compared with the finite element simulation results. The comparison results show that the prediction results of the two methods are basically the same, but the neural network based method can predict details, which is difficult to be achieved by finite element simulation, and thus can achieve more detailed division of microstructure distribution.