【作者】 王鑫； 门月； 张浩宇； 周舸； 张思倩； 李锋； 陈立佳；

【Author】 WANG Xin;MEN Yue;ZHANG Haoyu;ZHOU Ge;ZHANG Siqian;LI Feng;CHEN Lijia;School of Materials Science and Engineering,Shenyang University of University of Technology;

【机构】 沈阳工业大学材料科学与工程学院；

【摘要】 对α、β双相Ti-6Al-4V钛合金不同变形条件(温度:650-850℃,应变速率:5×10^-4～5×10^-2 s^-1)进行了超塑性拉伸试验,并对该合金超塑性变形组织演变进行了研究。研究过程采用BP人工神经网络技术,构建了基于超塑性变形温度、应变速率、延伸率为输入变量,晶粒尺寸、相比例为输出变量的三层BP人工神经网络模型,预测了该合金不同超塑性变形实验条件下的微观组织演变。同时,对该模型的隐含层数、神经元个数、输入输出数据、算法函数以及数据进行了科学合理的优化。研究结果表明:采用该方法的预测精度较高,网络预测值与试验数据值相对误差小于4.9%,能够较好的预测Ti-6Al-4V合金的超塑性变形过程的组织演变规律。更多还原

【Abstract】 A three-layer BP artificial neural network model of Ti-6 Al-4 V dual-phase titanium alloy was constructed by artificial neural network technology,with superplastic deformation temperature,strain rate and elongation as input variables,grain size and phase ratio as output variables.Microstructural evolution of the alloy under different superplastic deformation process(Temperature:650-850℃,Strain rate:5×10^-4～5×10^-2 s^-1) conditions were predicted.In order to obtain better prediction results,the hidden layers of the model,the number of neurons,input and output data,algorithm functions and data have been scientifically and rationally optimized.The results show that the prediction accuracy of this method is higher,and the relative error between network prediction value and experimental data value is less than4.9%,which could predict the superplastic Microstructural evolution of Ti-6 Al-4 V alloy very well.更多还原