基于深度学习的软钢力学特性

目的针对机械工程中软钢材料在大塑性拉伸载荷下力学特性分析的问题,提出一种基于深度学习的分析方法来预测其力学特性。方法首先对软钢材料不同台阶角度展开拉伸实验,并将采集到的实验数据利用智能技术进行预测分析。实验模型设计为双层结构,第1层结构采用共享全连接层特征输入,第2层使用极端随机树和长短时记忆网络做联合深度训练,并对训练结果经过激活函数计算后统一输出。采用联合训练模型在实验测试集上能较好地反映出应变与应力的变化趋势、速度和数值关系。结果实验结果显示,利用联合训练模型比单一ET和LSTM预测技术在拟合效果上分别提高了28.3%和63.5%。结论利用新模型取得较好的预测效果,这为分析金属阻尼器大塑性拉伸载荷下软钢材料力学特性的分析提供了重要的参考。

Mechanical Properties of Mild Steel Based on Deep Learning

The work aims to propose an analysis method based on deep learning to predict the mechanical properties of mild steel, in order to analyze the mechanical properties of mild steel under large plastic tensile load in mechanical engineering. Firstly, the tensile experiments were carried out to the mild steel materials with different step angles and the collected experiment data were analyzed by intelligent technology. The experiment model was designed to be a two-layer structure. The first layer was shared full connection layer for feature input. The second layer adopted extreme random tree and long-term and short-term memory network to carry out parallel depth training, and output the training results after activation function calculation. The parallel training model could better reflect the strain and stress change trend, velocity and numerical relationship in the experimental test set. The experimental results showed that the parallel training model could improve the fitting effect by 28.3% and 63.5% respectively compared with the single ET and LSTM prediction technology. Good prediction results can be obtained through the new model, which provides an important reference for the analysis of mechanical properties of mild steel materials in the metal damper under large plastic tensile load.