神经网络对混凝土动态特性的预测及试验

为研究混凝土高应变率下的动态特性,基于神经网络对非线性系统的辨识和预测功能,结合Leven-berg-M arquardt算法,利用变截面Hopk inson压杆对聚丙烯纤维混凝土的3种应变率下冲击压缩试验数据,采用BP网络对其峰值应力和对应的应变进行预测,并与试验结果进行了比较。分析表明,预测仿真结果与试验结果是相吻合的,所建立的网络模型可为研究混凝土高应变率下的应力应变关系提供参考。

Experiment and forecast on dynamic characteristics ofconcrete with neural network

Based on the neural network with the capability to recognize and forecast in the non-linear system,the BP neural network with Levenberg-Marquardt method was adopted to forecast the dynamic characteristics of the polypropylene fiber reinforced concrete.Using Hopkinson pressure bar with variable cross-sections, a shock compression test with three kinds of strain rates was carried out and then the neural network was tested with the experimental data.The analysis indicates that the forecasted results are in accordance with experimental reasults and the neural network provides us a new method to study the relation between stress and strain of the concrete with high strain rate.