为了对裂纹疲劳寿命进行快速准确预测,针对常用的疲劳实验和数值仿真实验中各自存在的实验条件受限、计算步骤繁琐、计算过程耗时较长等问题,以大型桥梁钢桁架中的裂纹为研究对象,提出一种结合裂纹分析软件FRANC3D和长短期记忆(long short term memory,简称LSTM)网络模型的裂纹疲劳寿命预测方法.首先,建立钢桁架的有限元模型,通过裂纹分析软件FRANC3D进行疲劳裂纹扩展分析并建立相关数据集;其次,通过疲劳裂纹的扩展仿真数据关注裂纹的扩展状况,收集裂纹特征信息对裂纹疲劳寿命进行预测.实验结果表明,所提出的方法具有较高的预测准确性,同时极大地提升了训练速度,可为裂纹疲劳寿命预测提供参考.
Crack Life Prediction of Bridge Steel Truss Based on FRANC3D and LSTM
To quickly and accurately predict the fatigue life of cracks,a crack fatigue life prediction method combining crack analysis software FRANC3D and long short term memory(LSTM)network model is pro-posed to address the limitations of experimental conditions,time-consuming calculations,and other problems existed in fatigue and numerical simulation experiments.In this paper,cracks in large bridge steel trusses are studied as the object.Firstly,a finite element model of the steel truss is established,fatigue crack propagation analysis is conducted using crack analysis software FRANC3D,and the analysis data is stored and established to create a crack propagation dataset.Then,the key feature of fatigue crack propagation is collected and extracted to predict the crack fatigue life by the proposed LSTM model-based prediction method.The experimental re-sults demonstrate that the proposed method has superior prediction accuracy and time-saving prediction speed,which can provide theoretical and practical value for predicting crack fatigue life.
NETL
NSTL
万方数据
