神经网络在锈蚀钢筋疲劳寿命预测中的应用

Application of artificial neural net in fatigue life prediction of corroded steel bars

在锈蚀钢筋轴向拉伸疲劳试验的基础上,利用BP神经网络较强的高次非线性能力和自学习能力,建立预测锈蚀钢筋等幅疲劳寿命的BP神经网络模型。基于非线性疲劳损伤理论,把等幅疲劳荷载作用下锈蚀钢筋的疲劳寿命预测神经网络模型进一步发展成变幅疲劳荷载作用下的锈蚀钢筋疲劳寿命预测神经网络模型。通过利用锈蚀钢筋混凝土梁的疲劳试验数据对该模型进行训练和检验,证明所建立的预测锈蚀钢筋变幅疲劳寿命的BP神经网络预测模型精度较高,适用性强,可为锈蚀混凝土桥梁的剩余疲劳寿命评估提供必要的依据。

Based on the fatigue test of corroded steel bars,a BP neural net model,by utilizing the high order nonlinear and the self-learning ability,is established for predicting the fatigue life of corroded steel bars under repeated loadings with constant amplitude.On the basis of nonlinear cumulative damage theory,a new BP neural net prediction model is founded for predicting the fatigue life of corroded steel bars subjected to repeated loadings with variant amplitude.The model is trained and examined with the fatigue test data of corroded reinforced concrete beams.The results show that the predicted value of fatigue life of the reinforced concrete beam is in a good agreement with the test value.The new model can be used to provide necessary ground for evaluating the residual fatigue life of corrosion-damaged concrete bridges.