中文摘要：

基于应力状态下构件快速碳化试验,开展应力状态下混凝土碳化深度的计算机仿真,采用Matlab神经网络工具箱对应力状态下混凝土碳化深度进行训练、仿真和预测.提出采用分项系数γc来反映结构可靠指标的结构耐久性设计方法,给出一定设计使用年限、一定可靠指标下结构耐久性设计时保护层厚度的计算表达式,并对算例中各框架预应力筋的保护层厚度进行计算.结果表明,建立的神经网络模型有较高的预测与仿真能力;提出的设计方法简单、实用,可以作为结构设计和耐久性验算复核的手段.

英文摘要：

Based on the fast carbonation experiment on the effect of stress level in the concrete,numerical simulation of the prestressed concrete carbonation depth was carried out.Network models were built and the neural network toolbox of Matlab was used.The complete numerical experiments included train,simulation and prediction.Subentry coefficient γc was provided to reflect structural reliability index in this durability design method.The formula of cover thickness in the durability design was presented according to certain design life and certain reliability index.The cover thickness of the tendons in a framework was calculated to prove this method.The results show that the neural network model has high prediction and simulation capability and the design method is simple and practical,which can be used in structural durability design.