基于分布式光纤测温的施工期混凝土坝温度场神经网络重构

Neural network reconstruction of temperature field during concrete dam construction period based on distributed optical fiber temperature monitoring

施工期混凝土坝的温度场是复杂的不稳定温度场,显式温度统计模型有时难以合理反映不同温控因素和混凝土坝温度之间的非线性关系.首先从施工期温度影响各因素的理论解析解入手,挖掘并选取施工期混凝土坝温度场的影响因子,以挖掘出的温度影响因子作为输入矢量,实测温度为输出矢量,建立了施工期混凝土坝温度的神经网络隐式时空分布模型.结合西南某在建混凝土坝工程的分布式光纤测温资料,分别建立了显式温度时空分布模型以及神经网络隐式时空分布模型.分析表明,相对于显式温度时空分布模型,建立的神经网络隐式温度时空分布模型重构的温度场精度更高,可准确反映不同时刻施工期混凝土坝的温度状态.

In light of the explicit temperature space-time distribution model can not reflect the nonlinear re- lationship between temperature field messages during construction, an implicit temperature space-time dis- tribution model is established by using neural network, starting with theory analytical solution of tempera- ture influencing factors of concrete dam during construction, then finding and choosing that influencing factors as input items, making measured temperature as output data. Using the temperature data of optical fiber system in a concrete dam project in Southwest China during construction to establish explicit and im- plicit temperature space-time distribution models, through analyzing many projects used as case studies, it is shown that the temperature field recreated from implicit temperature space-time distribution model based on the distributed optical fiber temperature measuring information of typical dam section, is more precise; it can show temperature state of concrete dam during different construction periods.