中文摘要：

为了获取纤维混凝土的热力学参数,自制含内热源的加热-测温装置,依据实测温度值,利用BP神经网络对聚丙烯-玄武岩混杂纤维混凝土热力学参数进行反分析预测。选取导热系数、质量热容和对流系数为待反演参数,利用正交设计得到各参数不同水平组合的试验方案,用ANSYS根据方案进行温度场正分析模拟试验,获取足够数量的训练样本。最后创建BP神经网络模型,实现参数预测。结果表明：最优参数为导热系数1.87 W（/m.K）,质量热容1 218.70 J（/kg.K）,对流系数12.91 W（/m2.K）。反演参数求解的温度值和实测值误差在1.81%~-13.64%范围内,参数符合工程要求的合理值。

英文摘要：

In order to acquire thermodynamic parameter of hybrid fiber reinforced concrete with poly-propylene and basalt fiber, a back analysis prediction was carried out with BP neural network,which based on the testing data of measured temperature gained form homemade device with inner heat source for heating-measuring.Taking thermal conductivity, specific heat and convection coefficient as stay inversion parameters, different experiment schemes of parameter combination were obtained by the orthogonal design, a sufficient number of training samples were provided by ANSYS thermal analysis, the prediction model of BP neural network was obtained and parameter prediction was realized.The results showed that optimum parameters of thermal conductivity, specific heat and convection coefficient were 1.87 W/（m. K）, 1 218.70 J/（kg. K） and 12.91 W/（m2. K）.The error between inversion temperature value and the measured temperature value was 1.81%-- 13.64%, which met the engineering requirements.