低温钢纤维混凝土强度试验与预测

Experiment and Prediction of Steel Fiber Reinforced Concrete Strength at Low Temperature

以不同温度（0℃、-5℃、-10℃、-15℃、-20℃）、不同钢纤维掺量及不同水灰比的钢纤维混凝土的抗压、抗拉、抗折与抗剪强度试验结果，建立以温度、钢纤维掺量以及水灰比作为输入矢量，混凝土预测强度作为输出矢量的网络模型。用人工神经网络分别为抗压强度、抗折强度、抗拉强度及抗剪强度建立了合适的网络模型，输入层和隐含层均采用双曲线正切S型传递函数，输出层采用线性传递函数。网络采用Levenberg—Marquardt算法进行训练，对低温钢纤维混凝土的强度进行了预测，预测的相对误差在0-0．05的范围内波动，各训练总标准差与仿真总标准差均在0．3的范围内，取得了满意的结果，这对低温条件下钢纤维混凝土强度预测有一定实用价值。

Based on experimental results of compressive, tensile, fracture and shear strength of steel fiber reinforced concrete with different fiber content and different water cement ratio at 0 ℃,-5 ℃,-10 ℃, -15 ℃ and -20 ℃, the network model was established with temperature, steel fiber volume and water cement ratio as input parameters, and the predicted concrete strength as output parameter. The network model of compressive, tensile, fracture and shear strength was set up respectively by artificial neural network, in which hyperbolic tangent model S transfer function is used for input and hidden layers and linear transfer function for output layer. The Levenberg-Marquardt algorithm was used in network to predict the strength of steel fiber reinforced concrete at low temperature, and its result is acceptable with the relative error in the range of 0-0. 05. the total standard deviations of each training and simulation in the range of 0.3. It has practical value in predicting strength of steel fiber reinforced concrete at low temperature.