多因素作用下混凝土寿命的BP神经网络预测

Application of BP Artificial Neural Network Modeling in Service Life Prediction of Concrete under Multifactors

影响混凝土及其结构使用寿命的因素主要包括荷载和自然环境的变化,它们对混凝土结构使用寿命的影响过程是错综复杂的,难以用准确的数学式表达。人工神经网络方法具有无需输入变量与输出变量间复杂的相关假设,也无需确定各种计算参数,从而以消除计算参数确定过程中产生的计算误差的特点,使得其在土木结构工程耐久性方面具有广泛的应用。采用动量-自适应学习速率调整算法以及规则化调整对BP神经网络的泛化能力进行了改进,使其误差平方和达到0.000918,提高了BP神经网络的泛化能力;并用改进的BP神经网络对荷载-复合离子-干湿交替作用下混凝土材料的使用寿命进行了预测,避免了在确定计算参数过程中所产生的计算误差,拓宽了多因素作用下结构混凝土寿命预测新方法。

The diffusion mechanism of load and the ambient condition is difficult to give a mathematic expression accurately. Artificial Neural Networks （ANN） have received increasing attention for use in predicting the property of materials in civil engineering. The total chloride concentration of different depths in concrete undergoing load and the ambient condition governs the service life of such conerete. This study investigates the usability of BP artificial neural networks （ANN） to estimate the service life of such eoncrete under load, ions and dry and wet cycles conditions： For this purpose, the parameters considered for the BP ANN inputs are the total chloride at different diffusion times and depths. The test results obtained from the BP ANN are compared with the experimental results. It is seen that the training and testing results are similar to the experimental results.