自适应模糊神经推理系统在混凝土强度评定中的应用

混凝土强度评定的准确性对结构物的安全性评价具有很大的影响。为了充分利用钻芯法和回弹法这2种常用混凝土测强方法的特点,建立自适应模糊神经推理系统模型来综合评定结构的混凝土强度。将回弹值的常用对数和碳化深度值作为模型的输入,钻芯值的常用对数作为模型的输出。模型参数采用混合算法确定。其中,条件参数采用梯度下降法来调整;结论参数采用最小二乘法来调整。该模型可以有效地映射出训练数据之间复杂的非线性关系。通过对已有的钻芯、回弹试验数据的对比计算,自适应模糊神经推理系统方法的强度预测精度高于常规的回归方法。

Application of Adaptive Neuro-Fuzzy Inference Systems to Concrete Strength Evaluation

The accuracy of concrete strength evaluation has a great influence on the safety assessment of structures. To take full advantage of the common concrete strength testing methods-the drill method and the rebound mcthod, an adaptive neuro-fuzzy inference system (ANFIS) model is built up to evaluate concrete strength. The common logarithms of rebound values and carbonation depth values are chosen as the inputs of the model. The common logarithms of drill values are chosen as the outputs of the model. The model parameter sets are determined by a compound algorithm. In the model, the condition parameters are adjusted by the gradient descend method and the conclusion parameter sets are adjusted by the least square method. The model efficiently maps the complex non-linear relationship of the training data. Through comparison between ANFIS simulation and the traditional regression method using exiting test data, it is found that the predicted result of the ANFIS is more accurate than that of the latter.