基于回声状态网络的混凝土抗压强度预测模型

Prediction model of Concrete Compressive Strength Based on Echo State Network

为了更准确地实现混凝土抗压强度的预测,建立了基于回声状态网络(ESN)的混凝土抗压强度预测模型,以混凝土中水泥含量等作为模型输入,混凝土抗压强度为输出,对260组混凝土28天抗压强度数据进行训练测试。结果表明,与以往BP神经网络相比,ESN将预测精度提高了3个数量级,能够准确快速地预测混凝土强度,具有较强的非线性逼近能力。

To predict concrete compressive strength more accurately,a concrete compressive strength prediction model based on echo state networks(ESN)is built.Taking the cement of concrete and other index as the model input and compressive strength as the output,260 groups compressive strength data of 28d are trained and tested.Simulation results reveal that the method based on ESN is 3 times more accurate than the BP neural network,shows the effectiveness and applicability,and ESN has strong nonlinear approximation ability.