基于神经网络的混凝土强度预测

Prediction Model of Concrete Strength by Artificial Neural Network

在传统预测混凝土强度的基础上,提出一种基于人工智能的新的预测方法,建立了两种神经网络模型:BP神经网络和RBF神经网络,实现了从新拌混凝土成分及其特性到硬化后混凝土强度之间的复杂的非线性映射。通过对试验数据的学习,网络结构可以早期预测混凝土28d抗压强度。另外,还利用BP神经网络模拟分析了混凝土成分质和量的变化对抗压强度的影响,其结果符合已知的经典混凝土强度变化规律,表明神经网络模型具有较高的精度和较强的泛化能力。

It has an important significance to improve the quality of engineering project and to speed up the progress of construction, that the 28-day compression strength value of concrete is realized earlier than planned time. In the paper, on the basis of methods for prediction of concrete strength at home and abroad in the past decades, hence, a new prediction approach based on artificial neural network is suggested, and establishing the multi-layer feed-forward neural network models to implement the complex non-linear mapping from the grading and characteristic values of fresh concrete to the strength to the strength of hardened concrete. Through the study for the plenty experimental data of concrete, that the 28-day compression strength of concrete can be predicted at earlier time by the intelligent system. Furthermore, the system is also used for simulated analyses of the effects of the variations for quality and quantity of concrete components on the compression strength of concrete, and the gainable results correspond to some known regularity for the variation of concrete strength, thus the high accuracy and strong generalization ability of system are shown.