Neural Networks with an Optimization Algorithm to Gauge Concrete Strength
一种优化算法的神经网络在混凝土强度判定中的应用

This paper discussed characteristics of concrete strength prediction, and upon such analysis, built a multi-layer feed-forward neural network model to perform complex non-linear mapping. Also an emphasis is placed on some algorithms, among which, the basic BP algorithms prove to be incapable because of local minimum, and the introduced GOBPA (Global Optimization BP Algorithm) proves to be successful by results from computer simulation studies. A lot of tests on concrete specimens are performed and the studies show that neural networks perform well and the application of multi-layer feed-forward neural network to concrete strength prediction is feasible with higher accuracy.