Compressive Strength Prediction and Analysis of Concrete Using Hybrid Artificial Neural Networks

Back-propagation (BP)-trained artificial neural networks (ANN) are often used to simulate material behavior characterized by nonlinear, complex, or unknown interactions among the many components of the material. Despite the widespread use of back-propagation neural networks (BPNNs) in various fields, the BP technique has the inherent drawback of being stuck at local minima rather than progressively converging to a maximum global value. This chapter models an accurate and fast concrete mix compression strength estimation technique using ANN and genetic algorithms (GA). The ANN-GA model underpredicts compressive strength values since its NMBE statistics for the BPNN, and ANN-GA models are 0.006 and 0.02, respectively. BPNN, on the other hand, was able to produce a prediction of compressive strength values that were close to optimum. The chapter found that the BPNN model produced more consistent predictions than the frequently utilized hybrid ANN-GA model.