Abstract:
The present study applies group method of data handling (GMDH) to predict compressive strength of normal strength concrete based on experimentally determined weight, ultrasonic pulse velocity and extraterrestrial solar radiations absorbed by concrete specimen. GMDH are widely used as mathematical modelling and non-linear regression algorithms, and are assumed as specific type of supervised artificial neural networks. Concrete being a multi-phase porous and non-linear material justifies usage of such algorithm as GMDH employs the idea of natural selection to control size, complexity and accuracy of networks being used for various applications like function approximation, non-linear regression and pattern recognition. The effectiveness of algorithm is validated when 60%, 70%, 80% and 100% of normalized and non-normalized data is used for training. GMDH being an intelligent algorithm with ability of learning and adaptation can be conveniently used as an appropriate prediction tool for non-linear complex systems like concrete.