Predicting Flexural Strength of Concretes Incorporating River Gravel Using Multi-Layer Perceptron Networks: A Case Study of Eastern Nigeria

  • IE Umeonyiagu
  • CC Nwobi-Okoye
Keywords: Artificial Neural Network, Concrete, Washed gravel, Regression, Modelling


Cases of collapsed buildings and structures are prevalent in Nigeria. In most of these cases the cause of the collapse could be traced to the strength of the construction materials which is usually concrete. Secondly, experimental determination of the strength of concrete materials used in buildings and structures is quite expensive and time consuming. This work shows the development of a computational model, based on artificial neural networks for the determination of flexural strength of concrete materials made from prevalent coarse aggregate components from Nigeria. The work involves building a multi-layer perception neural network model which uses experimental data obtained from flexural strength test of concrete made from washed gravel. The flexural strength predictions were compared with predictions from an alternative model based on regression analysis. The results of the study show that for the washed gravel based concrete the regression model prediction has a correlation coefficient of 0.92687 and a sum of squares error of 0.51954100, while the neural network model prediction has a correlation coefficient of 0.98364 and a sum of squares error of 0.00629630. Generally, the models predicted well, but the neural network model predicted better than the regression model. The result of the study has adequately demonstrated a cheap, simple, very quick and accurate alternative to experimental method of concrete strength determination. The method is also simpler and quicker than analytical methods based on regression analysis. Results obtained suggests an option of immense benefit to civil engineers and construction professionals, which would help in economically determining the strength, as well as economical selection of appropriate mix of construction materials, that is a prelude to building strong and cheap buildings and structures.

Building, Civil & Geotechical Engineering

Journal Identifiers

eISSN: 2467-8821
print ISSN: 0331-8443