Strength Prediction of FRCM Confined Cylindrical Concrete Specimens by Artificial Neural Network

Document Type : Original Article


Building and Construction Engineering Department, University of Garmian, Kurdistan Region, Iraq



This paper presents developing a new model for predicting the strength capacity of cylindrical concrete specimens confined with Fiber-Reinforced Cementitious Matrix (FRCM) by using the artificial neural network (ANN) technique. For this purpose, a database of 127 reliable specimen results was assembled from the literature. The most sensitive parameters in the strength enhancement were used as input values for the development of the new model which sequentially are; the strength of unconfined concrete, the tensile strength of fabric meshes, the strength of the matrix, the mechanical reinforcement ratio, and the thickness of the matrix. The new model was trained, validated, and tested using MATLAB, which produced a model with a mean square error of 0.00105 and an R-value of 0.9921 that had excellent prediction capacity and high accuracy. Moreover, to evaluate the reliability and validity of the ANN simulated model, the new model was verified against other available models and design code equations in the literature by using different specimens than that used for model development. The new model showed excellent results compared to other models and demonstrated the least rate of average absolute error of about 9%. Finally, a parametric study was investigated to evaluate the effectiveness of sensitive variables on confinement efficiency. The outcomes demonstrated that the new model's predictions for all parameters and the physical performance of the test results were in good agreement.


Main Subjects