an innovative approach to estimate chloride diffusion coefficient in submerged concrete structures using soft computing

Corrosion is one of the most important and common factors in the destruction of structures. among all kinds of structures, corrosion of submerged structures is of great importance and prevalence due to the impossibility of direct visibility, high reconstruction cost and special environmental conditions. the work done in the field of corrosion of these structures has mainly dealt with modeling the problem in the form of mathematical formulation or using soft computing methods. the work that has established the connection between these two methods has not been done, to the best of our knowledge. this article aims to develop a model in order to estimate the chloride diffusion coefficient in rebar corrosion in submerged concrete structures. present study seeks to address the estimation of chloride diffusion coefficient, which is one of the determinant factors in computing the corrosion time/rate of rebar’s. in this article, using the monte carlo sampling method and the formulas available for chloride diffusion coefficient, we produced 2000 artificial data samples. a variety of methods such as support vector machines (e.g., linear, quadratic, cubic, gaussian), k-nearest neighbors (fine, medium, coarse knn), and two methods of ensemble learning (bagged tree, subspace discriminant) are applied to predict the chloride diffusion coefficient. the results indicated that the quadratic support vector method (with 93.5% accuracy) is the best technique in estimating the chloride diffusion coefficient. best knn model (medium knn) and best ensemble method (bagged tree) have accuracy of 59.9% and 81.3%, resp.