Modeling the ultimate confined compressive strength and increase in strength of carbon-reinforced concrete columns: An artificial neural network approach

• Main Authors: Chan, Kenley Lawrence O., Manalansan, Enrico Francisco, Trespeses, Mariela Ana M.
• Format: text
• Language: English
• Published: Animo Repository 2008
• Online Access: https://animorepository.dlsu.edu.ph/etd_bachelors/11876
• Institution: De La Salle University

Carbon Fiber Reinforced Polymer (CFRP) has proven to be a method of increasing the ultimate confined compressive strength. This paper utilizes the MATLAB software to apply artificial neural network (ANN) modeling through Self Organizing Map (SOM) to classify data sets with similarities observed by the program, and Back-Propagation (BP) to predict the increase in compressive strength of CFRP. Various parameters were considered in the ANN models such as volumetric ratio of steel (us), volumetric ratio of carbon fiber (ucfrp), diameter (D), Length (L), Ultimate confined compressive strength (fcc), Unconfined compressive strength (fco). SOM was used to group data with similar behavior. Each group classified through SOM was observed, analyzed and screened by the proponents of the study to be trained and tested. In order to obtain the best model, organized groups of data that showed evident relationship among its parameters were then used for back-propagation. Back-propagation was applied to determine the output values from the organized data. Through linear regression the R-value which reflects the extent of the linear relationship between target and output was determined. MAE as well as MSE were utilized as error analyses. A parametric study was done to determine the behavior of the model. Chosen models were also compared with the models of the 10 authors of the gathered data. The researchers further enhanced the study by testing the model determined by back propagation. For each group, a combination of two specific parameters was considered as the varying variables while the other remaining parameters were held constant. Results were plotted in 3D through MATLAB and analyses were conducted to find out the relationship of the varying variables to the strength of carbon fiber.