Artificial neural network modeling of confinement of carbon fiber reinforced polymer as retrofitting materials in columns

• Main Authors: Mendoza, Ken Patrick G., Tecson, Ryan Ray R., Victorino, Maila Marie T.
• Format: text
• Language: English
• Published: Animo Repository 2011
• Subjects: Neural networks (Computer science); Carbon fibers; Columns, Concrete > Reinforcement; Civil Engineering
• Online Access: https://animorepository.dlsu.edu.ph/etd_bachelors/10600
• Institution: De La Salle University

The utilization of Carbon Fiber Reinforced Polymer (CFRP) as a retrofitting material has proven its strengthening effects on existing circular, square and rectangular concrete columns. To further develop the use CFRP in the field of civil engineering, theoretical models are needed to predict the effectiveness of CFRP confinement in columns with regards to its compressive strength. The Self-Organizing Map (SOM) toolbox was used to classify data according to the parameters that have similarities which is observed by the toolbox. Each grouping classified through SOM that showed evident relationship among its parameters was observed, analyzed and was related to the ultimate confined compressive strength and increase in strength of the confined columns. From the analysis of the best SOM models, parameters which have significant effect on the CFRP-confinement were then chosen to be used for the back-propagation models. These back-propagation models were then compared to existing models by different authors to verify its accuracy. Three artificial neural networks consisting of circular and non-circular data were developed to predict the ultimate confined compressive strength (fcc). The parameters that were considered in the back-propagation model are volumetric ratio of carbon fiber (Pcfrp), volumetric ratio of steel (Ps), unconfined compressive strength (fco), and the columns geometrical properties (b, h, and L). The models performed better than some models with regard to their correlation (R).