Structural Member Strength Prediction Using Backpropagation Neural Network: A Tool for Retrofitting Intervention Integrating Non-linear Static Analysis

The research was derived from extensive literature reading and addressed the gap in strengthening existing buildings. The study aims to create a model that would correlate the concrete's compressive strength to nondestructive tests (NDTs), establish the strength of in-situ structural members of an existing building using the model, and propose retrofitting intervention strategies as mitigation measures against ground motions. The study presents the artificial neural network (ANN) as the governing model for strength predictions over multi-linear and quadratic regressions. Sensitivity analysis gives prevalent insights into which factor influences the forecast among the input variables. This prediction model has been initiated to evaluate the in-situ strength of the case study building for the analysis following nonlinear static procedures. Two retrofitting interventions were then developed to compare with the performance of the existing three-story building. Predominantly, a performance-based design employing pushover analysis was done where the idealized curves were generated, projecting the base shear and displacements concerning the behavior of the building (ductile or inelastic behavior). This research evaluates the passing criteria of the building based on the performance objectives provided by American Society of Civil Engineers (ASCE) 41–17. The structural member checks in terms of member chord rotations, member shear forces, joint shear stress, and inter-story drifts in connection with the base shear and target displacements evaluation proposed the best retrofitting intervention. The research showed that Case II (retrofitting by shear walls) intervention provided the lowest base shear and passed the considered member checks than RC jacketed with FRP wrapping interventions.