EXAMINING THE PROGRESS IN NEAR SURFACE MOUNT REINFORCEMENT METHODS WITH ANALYTICAL MODELS AND NEURAL NETWORKS

Abstract

Near-surface mount (NSM) strengthening techniques have emerged as effective methods for enhancing the strength and performance of reinforced concrete structures. Despite advancements, the lack of reliable models and standardized methods to predict NSM systems' mechanical behavior remains challenging. This study addresses these gaps with a two-phase methodology.

In the first phase, a database of over 200 experimental data points was analyzed using artificial neural networks (ANN), achieving a 7% absolute error rate and demonstrating strong predictive capabilities. In the second phase, ANN results were optimized with multiple linear regression (MLR), developing a simplified mathematical model with an 18% error rate, offering practicality and ease of use for field engineers.

The proposed models comply with ACI 440.2R safety guidelines and outperform traditional approaches in accuracy and usability, enabling more effective application of NSM techniques. This research advances NSM methodologies by integrating analytical and ANN-based approaches, contributing to developing standardized guidelines and durable reinforced concrete designs.