Structural analysis has significantly evolved with advancements in artificial intelligence, sensor technology, and cloud computing. Machine learning models can effectively capture structural behavior from real-time sensor data, providing insights into complex nonlinear mechanics. However, conventional monitoring systems often suffer from excessive sensor deployment, leading to increased costs and inefficiencies. This study proposes an optimized structural assessment framework that integrates an optical fiber sensor system with deep learning models, including Long Short-Term Memory (LSTM) and Transformer networks, alongside a Multi-Layer Perceptron model, also known as a fully connected neural network (FCNN). The objective is to enhance predictive accuracy while minimizing the number of required sensors. The research focuses on fiber-reinforced polymer (CFRP)-reinforced concrete structures, where a concrete core is externally reinforced with a CFRP layer, enhancing its durability and load-bearing capacity. Due to the nonlinear behavior of both the material properties and contact interactions, the structural response under loading exhibits significant complexity. The interface between the CFRP layer and the concrete core introduces additional nonlinearities, requiring precise modeling to capture the interaction effects. Simulation data were used to establish a nonlinear mapping between strain inputs and reaction force outputs via the FCNN model. Additionally, sensor placement was optimized by introducing new measurement points, S1G3.5 and S1G9.5, strategically positioned to improve structural response analysis. A comparative evaluation of deep learning models was conducted, highlighting the time-series modeling strengths of LSTM and Transformer networks. These models exhibited superior performance in capturing sequential structural behaviors. The optimized setup demonstrated high predictive accuracy while significantly reducing sensor requirements, improving both cost efficiency and computational performance. Model effectiveness was validated using R2, MSE, MAE, and SMAPE metrics. This study presents an effective methodology for sensor optimization and deep learning-based predictive modeling, offering a cost-efficient and accurate solution for assessing CFRP-reinforced concrete structures