Exploring optimized neural network model for accurate prediction of concrete strength in large bridges

A bridge's functionality and security are compromised by deck degradation. About 45,000 crossings in Ohio require inspection to guarantee that they are structurally sound. Predicting bridge corrosion 1 accurately and promptly is essential to averting accidents. The goal of this study was to create a precise simulation that could be used to forecast Ohio bridge decking characteristics. After thoroughly analyzing the existing research, it was discovered that earlier studies' methods to predict the deterioration of bridge decks were created using different characteristics and techniques. This research suggests combining Bayesian Neural Networks (BNN) with Seeker Optimization Algorithm (SOA) methods to improve the accuracy of bridge deck worsening and concrete strength predictions. This is because there is no certainty that the characteristics and algorithms used by previous researchers will produce precise findings for Ohio's bridges. Utilizing various feature-selection techniques, the structure's initial goal is to identify the "optima"l qualities that may be connected to deck deterioration circumstances, particularly concerning Ohio's bridges. Outcomes from the BNN-SOA techniques that employed the "optimal" characteristics as inputs had been compared with findings from identical deep learning methods that employed the "most frequent" characteristics used in earlier research to verify the structure. Whenever the "optimal" characteristics were used, the combination of DL methods was significantly better at predicting decking characteristics than individual models based on a data set sourced by the Ohio Department of Transportation (ODOT). The system can be effectively employed by other transportation authorities to estimate the degradation of bridge parts better and correctly, considering it was created utilizing ODOT information.