Extended artificial neural network for estimating the global response of a cable-stayed bridge based on limited multi-response data

A method that can estimate global deformation and internal forces using a limited amount of displacement data and based on the shape superposition technique and a neural network has been recently developed. However, it is difficult to directly measure sufficient displacement data owing to the limitations of conventional displacement meters and the high cost of global navigation satellite systems (GNSS). Therefore, in this study, the previously developed estimation method was extended by combining displacement, slope, and strain to improve the estimation accuracy while reducing the need for high-cost GNSS. To validate the proposed model, the global deformation and internal forces of a cable-stayed bridge were estimated using limited multi-response data. The effect of multi-response data was analyzed, and the estimation performance of the extended method was verified by comparing its results with those of previous methods using a numerical model. The comparison results reveal that the extended method has better performance when estimating global responses than previous methods.