Updating Effects on Equivalent Shear Beam Structure Models for Lower-Order Modes Based on Sensitivity-Based Methods and Artificial Neural Networks

Abstract

In this study, a building steel structure that adopted an equivalent shear beam structure model is analyzed, with the updating effects on this simplified analysis model discussed using sensitivity-based and artificial neural network (ANN) methods. Due to the limitations of the sensitivity-based structural model updating method, mean absolute relative error was used to evaluate the reasonable number of hidden layer nodes of the ANN multi-layer perceptron architecture to be applied to the updating equivalent shear beam structural model. A National Earthquake Engineering Research Center (Taipei) steel test structure was selected as the case study. The results reveal that the equivalent shear beam structural model updated modal parameter analysis of lower-order modes is more consistent with the modal test than the 3D finite element analysis. A comparison between the discrepancies between the sensitivity-based and ANN methods suggests that the latter outperforms the former, as indicated by its better performance in terms of predicting the first two modal natural frequencies. This finding demonstrates the applicability of the updated equivalent shear beam model and indicates that structural dynamic response analysis can be conducted using the updated stiffness values of each floor. Therefore, this simplified analysis model could be applied to the vibration analysis and design of multi-story structures (e.g., high-rise steel structures, scaffoldings, and vibrating shaking tables). Furthermore, these findings indicate that this simplified analysis model for multi-story structures could also be applied to the evaluation of old structures.