An accurate and efficient simulation of the hysteretic behavior of materials and components is essential for structural analysis. The surrogate model based on neural networks shows significant potential in balancing efficiency and accuracy. However, its serial information flow and prediction based on single-level features adversely affect the network performance. Therefore, a weighted stacked pyramid neural network architecture is proposed herein. This network establishes a pyramid architecture by introducing multi-level shortcuts to directly integrate features in the output module. In addition, a weighted stacked strategy is proposed to replace the conventional feature fusion method. The weights of the features are determined based on their levels. These basic principles are verified, and key network settings are discussed. Subsequently, the redesigned architectures are compared with other commonly used algorithms. Results show that the testing mean-square error (MSE) loss of the networks on varied datasets can be reduced by an average of 34.7%. The redesigned architectures outperform 87.5% of cases, and the proposed Pyramid-GA network has the best overall performance.
View on arXiv