A High-formwork support system (HFSS) is essential during the construction process for complex structures. A lack of an effective monitoring method occasionally causes the system's tragic collapse. In order to establish a reliable method for structural condition identification, experimental measurements on structures are always required. However, in practice, the experiments on sites are either expensive or difficult to conduct. A data-driven algorithm convolutional neural networks (CNNs) for working status monitoring of the HFSS structure is proposed in this study. First, a finite element (FE) model for the HFSS was developed and optimized by using the genetic algorithm (GA). Then, the optimal FE model was employed to produce structural response data for three statuses of the HFSS, such as normal working status, local instability, and fully unstable statuses. The generated dataset was adopted to train a CNNs classifier, which can correctly predict the working states of the structure. Finally, the potential of CNNs was validated on the experimental measurements derived from the HFSS, and the performance of CNNs and support vector machine (SVM) was compared. CNNs performed much better than SVM on the experimental dataset. Moreover, this study developed a Retrieval-Augmented Generation (RAG) model by leveraging a Generative Pre-Trained Transformer (GPT) to synthetically generate an SHM report to describe the structural condition of the given HFSS structures. A knowledge graph (KG) was also developed to enable comprehensive, reliable, informative SHM contexts. Multiple evaluation metrics were employed to assess the performance of the RAG model. The findings indicate that the RAG model could generate accurate and reasonable SHM reports for HFSS.
finite element model updating; CNNs; high-formwork support system; working status prediction; GPT; knowledge graph