Research on the Construction and Application of a Generative AI-Based Tilting Prediction Model for Ancient Buildings

Ancient buildings, as an integral part of human cultural heritage, face tilting issues that not only affect their aesthetic appeal but may also lead to structural instability or even collapse, posing serious threats to building safety and human lives. Traditional monitoring methods rely on physical sensors and manual inspections, which suffer from limitations such as data scarcity, low prediction accuracy, and poor real-time performance. To address these challenges, this study proposes a generative artificial intelligence (AI)-based tilting prediction model for ancient buildings. By leveraging Generative Adversarial Networks (GANs) for data augmentation, diverse combinations of tilting and contributing factor data are generated, effectively resolving the issue of data scarcity. The model adopts a Multi-Layer Perceptron (MLP) architecture, with inputs including contributing factors such as foundation settlement, vibration frequency, temperature, and humidity, and outputs including tilt angles and tilt rates, enabling accurate capture of complex nonlinear relationships. Experimental results demonstrate that the model significantly outperforms traditional methods in terms of prediction accuracy and generalization ability, and its effectiveness is validated through a real-world monitoring system. This study provides scientific and technical support for early warning and preventive conservation of ancient building tilting, offering significant theoretical value and practical application.