High slope stability prediction based on quantum-inspired heuristic optimization algorithm and transfer learning

ObjectiveThis study focuses on the complex issues of high slope stability prediction. A novel prediction model was proposed, combining quantum-inspired heuristic optimization (QHO) algorithm with transfer learning (TL), i.e., QHO-TL model.MethodFirst, a detailed 3D numerical simulation model was developed. It was used to study the slope stability evolution rule in various conditions, e.g., natural state, heavy rainfall, earthquake, and excavation. Next, the performances of the proposed model were compared with traditional machine learning methods and existing models. The comparison considered different geological strength indices, rainfall intensities, and noise levels. It verified the model's adaptability in complex environments. Finally, the global sensitivity analysis was conducted to identify key parameters affecting slope stability and their contributions, supporting slope engineering design and risk control.ResultQHO-TL model achieves an error rate of only 1.6% in storm conditions, which is 48.4% lower than that with random forest model. The proposed model effectively integrates rainfall time-series features and intelligently screens key influencing factors. It could predict dynamic changes in the seepage field ahead of time, improving disaster warning.ConclusionThe proposed model exhibits robust generalization ability for low-frequency yet high-hazard events by accurately identifying critical states during the creep-acceleration phase. It provides scientific support for intervention decisions prior to slope instability occurrences and holds substantial practical value for ensuring safety measures against slope failures in complex geological conditions.