An evaluation method for multisource information fusion in tunneling water inrush disasters

During tunnel construction, a substantial volume of data is generated. However, presently, there is a challenge in effectively integrating those data to conduct a precise risk assessment of water inrush. This study develops a novel multisource information fusion method that merges a probabilistic support vector machine, cloud model, evidential reasoning (ER) rule, and Monte Carlo (MC) simulation method to support water-inrush risk assessment under uncertainty. Different models train a variety of information sources to analyze the water-inrush risk value. The evaluation of each model's judgment is based on its performance, which is determined by its reliability and the significance of its weights. Finally, these multiple assessment results are fused at the decision level to achieve an overall water-inrush risk evaluation using the ER rule. The MC simulation method was used to model the uncertainty and randomness underlying the limited number of observations. The Heshan tunnel case in China is used to demonstrate the feasibility and effectiveness of the developed method. The outcomes suggest that the multisource information fusion technique proposed is capable of (1) demonstrating a good ability to handle high conflict information; (2) exhibiting an outstanding assessment performance (90% accuracy) compared to that of single-source assessment models (lower than 70% accuracy); and (3) performing strongly with bias, as it can achieve acceptable assessment accuracy under 5% bias. Therefore, the newly proposed method for fusing multiple sources of information can serve as a practical reference for water-inrush risk assessment and management.