Enhancing Submersible Safety: Predictive Modeling and Search Model for Lost Submersibles

Abstract

The Titan submersible accident has raised global concerns regarding submersible safety. To mitigate such risks and enhance survival probabilities, this study develops a safety support system aimed at predicting submersible trajectories and optimizing search operations. A Submersible Position Prediction Model is established, incorporating seawater salinity, temperature, current, and seabed elevation data from the Ionian Sea. Force analysis based on computer simulations is used to generate minute-by-minute positional data, resulting in a probability distribution of its location. Model validation is conducted using Caribbean Sea data. Furthermore, a Two-Dimensional Kernel Density Estimation Search Model is proposed to minimize search time for a lost submersible. The method partitions the probability distribution into four regions, each searched by a robot moving outward from the center, significantly improving search efficiency. A function relating search success probability to time is derived based on robot speed and detection range. The results show that the lost submarine can be found within 1.3 hours. Finally, a multi-submersible position prediction model is introduced, which updates the dive path of one submersible based on real-time positions of others, enhancing coordination and emergency response capabilities.