Research on Submersible Safety Procedures through Bayesian Inference

Aiming to reduce the risk of commercial sightseeing submarines, the paper investigates the motion trajectory and over-time trajectory-prediction-rescue model of submarines after loss of contact in the ocean. We have selected the three most significant uncertainties - ocean currents, water temperature, and density. Then we calculate the coordinates of the submersible at different time periods from start to loss of contact. Based on these coordinates, we introduced uncertainties and constructed a random disturbance model, predicted a more realistic submersible motion trajectory. Based on the trajectory model of problem 1, we established a probability model for submersible position through Bayesian inference. Then, we selected 19 key coordinates using the circular search principle and used simulated annealing algorithm to plan the global optimal path between them. We derived the deployment points of rescue vessel. Based on relevant data from the Caribbean Sea, we replaced the ocean current disturbance factors and then used the same method as section 3 to derive the motion trajectory models of unpowered and disconnected submersibles. To address the problem of joint diving of multiple submersibles, a single wheeled vehicle model combined with a circular formation control model is used to construct a multiple-submersible-formations-rescue model. This paper ensures coordinated and comprehensive coverage of environmental and submersible data in various situations, and reduces the rescue time of the host ship. Finally, we obtained the global optimal search path, rescue boat deployment model, and multi submarine formation rescue model which are suitable for most complex seabed situations. With these models, the search and rescue efficiency of a lost submersible will be greatly increased, and the tour risk of the submersible will be reduced. Ensuring the safety of passengers on the submersible can promote the commercial project of submarine sightseeing widely.