Application Of Deep Neural Network Based On Hidden Fluid Mechanics In Martian Dust Storm Field

With the continuous deepening of human exploration of the universe,Mars has become one of the popular destinations for exploring the solar system.However,sandstorms in the Martian atmosphere are still a difficult problem for scientists and engineers to study,as their impact on the Martian atmosphere is very complex and nonlinear.Traditional numerical simulation methods require a large amount of computational resources and time,and still find it difficult to accurately predict the occurrence and evolution of dust storms.For this reason,this study adopts a data-driven method,and uses a depth learning model based on hidden fluid mechanics to study the Martian dust storm field.This neural network combines physical partial differential equations and deep learning,which can fully utilize the characteristics of physical equations and achieve fast and accurate prediction of complex physical phenomena under the constraints of physical equations.The experimental results show that using the hidden fluid deep learning model to predict the Martian wind field is significantly better than traditional models,especially compared to the pure physical model(MGITM),the wind field error predicted by the model proposed in this study is smaller.In addition,in the test set,the R2 index of carbon dioxide density remained at 0.954,indicating that the model has good fitting performance.In addition,this article found through this model that during sandstorms,the changes in the Martian wind field are more complex,and have a greater impact on atmospheric circulation.In contrast,the changes in the Martian wind field during the absence of sandstorms are more stable and relatively simple,indicating that Martian sandstorms can affect the overall atmospheric circulation of Mars.This article also found that predicting the Martian wind field can provide important reference value for the Martian thermosphere circulation,and is also of great significance for the planning of future Martian human exploration missions.The innovation of this study lies in the fact that on the one hand,we enhanced the training data by combining numerical inversion of physical formulas with deep neural networks.On the other hand,the combination of deep neural networks and physical formulas enables the model in this paper to infer unknown wind field information on Mars through the density and pressure of the Martian atmosphere,achieving accurate prediction of the Martian wind field and providing important support for in-depth research on the Martian environment.In addition,this paper has carried out Exploratory research on Martian dust storms,atmospheric circulation and other fields,and provided new ideas and methods,providing important reference for future Mars exploration missions.The methods and conclusions of this study can not only be applied to Mars exploration,but also provide assistance for research in fields such as natural disasters and climate change on Earth..