Machine learning-based CFD simulations: a review, models, open threats, and future tactics

This review targets various scenarios where CFD could be used and the logical parts needed for exemplary computations. The machine learning aspect with algorithms that have been implemented suggests design parameters to an algorithm that can be used for bodies in flights and different research-based algorithms that have been used and outlines the advantages, disadvantages, and tools used for computing the algorithm. Since fluid behavior is quite erratic, a single algorithm may not be versatile in every case. In some cases, multiple algorithms are combined for successful simulations. The uniqueness of the review lies in the combination of algorithms for every different case with theoretical analysis and disadvantages, which could be avoided by clubbing another algorithm that overcomes the problem. Since ML is not fully mature yet to provide high accuracy without bit preprocessing in the form of the numerical method, this is one of the heavy limitations that are briefly discussed.

Automated summary

This review explores the applications of CFD in various scenarios and the logical components required for exemplary computations. It examines the influence of machine learning on design parameters with implemented algorithms, as well as research-based algorithms used in different situations. It also analyzes the advantages, disadvantages, and computing tools of the algorithms.