Deep reinforcement learning in fluid mechanics:A promising method for both active flow control and shape optimization

In recent years,artificial neural networks(ANNs)and deep learning have become increasingly popular across a wide range of scientific and technical fields,including fluid mechanics.While it will take time to fully grasp the potentialities as well as the limitations of these methods,evidence is starting to accumulate that point to their potential in helping solve problems for which no theoretically optimal solution method is known.This is particularly true in fluid mechanics,where problems involving optimal control and optimal design are involved.Indeed,such problems are famously difficult to solve effectively with traditional methods due to the combination of non linearity,non convexity,and high dimensionality they involve.By contrast,deep reinforcement learning(DRL),a method of optimization based on teaching empirical strategies to an ANN through trial and error,is well adapted to solving such problems.In this short review,we offer an insight into the current state of the art of the use of DRL within fluid mechanics,focusing on control and optimal design problems.

Numerical simulation of condensation shock in partial cavitating flow on a hydrofoil

We present a numerical simulation of the so-called condensation shock on a NACA 0015 hydrofoil with a finite mass transfer model for the first time.Most recently,condensation shock was indentified in the experimental measurement as a mechanism for partial cavity shedding in the flow past a hydrofoil.Compressible solvers,which were adopted in the previous simulations to study such phenomenon numerically,are extremely time consuming due to the limitation of acoustic Courant number.In this work,we consider a finite mass transfer model for cavitation flow with slightly modification.Our numerical results show that the finite mass transfer model can be successfully applied for calculating the condensation shock in the flow past a hydrofoil.The dynamics of the condensation shock on the hydrofoil is also discussed.The model is proved to be useful for further understanding of the underlying phyiscs of such flow.

Applying extended intrinsic mean spin tensor in evolution algorithm for RANS modelling of turbulent rotating channel flow

We present a machine learning based method for RANS modeling in the rotating frame of reference(RFR).The extended intrinsic mean spin tensor(EIMST)is adopted in a novel expansion of the evolution algorithm,named multi-dimensional gene expression programming(MGEP).Based on DNS data,a constrain free model for Reynolds stress is created by considering system rotating.The anisotropy behavior of Reynolds stress is considered in the model,which is then for the first time applied for modeling turbulent flow inside a rotating channel.Compared with the traditional RANS model,the new model can predict the non-symmetric profile of Reynolds stress.Meanwhile,the Taylor-Gortler vortex is captured in our simulations with the new model.It is demonstrated that the application of EIMST in MGEP can be successfully adopted for RANS modeling in the RFR.