Search

Creating new air combat tactics and discovering novel maneuvers can require numerous hours of expert pilots' time. Additionally, for each different combat scenario, the same strategies may not work since small changes in equipment performance may drastically change the air combat outcome. For this reason, we created a reinforcement learning environment to help investigate potential air combat tactics in the field of beyond-visual-range (BVR) air combat: the BVR Gym. This type of air combat is important since long-range missiles are often the first weapon to be used in aerial combat. Some existing environments provide high-fidelity simulations but are either not open source or are not adapted to the BVR air combat domain. Other environments are open source but use less accurate simulation models. Our work provides a high-fidelity environment based on the open-source flight dynamics simulator JSBSim and is adapted to the BVR air combat domain. This article describes the building blocks of the environment and some use cases., Comment: 8 pages, 8 figures

As the effective range of air-to-air missiles increases, it becomes harder for human operators to maintain the situational awareness needed to keep a UAV safe. In this work, we propose a decision support tool to help UAV operators in Beyond Visual Range (BVR) air combat scenarios assess the risks of different options and make decisions based on those. Earlier work focused on the threat posed by a single missile, and in this work, we extend the ideas to several missile threats. The proposed method uses Deep Neural Networks (DNN) to learn from high-fidelity simulations to provide the operator with an outcome estimate for a set of different strategies. Our results demonstrate that the proposed system can manage multiple incoming missiles, evaluate a family of options, and recommend the least risky course of action.

We further elaborate on the generalized formulation for cubic equation of state proposed by Cismondi and Mollerup [Fluid Phase Equilib. 232 (2005)]. With this formulation all well-known cubic equations of state can be described with a certain pair of values, which allows for a generic implementation of different equations of state. Based on this generalized formulation, we derive a complete thermodynamic model for computational fluid dynamics (CFD) simulations by providing the resulting correlations for all required thermodynamic properties. For the transport properties, we employ the Chung correlations. Our generic implementation includes the often used equations of state Soave-Redlich-Kwong and Peng-Robinson and the Redlich-Kwong-Peng-Robinson (RKPR) equation of state. The first two assume a universal compressibility factor and are therefore only suitable for fluids with a matching critical compressibility. The Redlich-Kwong-Peng-Robinson overcomes this limitation by considering the equation of state parameter as function of the critical compressibility. We compare the resulting thermodynamic modeling for the three equations of state for selected fluids with each other and CoolProp reference data. As supplementary material to this paper, we provide a Python tool called real gas thermodynamic python library (realtpl). This tool can be used to evaluate and compare the results for a wide range of different fluids. Additionally, we also provide the implementation of the generalized form in OpenFOAM.

We study the segregation of emulsions in decaying turbulence using direct numerical simulations (DNS) in combination with the volume of fluid method (VOF). To this end, we generate emulsions in forced homogeneous isotropic turbulence and then turn the forcing off and activate gravitational acceleration. This allows us to study the segregation process in decaying turbulence and under gravity. We consider non-iso-density emulsions, where the dispersed phase is the lighter one. The segregation process is driven by both the minimization of the potential energy achieved by the sinking of the heavier phase, as well as the minimization of the surface energy achieved by coalescence. To study these two processes and their impacts on the segregation progress in detail, we consider different buoyancy forces and surface tension coefficients in our investigation, resulting in five different configurations. The surface tension coefficient also alters the droplet size distribution of the emulsion. Using the three-dimensional simulation results and the monitored data, we analyze the driving mechanisms and their impact on the segregation progress in detail. We propose a dimensionless number that reflects the energy release dominating the segregation. Moreover, we evaluate the time required for the rise of the lighter phase and study correlations with the varied parameters gravitational acceleration and surface tension coefficient.

A new premixed turbulent combustion model is proposed. It is based on one-dimensional (1D) filtering of density times progress variable and of the reaction source term of laminar premixed flame profiles using a filter kernel which reflects the variation of the slicing area of planar flame fronts as they move through multidimensional filter volumes. It is shown that these multidimensional effects qualitatively change the relation between the filtered reaction source term and the Favre-filtered reaction progress variable compared to 1D filtering, particularly at large filter widths. Analytical results for the filtered quantities are achieved by approximating density times progress variable and reaction source term by suitable Ansatz functions. Filtered data from Direct Numerical Simulations (DNS) of statistically planar turbulent premixed flames at different free stream turbulence levels and heat release parameters is used to develop and validate the model. Two wrinkling factor models as function of filter width and subgrid turbulence level are proposed. For small filter widths up to two times the laminar flame thickness, minor effects from subgrid flame folding are observed. For larger filters, the filtered reaction source term rises linearly with filter width at a rate which increases with subgrid turbulence level. The modelled reaction source term as function of Favre averaged progress variable and filter width shows excellent agreement with filtered DNS data for all investigated free stream turbulence levels, filter widths and heat release parameters., Comment: 33 pages, 27 figures, submitted to Combustion and Flame

While large training datasets generally offer improvement in model performance, the training process becomes computationally expensive and time consuming. Distributed learning is a common strategy to reduce the overall training time by exploiting multiple computing devices. Recently, it has been observed in the single machine setting that overparametrization is essential for benign overfitting in ridgeless regression in Hilbert spaces. We show that in this regime, data splitting has a regularizing effect, hence improving statistical performance and computational complexity at the same time. We further provide a unified framework that allows to analyze both the finite and infinite dimensional setting. We numerically demonstrate the effect of different model parameters.

In the semiclassical limit h to 0, we analyze a class of self-adjoint Schr\"odinger operators H_h = h^2 L + h W + V id_E acting on sections of a vector bundle E over an oriented Riemannian manifold M where L is a Laplace type operator, W is an endomorphism field and the potential energy V has non-degenerate minima at a finite number of points m_1, ... m_r in M, called potential wells. Using quasimodes of WKB-type near m_j for eigenfunctions associated with the low lying eigenvalues of H_h, we analyze the tunneling effect, i.e. the splitting between low lying eigenvalues, which e.g. arises in certain symmetric configurations. Technically, we treat the coupling between different potential wells by an interaction matrix and we consider the case of a single minimal geodesic (with respect to the associated Agmon metric) connecting two potential wells and the case of a submanifold of minimal geodesics of dimension l + 1. This dimension l determines the polynomial prefactor for exponentially small eigenvalue splitting., Comment: 23 pages