Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “APPROXIMATIONS”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 847 records · Page 47
All sourcesNASA NTRSDOE OSTI
DOE OSTI2020

Adaptive Activation Functions Accelerate Convergence in Deep and Physics-informed Neural Networks

We employ adaptive activation functions for regression in deep and physics-informed neural networks (PINNs) to approximate smooth and discontinuous functions as well as solutions of linear and nonlinear partial differential equations. In particular, we solve the nonlinear Klein-Gordon equation, which has smooth solutions, the nonlinear Burgers equation, which can admit high gradient solutions, and the Helmholtz equation. We introduce a scalable hyper-parameter in the activation function, which can be optimized to achieve best performance of the network as it changes dynamically the topology of the loss function involved in the optimization process. The adaptive activation function has better learning capabilities than the traditional one (fixed activation) as it improves greatly the convergence rate, especially at early training, as well as the solution accuracy. To better understand the learning process, we plot the neural network solution in the frequency domain to examine how the network captures successively different frequency bands present in the solution. We consider both forward problems, where the approximate solutions are obtained, as well as inverse problems, where parameters involved in the governing equation are identified. Our simulation results show that the proposed method is a very simple and effective approach to increase the efficiency, robustness and accuracy of the neural network approximation of nonlinear functions as well as solutions of partial differential equations, especially for forward problems. We theoretically prove that in the proposed method, gradient descent algorithms are not attracted to suboptimal critical points or local minima.

machine leaning, Bad minima, Inverse problems, Phy↗
DOE OSTI2020

Multigrid deflation for Lattice QCD

Computing the trace of the inverse of large matrices is typically addressed through statistical methods. Deflating out the lowest eigenvectors or singular vectors of the matrix reduces the variance of the trace estimator. This work summarizes our efforts to reduce the computational cost of computing the deflation space while achieving the desired variance reduction for Lattice QCD applications. Previous efforts computed the lower part of the singular spectrum of the Dirac operator by using an eigensolver preconditioned with a multigrid linear system solver. Despite the improvement in performance in those applications, as the problem size grows the runtime and storage demands of this approach will eventually dominate the stochastic estimation part of the computation. In this work, we propose to compute the deflation space in one of the following two ways. First, by using an inexact eigensolver on the Hermitian, but maximally indefinite, operator. Second, by exploiting the fact that the multigrid prolongator for this operator is rich in components toward the lower part of the singular spectrum. We show experimentally that the inexact eigensolver can approximate the lower part of the spectrum even for ill-conditioned operators. Also, the deflation based on the multigrid prolongator is more efficient to compute and apply, and, despite its limited ability to approximate the fine level spectrum, it obtains similar variance reduction on the trace estimator as deflating with approximate eigenvectors from the fine level operator.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2020

An arbitrarily high-order three-dimensional Cartesian-grid method for reconstructing interfaces from volume fraction fields

Here Tthis work describes a newly developed, arbitrarily high-order Cartesian-grid method for reconstructing material interfaces from a volume fraction field. The method begins by identifying all of the grid cells in the volume fraction field that are intersected by the interface and need to be approximated by the reconstruction scheme. Finite-differences are used to calculate the gradient of the volume fraction field and provide an estimate of the surface normal in all of the interfacial grid cells. Groups of connected grid cells are then identified which all have the same dominant component of the normal vector. This grouping by orientation determines the proper dependent variable to use in the surface reconstruction (e.g. for a 2D curve, this step determines if the surface will be approximated by a function of x or y). A cumulative integral over the surface is constructed and fit using b-splines for two-dimensional problems or tensor-product b-splines for three-dimensional problems. This construction allows for the interface to be recovered through application of the second fundamental theorem of calculus. Fitting the cumulative integral with $\mathscr{N}$ th-order b-splines (or tensor-product b-splines) yields an ($\mathscr{N}$-1) th-order convergence rate of the interface shape. Differentiation of the b-spline interface function(s) allows for the high-order approximation of the normal vector and curvature to be obtained directly anywhere along b-spline. Together, the proposed reconstruction technique can achieve arbitrarily high mesh convergence rates. Validation tests are presented with mesh convergence rates ranging from fourth- to tenth-order.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Multi-fidelity Bayesian neural networks: Algorithms and applications

Here we propose a new class of Bayesian neural networks (BNNs) that can be trained using noisy data of variable fidelity, and we apply them to learn function approximations as well as to solve inverse problems based on partial differential equations (PDEs). These multi-fidelity BNNs consist of three neural networks: The first is a fully connected neural network, which is trained following the maximum a posteriori probability (MAP) method to fit the low-fidelity data; the second is a Bayesian neural network employed to capture the cross-correlation with uncertainty quantification between the low- and high-fidelity data; and the last one is the physics-informed neural network, which encodes the physical laws described by PDEs. For the training of the last two neural networks, we first employ the mean-field variational inference (VI) to maximize the evidence lower bound (ELBO) to obtain informative prior distributions for the hyperparameters in the BNNs, and subsequently we use the Hamiltonian Monte Carlo (HMC) method to estimate accurately the posterior distributions for the corresponding hyperparameters. We demonstrate the accuracy of the present method using synthetic data as well as real measurements. Specifically, we first approximate a one- and four-dimensional function, and then infer the reaction rates in one- and two-dimensional diffusion-reaction systems. Moreover, we infer the sea surface temperature (SST) in the Massachusetts and Cape Cod Bays using satellite images and in-situ measurements. Taken together, our results demonstrate that the present method can capture both linear and nonlinear correlation between the low- and high-fidelity data adaptively, identify unknown parameters in PDEs, and quantify uncertainties in predictions, given a few scattered noisy high-fidelity data. Finally, we demonstrate that we can effectively and efficiently reduce the uncertainties and hence enhance the prediction accuracy with an active learning approach, using as examples a specific one-dimensional function approximation and an inverse PDE problem.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Physics-informed neural networks for solving forward and inverse flow problems via the Boltzmann-BGK formulation

We report the Boltzmann equation with the Bhatnagar-Gross-Krook collision model (Boltzmann-BGK equation) has been employed to describe multiscale flows, i.e., from the hydrodynamic limit to free molecular flow. In this study we employ physics-informed neural networks (PINNs) to solve forward and inverse problems via the Boltzmann-BGK formulation (PINN-BGK), enabling PINNs to model flows in both the continuum and rarefied regimes. In particular, the PINN-BGK is composed of three sub-networks, i.e., the first for approximating the equilibrium distribution function, the second for approximating the non-equilibrium distribution function, and the third one for encoding the Boltzmann-BGK equation as well as the corresponding boundary/initial conditions. By minimizing the residuals of the governing equations and the mismatch between the predicted and provided boundary/initial conditions, we can approximate the Boltzmann-BGK equation for both continuous and rarefied flows. For forward problems, the PINN-BGK is utilized to solve various benchmark flows given boundary/initial conditions, e.g. Kovasznay flow, Taylor-Green flow, cavity flow, and micro Couette flow for Knudsen number up to 5. For inverse problems, we focus on rarefied flows in which accurate boundary conditions are difficult to obtain. We employ the PINN-BGK to infer the flow field in the entire computational domain given a limited number of interior scattered measurements on the velocity without using the (unknown) boundary conditions. Results for the two-dimensional micro Couette and micro cavity flows with Knudsen numbers ranging from 0.1 to 10 indicate that the PINN-BGK can infer the velocity field in the entire domain with good accuracy. Finally, we also present some results on using transfer learning to accelerate the training process. Specifically, we can obtain a three-fold speedup comparted to the standard training process (e.g., Adam plus L-BFG-B) for the two-dimensional flow problems considered in our work.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2022

Efficient hybrid explicit-implicit learning for multiscale problems

Splitting method is a powerful method to handle application problems by splitting physics, scales, domain, and so on. Many splitting algorithms have been designed for efficient temporal discretization. Here, in this paper, our goal is to use temporal splitting concepts in designing machine learning algorithms and, at the same time, help splitting algorithms by incorporating data and speeding them up. We propose a machine learning assisted splitting scheme which improves the efficiency of the scheme meanwhile preserves the accuracy. We consider a recently introduced multiscale splitting algorithms, where the multiscale problem is solved on a coarse grid. To approximate the dynamics, only a few degrees of freedom are solved implicitly, while others explicitly. This splitting concept allows identifying degrees of freedom that need implicit treatment. In this paper, we use this splitting concept in machine learning and propose several strategies. First, the implicit part of the solution can be learned as it is more difficult to solve, while the explicit part can be computed. This provides a speed-up and data incorporation for splitting approaches. Secondly, one can design a hybrid neural network architecture because handling explicit parts requires much fewer communications among neurons and can be done efficiently. Thirdly, one can solve the coarse grid component via PDEs or other approximation methods and construct simpler neural networks for the explicit part of the solutions. We discuss these options and implement one of them by interpreting it as a machine translation task. This interpretation of the splitting scheme successfully enables us using the Transformer since it can perform model reduction for multiple time series and learn the connection between them. We also find that the splitting scheme is a great platform to predict the coarse solution with insufficient information of the target model: the target problem is partially given and we need to solve it through a known problem which approximates the target. Our machine learning model can incorporate and encode the given information from two different problems and then solve the target problems. We conduct four numerical examples and the results show that our method is stable and accurate.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

A new operator-split Compton scattering method

In high-energy-density physics, when the temperatures are high (keV-range) and/or the densities are low, Compton scattering becomes the main energy transfer mechanism between the radiation field and the material. The interaction between the radiation field and the material is non-linear in the radiation field – due to induced processes caused by photons being bosons – and in the material temperature. Many existing codes use a Compton-Fokker-Planck (CFP) approximation – which is inaccurate for high photon energies or material temperatures – or use linear approximations within a time-step. Here, we have implemented a new Compton-heating method called OS-FBC (Operator-Split Full-Boltzmann Compton) that solves the full, nonlinear Compton-heating system with few approximations.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Data-driven models of nonautonomous systems

Nonautonomous dynamical systems are characterized by time-dependent inputs, which complicates the discovery of predictive models describing the spatiotemporal evolution of the state variables of quantities of interest from their temporal snapshots. When dynamic mode decomposition (DMD) is used to infer a linear model, this difficulty manifests itself in the need to approximate the time-dependent Koopman operators. Our approach is to approximate the original nonautonomous system with a modified system derived via a local parameterization of the time-dependent inputs. The modified system comprises a sequence of local parametric systems, which are subsequently approximated by a parametric surrogate model using the DRIPS (dimension reduction and interpolation in parameter space) framework. The offline step of DRIPS relies on DMD to build a linear surrogate model, endowed with reduced-order bases for the observables mapped from training data. The online step interpolates on suitable manifolds to construct a sequence of iterative parametric surrogate models; the target/test parameter points on these manifolds are specified by a local parameterization of the test time-dependent inputs. Here, we use numerical experimentation to demonstrate the robustness of our method and compare its performance with that of deep neural networks.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Assessing feasible H2–CO2 sources in the US as Feedstocks for Sustainable Aviation Fuel Precursors: Acetic Acid and Ethanol Production via Hydrogenotrophic Pathways

The environmental impact of carbon dioxide emissions is significant, and research is focused on mitigating these emissions and developing eco-friendly technologies in line with green chemistry principles. Waste-to-energy technologies play a crucial role in converting waste into renewable energy and valuable biofuels and bioproducts. This study specifically explores the utilization of waste gas emissions, particularly carbon dioxide, from various sources in the United States for the production of sustainable aviation fuel (SAF) precursors, such as ethanol and acetic acid. The study categorizes and quantifies the volumes of carbon dioxide emissions into three types: non-biogenic, biogenic, and biogenic emissions from ethanol production facilities. Stoichiometric calculations are applied to compare the amounts of carbon dioxide from each category with the available hydrogen production capacity, determining if sufficient hydrogen is present for converting carbon dioxide into SAF precursors. The study reveals two key findings. Firstly, there is a significant reserve of carbon dioxide, approximately 1648 million metric tons per year (MMTy), combining all three categories, which would require a substantial increase of approximately 35–40 times in the existing hydrogen production capacity of 4.988 MMTy. This increased hydrogen production has the potential to yield approximately 1067.82 MMTy of acetic acid and 189.19 MMTy of ethanol annually. Secondly, upon analyzing the quality and application of the three sources of carbon dioxide with the currently available hydrogen production capacity, it is found that biogenic carbon dioxide from ethanol plants is the most suitable choice for immediate production of SAF precursors. This would theoretically result in an annual production of 1.36 MMTy of ethanol and 1.772 MMTy of acetic acid. The other two sources of carbon dioxide can be considered potential reserves for future utilization when additional hydrogen production facilities are established. The study provides a foundation for assessing the aggregation potential required for acetic acid and ethanol production. By optimizing the use of waste gases as raw materials, the study not only enables the production of SAF precursors but also contributes to the passive reduction of greenhouse gas emissions.

Carbon Dioxide↗
DOE OSTI2021

Conditional Point Sampling: A Monte Carlo Method for Radiation Transport in Stochastic Media.

Current methods for stochastic media transport are either computationally expensive or, by nature, approximate. Moreover, none of the well-developed, benchmarked approximate methods can compute the variance caused by the stochastic mixing, a quantity especially important to safety calculations. Therefore, we derive and apply a new conditional probability function (CPF) for use in the recently developed stochastic media transport algorithm Conditional Point Sampling (CoPS), which 1) leverages the full intra-particle memory of CoPS to yield errorless computation of stochastic media outputs in 1D, binary, Markovian-mixed media, and 2) leverages the full inter-particle memory of CoPS and the recently developed Embedded Variance Deconvolution method to yield computation of the variance in transport outputs caused by stochastic material mixing. Numerical results demonstrate errorless stochastic media transport as compared to reference benchmark solutions with the new CPF for this class of stochastic mixing as well as the ability to compute the variance caused by the stochastic mixing via CoPS. Using previously derived, non-errorless CPFs, CoPS is further found to be more accurate than the atomic mix approximation, Chord Length Sampling (CLS), and most of memory-enhanced versions of CLS surveyed. In addition, we study the compounding behavior of CPF error as a function of cohort size (where a cohort is a group of histories that share intra-particle memory) and recommend that small cohorts be used when computing the variance in transport outputs caused by stochastic mixing.

61 RADIATION PROTECTION AND DOSIMETRY↗
DOE OSTI2023

Modeling and characterizing the elastodynamic response of octet-truss lattice structures using resonance techniques

Characterization of additively manufactured materials and structures is an ongoing effort for the advanced manufacturing community. This article will investigate an approach using resonant spectral ultrasound (RUS) to measure the effective elastic constants of an Octet Truss lattice and apply the results to continuum bases models representing the lattice regions in different structures. The study is focused on simple lattices structures fabricated from Ti5553 using a laser powder bed fusion process (LBPF). Solid and lattice samples are measured to determine RUS estimates for the elastic properties of the bulk material and the effective elastic properties of the lattice structure. The estimated elastic properties are then incorporated into 3d finite element models using a continuum approximation of the physical AM parts. Comparisons show good agreement between the experimentally measured eigen frequencies of the Ti5553 LBPF parts and the eigen frequencies calculated using the continuum approximations based on the effective RUS elastic properties. The current results suggest there is additional physics and geometrical effects that are accounted for in the RUS continuum approximation of the lattice that are not captured in the full 3d finite element model of the parts utilizing only the base material properties.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Distinguishing fissile uranium isotopes using an active well neutron coincidence counter

Proposed thorium-based nuclear fuel cycles are likely to require quantification and verification of 233 U within nuclear material. Because of their similar fission cross sections, active neutron nondestructive assay (NDA) systems may respond similarly to 233 U and 235 U. Traditional safeguards equipment has been optimized for 235 U and 238 U quantification associated with conventional uranium/plutonium fuel cycles and may not be directly applicable to 233 U quantification when mixed with other actinides. This work used models of the large volume active well coincidence counter (LV-AWCC) at Oak Ridge National Laboratory to evaluate the performance of this neutron NDA system to differentiate fissile uranium isotopes. The models were developed to simulate NDA system performance in response to a number of triangular radiation signature training device sources within the central cavity or well. This work predicted that the LV-AWCC can effectively differentiate 233 U from 235 U in certain modes of operation. In active mode, the LV-AWCC with the cadmium liner results in different doubles count rates between the fissile isotopes for a given fissile uranium mass. Without the cadmium liner, the uranium isotopes provide a statistically indistinguishable doubles count rate response for the fissile masses considered in this work (up to approximately 150 g). The cadmium liner serves to harden the neutron interrogation spectrum, which better exploits the notable difference in the 233 U and 235 U fission cross sections at approximately 1 eV. In passive mode, the two fissile isotopes exhibit different doubles and singles count rates regardless of liner presence because the passive source strength of 233 U is approximately 2 orders of magnitude stronger than that of 235 U due to the shorter half-life and correspondingly higher (α, n) yield. We conclude that using neutron interrogation in the LV-AWCC, two measurements are needed to quantify 233 U content in mixed uranium items. The first measurement is used to determine the total fissile uranium mass using a mode that cannot distinguish fissile isotopes (i.e., where a similar response is observed for both fissile uranium isotopes such as active doubles without cadmium or using a thermal neutron interrogation source). In conclusion, the second measurement is used to determine the 233 U content by using a differentiating technique (e.g., passive doubles, passive doubles to singles ratio, active doubles with cadmium).

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2025

Simulation and measurement of the dynamic strain response in a prototypical spallation target

The Second Target Station (STS) at the Spallation Neutron Source (SNS) will address emerging scientific challenges by providing a source of intense cold neutrons to instruments optimized for this source. The STS target will use rotating tungsten blocks and will receive 1.3 GeV proton beam pulses from the SNS accelerator at a repetition rate of 15 Hz. The facility life is planned for 40 years, and each target assembly life is expected to be approximately 10 years. An accurate strain prediction is then critical for fatigue life assessment of STS target blocks because they will be subject to approximately 10 8 beam pulses per lifetime. As an R&D activity, the Los Alamos Neutron Science Center (LANSCE) Weapons Neutron Research (WNR) Target 2 (Blue Room) facility was used to test the strain response of prototypical target blocks to the thermal shock of a proton pulse. The blue room was well suited for a pulsed proton beam impact test of subscale STS target blocks; the 800 MeV proton energy is approximately 60 % of the 1.3 GeV proton energy expected from the SNS accelerator to the STS. The LANSCE Proton Storage Ring (PSR) and SNS are both short-pulse proton beam sources with nominal pulse widths of 250 ns and 661 ns, respectively, so the energy deposition in the target occurs in <1 μs pulse duration. Strain measurements on the outer surface of three target blocks (bare tungsten, tantalum-clad tungsten, niobium-clad tungsten) were recorded for comparison against neutronics and structural simulations. In conclusion, this experiment and the supporting simulations satisfied the following primary research goals for the STS target.

Dynamic strain response↗
DOE OSTI2022

Assessments of epistemic uncertainty using Gaussian stochastic weight averaging for fluid-flow regression

Here, we use Gaussian stochastic weight averaging (SWAG) to assess the epistemic uncertainty associated with neural-network-based function approximation relevant to fluid flows. SWAG approximates a posterior Gaussian distribution of each weight, given training data, and a constant learning rate. Having access to this distribution, it is able to create multiple models with various combinations of sampled weights, which can be used to obtain ensemble predictions. The average of such an ensemble can be regarded as the 'mean estimation', whereas its standard deviation can be used to construct 'confidence intervals', which enable us to perform uncertainty quantification (UQ) with regard to the training process of neural networks. We utilize representative neural-network-based function approximation tasks for the following cases: (i) a two-dimensional circular-cylinder wake; (ii) the DayMET dataset (maximum daily temperature in North America); (iii) a three-dimensional square-cylinder wake; and (iv) urban flow, to assess the generalizability of the present idea for a wide range of complex datasets. SWAG-based UQ can be applied regardless of the network architecture, and therefore, we demonstrate the applicability of the method for two types of neural networks: (i) global field reconstruction from sparse sensors by combining convolutional neural network (CNN) and multi-layer perceptron (MLP); and (ii) far-field state estimation from sectional data with two-dimensional CNN. We find that SWAG can obtain physically-interpretable confidence-interval estimates from the perspective of epistemic uncertainty. This capability supports its use for a wide range of problems in science and engineering.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Solid-state polymer magnesium supercapacitor

Here, the present article focuses on the development of a highly ion-conductive, solvent-free, solid-state polymer electrolyte membrane (PEM) via photopolymerization of polyethylene glycol diacrylate (PEGDA) network from its homogeneous melt mixtures containing succinonitrile (SCN) plasticizer and magnesium bis(trifluoromethane sulfonyl) imide Mg(TFSI) 2 salt. The above solid-state Mg-PEM exhibits a Helmholtz electric double-layer capacitor (EDLC) behavior in its supercapacitive symmetric carbonaceous electrode configuration. The electrochemical stability for this PEM membrane (20/40/40 PEGDA/SCN/Mg (TFSI) 2 ) was found to be approximately 3 V from linear sweep voltammetry with a specific capacitance of about 44 F/g from cyclic voltammetry and an energy density of approximately 17 Wh/kg in the 0–2 V range from the constant current density (CCD) experiment. Of particular interest is its excellent capacity retention of over 11,200 cycles, thus tested with the Coulombic efficiency of over 87%. Moreover, the energy density after extensive cycling for 11,200 has increased to approximately 56 Wh/kg at 10 mV/s in the potential range of 0–3 V relative to 17 Wh/kg in the potential range of 0–2 V at the same scan rate, attesting the excellent electrochemical stability and long life of the present Mg-PEM supercapacitor.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2022

Optimizing lane reversals in transportation networks to reduce traffic congestion: A global optimization approach

This paper studies how to reduce the overall travel time of commuters in a transportation network by reversing the direction of some lanes in the network using a macroscopic network-wide perspective. Similar to the Network Design Problem, the lane reversal problem has been shown to be NP-hard given the dependence of the users’ route selection on the lane direction decision. Herein, we propose and compare three efficient methods to solve the routing and lane reversal problem jointly. First, we introduce an alternating method that decouples the routing and lane assignment problems. Second, we propose a Frank–Wolfe method that jointly takes gradient steps to adjust both the lane assignment and routing decisions. Third, we propose a convex approximation method that uses a threshold-based approach to convexify the joint routing and lane reversal objective. The convex approximation method is advantageous since it finds a global optimum solution for the approximated problem and it enables the possibility to include linear constraints. Using this method, we extend the main formulation to be able to limit a maximum number of reversed lanes, as well as to incorporate multiple origin–destination (OD) patterns. We test the proposed methods in a case study using the transportation network of Eastern Massachusetts where our results indicate an overall reduction in travel times of 4.7% by selecting the best 15 reversals. Moreover, using a small test network, we investigate the performance of the lane reversal strategies as a function of the OD demand symmetry. As expected, we observe that when the OD demand is very asymmetric (e.g., for a single OD pair, evacuations, large events), the reduction in travel times is larger than the symmetric case, reaching travel time reductions of 60%.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2024

An extension of Thwaites’ method for turbulent boundary layers

Thwaites developed an approximate method for determining the evolution of laminar boundary layers. The approximation follows from an assumption that the growth of a laminar boundary layer in the presence of pressure gradients could be parameterized solely as a function of the Holstein–Bohlen flow parameter, thus reducing the von Kármán momentum integral to a first-order ordinary differential equation. This method is useful for the analysis of laminar flows, and in computational potential flow solvers to account for the viscous effects. In this work, an approximate method for determining the momentum thickness of a two-dimensional, turbulent boundary layer is proposed following Thwaites’ work. It is shown that the method provides good estimates of the momentum thickness for multiple boundary layers, including both favourable and adverse pressure gradient effects, up to the point of separation. In the limit of high Reynolds numbers, it is possible to derive a criterion for the onset of separation from the proposed model, which is shown to be in agreement with prior empirical observations. The sensitivity of the separation location with respect to upstream perturbations is also analysed through this model for the NASA/Boeing speed bump and the transonic Bachalo–Johnson bump.

17 WIND ENERGY↗