arXiv subjects

Balaji Jayaraman

Publications and source records attributed to Balaji Jayaraman.

11 recordsLinked to original sources

arXiv2026

FLARE: A Data-Efficient Surrogate for Predicting Displacement Fields in Directed Energy Deposition

Directed energy deposition (DED) produces complex thermo-mechanical responses that can lead to distortion and reduced dimensional accuracy of a manufactured part. Thermo-mechanical finite element simulations are widely used to estimate these effects, but their computational cost and the complexity of accurately capturing DED physics limit their use in design iteration and process optimization. This paper introduces FLARE (Field Prediction via Linear Affine Reconstruction in wEight-space), a data-efficient surrogate modeling framework for predicting post-cooling displacement fields in DED from geometric and process parameters. We develop a predefined-geometry DED simulation workflow using an open-source finite element framework and generate a dataset of simulations with varying geometry, laser power, and deposition velocity. Each simulation provides full-field displacement, stress, strain, and temperature data throughout the manufacturing process. FLARE encodes each simulation as an implicit neural field and regularizes the corresponding neural-network weights so that they follow the affine structure of the input parameter space. This enables prediction of unseen parameter combinations by reconstructing network weights through affine mixing of training examples. On this DED benchmark, the method shows improved accuracy compared to baseline methods in both in-distribution and extrapolation settings. Although the present study focuses on DED displacement prediction, the proposed affine weight-space reconstruction framework offers a promising approach for data-efficient surrogate modeling of physical fields.

cs.LG↗

arXiv2020

Modification of Turbulent Boundary Layer Coherent Structures with Drag Reducing Polymer Solution

The modification of dominant coherent structures that extend through the log-region of a drag reduced turbulent boundary layer is studied via examination of two-point correlations from time-resolved particle-image-velocimetry. Measurements were acquired in polymer oceans (uniform concentration) at drag reduction levels corresponding to the low drag reduction regime (< 40%), the high drag reduction (HDR) regime (>40%), and at an intermediate level (46%). The mean velocity profiles and two-point correlations were compared with that of water (Newtonian, DR = 0%). These results show that with increasing drag reduction the inclination of these dominant coherent structures decrease, their streamwise extent increases, and the fluctuations in the correlations are suppressed (especially at HDR). These observations are examined in comparison with the coherent structure literature (Newtonian and polymeric).

physics.flu-dyn↗

arXiv2019

Slope Dependent Turbulence over Two-dimensional Wavy Surfaces

Knowledge of turbulent flows over non-flat surfaces is of major practical interest in diverse applications. Significant work continues to be reported in the roughness regime at high Reynolds numbers where the cumulative effect of surface undulations on the averaged and integrated turbulence quantities is well documented. Even for such cases, the surface topology plays an important role for transitional roughness Reynolds numbers that is hard to characterize. In this work, we attempt to develop a bottom up understanding of the mechanisms underlying turbulence generation and transport, particularly within the region of the turbulent boundary layer (TBL) affected by the surface. We relate surface characteristics with turbulence generation mechanisms, Reynolds stress transport and the resulting drag increase. To this end, we perform a suite of direct numerical simulations of fully developed turbulent flow between two infinitely wide, two-dimensional sinusoidally wavy surfaces at a friction Reynolds number, $Re_τ=180$, with different mean surface slopes, $ζ$ (and fixed inner-scaled undulation height, $a^+=13$) corresponding to the `waviness' regime. The increase in wave slope enhances near surface turbulent mixing resulting in increased total drag, higher fraction of form drag, faster approach to isotropy and thereby, modulation of the buffer layer. The primary near-surface streamwise and vertical turbulence generation occurs in the leeward and windward side of the wave. In contrast, the spanwise variance is produced through pressure-rate-of-strain mechanism, primarily, along the windward side of the wave for flows with very little flow separation. We also observe significant dispersion effects in the production structure.

physics.flu-dyn↗

arXiv2019

Wind Estimation Using Quadcopter Motion: A Machine Learning Approach

In this article, we study the well known problem of wind estimation in atmospheric turbulence using small unmanned aerial systems (sUAS). We present a machine learning approach to wind velocity estimation based on quadcopter state measurements without a wind sensor. We accomplish this by training a long short-term memory (LSTM) neural network (NN) on roll and pitch angles and quadcopter position inputs with forcing wind velocities as the targets. The datasets are generated using a simulated quadcopter in turbulent wind fields. The trained neural network is deployed to estimate the turbulent winds as generated by the Dryden gust model as well as a realistic large eddy simulation (LES) of a near-neutral atmospheric boundary layer (ABL) over flat terrain. The resulting NN predictions are compared to a wind triangle approach that uses tilt angle as an approximation of airspeed. Results from this study indicate that the LSTM-NN based approach predicts lower errors in both the mean and variance of the local wind field as compared to the wind triangle approach. The work reported in this article demonstrates the potential of machine learning for sensor-less wind estimation and has strong implications to large-scale low-altitude atmospheric sensing using sUAS for environmental and autonomous navigation applications.

eess.SP↗

arXiv2019

Quad-rotor Flight Simulation in Realistic Atmospheric Conditions

In trajectory planning and control design for unmanned air vehicles, highly simplified models are typically used to represent the vehicle dynamics and the operating environment. The goal of this work is to perform real-time, but realistic flight simulations and trajectory planning for quad-copters in low altitude (<500m) atmospheric conditions. The aerodynamic model for rotor performance is adapted from blade element momentum theory and validated against experimental data. Large-eddy simulations of the atmospheric boundary layer are used to accurately represent the operating environment of unmanned air vehicles. A reduced-order version of the atmospheric boundary layer data as well as the popular Dryden model are used to assess the impact of accuracy of the wind field model on the predicted vehicle performance and trajectory. The wind model, aerodynamics and control modules are integrated into a six-degree-of-freedom flight simulation environment with a fully nonlinear flight controller. Simulations are performed for two representative flight paths, namely, straight and circular paths. Results for different wind models are compared and the impact of simplifying assumptions in representing rotor aerodynamics is discussed. The simulation framework and codes are open-sourced for use by the community.

physics.flu-dyn↗

arXiv2018

The Surprising Transition in Atmospheric Boundary Layer Turbulence Structure from Neutral to Moderately Convective Stability States and Mechanisms Underlying Large-scale Rolls

The vectoral wind structure of daytime atmospheric boundary layer (ABL) turbulence is strongly dependent on the balance between shear-driven turbulence production of horizontal fluctuations (driven by winds at the mesoscale), and buoyancy-driven turbulence production of vertical velocity fluctuations (driven by solar heating), characterized by the global instability state parameter $-z_i/L > 0$. In the fully shear-driven neutral limit $-z_i/L \rightarrow 0$, the surface layer is dominated by coherent streamwise-elongated concentrations of negative streamwise fluctuating velocity (low-speed streaks), while in the moderately convective state ($z_i/L \sim 10$ ) buoyancy generates streamwise-elongated thermal updraft sheets of concentrated vertical velocity fluctuations. Using large-eddy simulation (LES), we study the transition between the neutral and moderately convective states by quantifying correlations and integral scales as a function of $-z_i/L$ created by systematically increasing surface heat flux with fixed geostrophic wind. We discover a surprising sudden transition in ABL turbulence structure at $-z_i/L \approx 0.40$ with dramatic enhancement of streamwise coherence, particularly in the mixed layer, and a sudden change in ABL dynamic response to further increase in surface heating. In the supercritical ABL, continued increase in surface heat flux leads to a maximal coherence state at ($z_i/L \sim 1.0-1.5$) associated with helical large-scale roll structure and exceptionally coherent thermal updrafts, a process driven by two key dynamical effects: (a) a surprising continual increase in streamwise coherence of streamwise velocity fluctuations and shear-driven low-speed streaks and (b) increasing spatial correlation between the coherent low-speed streaks in the surface layer below and in the coherent thermal updrafts with in the mixed layer above.

physics.flu-dyn↗

arXiv2018

Characterization of a Canonical Helicopter Hub Wake

The current study investigates the long-age wake behind rotating helicopter hub models composed of geometrically simple, canonical bluff body shapes. The models consisted of a 4-arm rotor mounted on a shaft above a 2-arm (scissor) rotor with all the rotor arms having a rectangular cross section. The relative phase between the 2- and 4-arm rotors was either $0°$ (in-phase) or $45°$ (out-of-phase). The rotors were oriented at zero angle-of-attack and rotated at 30 Hz. Their wakes were measured with particle-image-velocimetry within a water tunnel at a hub diameter based Reynolds number of $820,000$ and an advance ratio of $0.2$. Mean profiles, fluctuating profiles and spectral analysis using time-series analysis as well as dynamic mode decomposition were used to characterize the wake and identify coherent structures associated with specific frequency content. The canonical geometry produced coherent structures that were consistent with previous results using more complex geometries. It was shown that the dominant structures (2 and 4 times per hub revolution) decay slowly and were not sensitive to the relative phase between the rotors. Conversely, the next strongest structure (6 times per hub revolution) was sensitive to the relative phase with almost no coherence observed for the in-phase model. This is strong evidence that the 6 per revolution content is a nonlinear interaction between the 2 and 4 revolution structures. This study demonstrates that the far wake region is dominated by the main rotor arms wake, the scissor rotor wake and interactions between these two features.

physics.flu-dyn↗

arXiv2018

Assessment of End-to-End and Sequential Data-driven Learning of Fluid Flows

In this work we explore the advantages of end-to-end learning of multilayer maps offered by feed forward neural-networks (FFNN) for learning and predicting dynamics from transient fluid flow data.While machine learning in general depends on data quality and quantity relative to the underlying dynamics of the system, it is important for a given learning architecture to make the most of this available information. To this end, we focus on data-driven problems where there is a need to predict over reasonable time into the future with limited data availability. Such function approximation or time series prediction is in contrast to many applications of machine learning such as pattern recognition and parameter estimation that leverage vast datasets. In this study, we interpret the suite of recently popular data-driven learning approaches that approximate the dynamics as Markov linear model in a higher-dimensional feature space as a multilayer architecture similar to neural networks.

physics.comp-ph↗

arXiv2018

On the Trajectory Dependence of Atmospheric Boundary Layer Turbulence Sensing using Small Unmanned Vehicles

Atmospheric turbulence, especially in the near-surface boundary layer is known to be under-sampled due to the need to capture a wide separation in length and time-scales and limitation in the number of sensors. Over the past decade, the use if Unmanned Aircraft Systems (UAS) technology is approaching ubiquitous proportions for a wide variety of applications, especially with the recent FAA relaxation of flying restrictions. From a geophysical sciences perspective, such technology would allow for sensing of large-scale atmospheric flows, particularly, atmospheric boundary layer (ABL) turbulence, air quality monitoring in urban settings where multitude of small, minimally-invasive and mobile sensors can drastically alter our ability to study such complex phenomena. Currently available observational data of atmospheric boundary layer physics is so sparse and infrequent which significantly limits analysis. With the quantity and resolution of the data that can be measured using a swarm of UAS, three-dimensional reconstruction and deduction of coherent structures in ABL turbulence may be feasible. However, key challenges remain in the form of identifying optimal trajectories to fly the UAS to obtain the relevant quantifications of the turbulence, interpretation of the sensor data from mobile sensors and understanding how representative are the sparse measurements of the overall turbulent boundary layer. This leads to many fundamentally interesting questions if which a few are explored in this article.

physics.flu-dyn↗

arXiv2018

Interplay of Sensor Quantity, Placement and System Dimensionality on Energy Sparse Reconstruction of Fluid Flows

Reconstruction of fine-scale information from sparse data is relevant to many practical fluid dynamic applications where the sensing is typically sparse. Fluid flows in an ideal sense are manifestations of nonlinear multiscale PDE dynamical systems with inherent scale separation that impact the system dimensionality. There is a common need to analyze the data from flow measurements or high-fidelity computations for stability characteristics, identification of coherent structures and develop evolutionary models for real-time data-driven control. Given that sparse reconstruction is inherently an ill-posed problem, the most successful approaches require the knowledge of the underlying basis space spanning the manifold in which the system resides. In this study, we employ an approach that learns basis from singular value decomposition (SVD) of training data to reconstruct sparsely sensed information at randomly sampled locations. This allows us to leverage energy sparsity with l2 minimization instead of the more expensive, sparsity promoting l1 minimization. Further, for unknown flow systems where only global operating parameters such as Reynolds (Re) number and raw data are available, it is often not clear what the optimal number of sensors and their placement for near-exact reconstruction needs to be. In this effort, we explore the interplay of data sparsity, sparsity of the underlying flow system and sensor placement on energy sparse reconstruction performance enabled by data- driven SVD basis. To this end, we investigate sparse convolution-based reconstruction performance by characterizing operational bounds for canonical laminar cylinder wake flows in both limit-cycle and transient regimes.

physics.comp-ph↗

arXiv2018

Sparse Convolution-based Markov Models for Nonlinear Fluid Flows

Data-driven modeling for nonlinear fluid flows using sparse convolution-based mapping into a feature space where the dynamics are Markov linear is explored in this article. The underlying principle of low-order models for fluid systems is identifying convolutions to a feature space where the system evolution (a) is simpler and efficient to model and (b) the predictions can be reconstructed accurately through deconvolution. Such methods are useful when real-time models from sensor data are needed for online decision making. The Markov linear approximation is popular as it allows us to leverage the vast linear systems machinery. Examples include the Koopman operator approximation techniques and evolutionary kernel methods in machine learning. The success of these models in approximating nonlinear dynamical systems is tied to the effectiveness of the convolution map in accomplishing both (a) and (b) mentioned above. To assess this, we perform in-depth study of two classes of sparse convolution operators: (i) a pure data-driven POD-convolution that uses left singular vectors of the data snapshots - a staple of Koopman approximation methods and (ii) a sparse Gaussian Process (sGP) convolution that combines sparse sampling with a Gaussian kernel embedding an implicit feature map to an inner product reproducing kernel Hilbert space.

physics.flu-dyn↗