arXiv subjects

Mehrshad Gholami Anjiraki

Publications and source records attributed to Mehrshad Gholami Anjiraki.

7 recordsLinked to original sources

arXiv2026

A novel machine learning framework based on proper orthogonal decomposition for turbine wake flow predictions in natural waterways

We developed a machine learning (ML) surrogate trained on instantaneous large eddy simulation (LES) data filtered using proper orthogonal decomposition (POD) to predict mean flow fields in natural rivers with complex geometries. The surrogate enables efficient prediction of mean flow fields downstream of marine hydrokinetic (MHK) turbine arrays at a fraction of the computational cost of high fidelity simulations, facilitating turbine layout optimization in natural waterways where bathymetry, turbulence, and turbine interactions influence the available energy resource. We demonstrate the capabilities of the model at a real world tidal river site under ebb tide conditions. Complex bed bathymetry and planform geometry generate substantial background turbulence that can obscure turbine wake signatures. POD is employed to extract relevant flow structures, improving the ability of the ML model to learn wake dynamics. First, LES is performed at the tidal site to generate training data. POD is then applied to decompose the instantaneous LES fields into spatial modes. Selected modes are used to train a three dimensional convolutional encoder decoder network to predict time averaged velocity fields. This selective filtering reduces the influence of dominant background fluctuations while preserving information relevant to turbine wakes. At equivalent network size and training cost, the POD filtered surrogate reduces the mean squared error by three orders of magnitude compared with the same network trained on unfiltered LES data.

physics.flu-dyn↗

arXiv2026

Computational and experimental study of multi-grain-size sediment transport and sorting process

Successful prediction of the interaction between turbulent flows and river ecosystems is essential for flood management, development of mitigation strategies, and infrastructure protection. Natural riverbeds consist of bed materials of multiple grain sizes that interact with turbulence differently, leading to spatial classification of particle sizes. Such particle sorting processes can only be captured by models capable of handling multiple grain sizes. We present a numerical framework that couples large-eddy simulation with a multi-grain-size sediment transport module to track multiple grain size classes, their local fractions within the bed-load layer, and bed evolution. A laboratory experiment was conducted in a flume with multi-size bed material to examine sediment particle redistribution and sorting. Measured sediment grain size distributions and bed elevations were used to validate the numerical framework. The results of the multi-grain-size sediment transport model for bed deformation were then compared against those of a uniform sediment transport model, demonstrating that the former reduces the prediction errors relative to the experimental measurements from 37.98 to 3.83 percent for maximum scour depth and from 48.35 to 11.54 percent for maximum deposition height. The validated numerical framework was also applied to large-scale meandering rivers with multi-grain bed material, simulating their bed deformation and sediment sorting. The simulation results revealed the formation of particle sorting patterns in meandering rivers, with coarser and finer sediment particles accumulating near the erosional depression (scour) and depositional (point bar) regions, respectively. These findings support the coupled framework's ability to predict sediment particle sorting and highlight the importance of representing bed material heterogeneity in numerical modeling of meandering rivers.

physics.flu-dyn↗

arXiv2025

Computational study of vertical-axis MHK turbines using a coupled flow-sediment-turbine modeling approach

We present a coupled large-eddy simulation (LES) and bed morphodynamics study to investigate the influence of sediment dynamics on the performance of a utility-scale marine hydrokinetic vertical-axis turbine (VAT) parametrized by an actuator surface model. By resolving the two-way interactions between turbine-induced flow structures and bed evolution, the study offers insights into the environmental implications of VAT deployment in riverine and marine settings. A range of tip speed ratios is examined to evaluate wake recovery, power production, and bed response. The actuator surface method (ASM) is implemented to capture the effects of rotating vertical blades on the flow, while the immersed boundary method accounts for fluid interactions with the channel walls and sediment layer. The results show that higher TSRs intensify turbulence, accelerate wake recovery over rigid beds, and enhance erosion and deposition patterns beneath and downstream of the turbine under live-bed conditions. Bed deformation under live-bed conditions induces asymmetrical wake structures through jet flows, further accelerating wake recovery and decreasing turbine performance by about 2%, compared to rigid-bed conditions. Considering the computational cost of the ASM framework, which is nearly $4\%$ of the turbine-resolving approach, it provides an efficient yet robust tool for assessing flow-sediment-turbine interactions.

physics.flu-dyn↗

arXiv2025

Wind farm layout optimization using a novel machine learning approach

We present a novel approach to optimize wind farm layouts for maximum annual energy production (AEP). The optimization effort requires efficient wake models to predict the wake flow and, subsequently, the power generation of wind farms with reasonable accuracy and low computational cost. Wake flow predictions using large-eddy simulation (LES) ensure high fidelity, while reduced-order models, e.g., the Gaussian-curl hybrid (GCH), provide computational efficiency. We integrate LES results and the GCH model to develop a machine learning (ML) framework based on an autoencoder-based convolutional neural network, allowing for a reliable and cost-effective prediction of the wake flow field. We trained the ML model using high-fidelity LES results as the target vector, while low-fidelity data from the GCH model serve as the input vector. The efficiency of the ML model to predict the AEP of the South Fork wind farm, offshore Rhode Island, was illustrated. Then, we integrated the ML model into a greedy optimization algorithm to determine the optimal wind farm layout in terms of turbine positioning. The optimized wind farm layout is shown to achieve a 2. 05\% improvement in AEP over the existing wind farm.

physics.flu-dyn↗

arXiv2025

On the interaction of fish and marine hydrokinetic turbines: Insights gained through experimental and computational observations

Tidal and riverine hydrokinetic turbines offer promising solutions for renewable energy generation in aquatic environments. However, their ecological impact, especially on fish behavior, warrants a thorough investigation. This study presents an integrated experimental and computational analysis of fish turbine interactions, combining laboratory observations with high fidelity large eddy simulations. The simulations capture essential flow features, including wake asymmetry, vortex shedding, and spatial variations in turbulence intensity, under varying flow and rotational regimes of a geometry resolved vertical axis turbine. Behavioral experiments with rainbow trout reveal a consistent tendency to avoid high turbulence and high shear regions, favoring low turbulence zones such as downstream sidewalls. Hydrodynamic stressors and energetic demands were characterized using turbulence metrics, including turbulence kinetic energy, Reynolds stress, and integral length scale, along with estimations of fish generated thrust force. Our results demonstrate that turbine induced turbulence significantly influences fish movement and habitat selection, highlighting the need to consider behavioral responses in conventional fish injury assessment frameworks. These findings provide critical insights for designing and operating hydrokinetic turbines in ecologically sensitive waters, ensuring a balance between renewable energy extraction and aquatic ecosystem protection.

physics.flu-dyn↗

arXiv2025

Large eddy simulation of a utility-scale vertical-axis marine hydrokinetic turbine under live-bed conditions

We present a coupled large-eddy simulation (LES) and bed morphodynamics study to investigate the impact of sediment dynamics on the wake flow, wake recovery and power production of a utility-scale marine hydrokinetic vertical-axis turbine (VAT). A geometry-resolving immersed boundary method is employed to capture the turbine components, the waterway, and the sediment layer. Our numerical findings reveal that increasing the turbine tip speed ratio (TSR) would intensify turbulence, accelerate wake recovery, and increase erosion at the base of the device. Furthermore, it is found that the deformation of the bed around the turbine induces a jet-like flow near the bed beneath the turbine, which enhances wake recovery. Analyzing the interactions between turbulent flow and bed morphodynamics, this study seeks to provide physical information on the environmental and operational implications of VAT deployment in natural riverine and marine environments.

physics.flu-dyn↗

arXiv2024

Toward ultra-efficient high-fidelity prediction of bed morphodynamics of large-scale meandering rivers using a novel LES-trained machine learning approach

Flood-induced deformation of the bed topography of fluvial meandering rivers could lead to river bank displacement, structural failure of the infrastructures, and the propagation of scour or deposition features. The assessment of sediment transport in large-scale meanders is, therefore, a key environmental issue. High-fidelity numerical models provide powerful tools for such assessments. However, high-fidelity simulations of large-scale rivers using the coupled flow and morphodynamics modules can be computationally expensive, owing to the costly two-way coupling between turbulence and bed morphodynamics. This study seeks to present a novel machine learning approach, which is trained using coupled large-eddy simulation (LES) and morphodynamics results. The proposed machine learning approach predicts bed shear stress and equilibrium bed morphology of large-scale meanders under bankfull flow conditions at a fraction of the cost of coupled LES-morphodynamics. We developed and evaluated the performance of a convolutional neural network autoencoder (CNNAE) algorithm to generate high-fidelity bed shear stress and equilibrium morphology of large-scale meandering rivers. The CNNAE algorithm utilizes instantaneous shear stress distribution and change of bed elevation of high-fidelity simulation results, along with geometric parameters of meanders as inputs to predict mean bed shear stress distribution and equilibrium bed elevation of rivers. The results demonstrated the feasibility, accuracy, and efficiency of the proposed CNNAE algorithm.

physics.flu-dyn↗