Somnath Roy
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Somnath Roy has a total of 66 co-authors. Here are the top co-authors by number of shared publications:
Manish Kumar
Malaviya National Institute of Technology
6 shared publicationsApurva Raj
4 shared publications- Siddharth Durgaprasad Sharma
Centre for Computational and Data Sciences, Indian Institute of Technology Kharagpur 2 , Kharagpur, West Bengal 721302, India
4 shared publications Debajyoti Kumar
3 shared publications- Debajyoti Kumar3 shared publications
Irshad Alam
3 shared publications
Latest Publications
- 27 October 2026
- Published by Elsevier in European Journal of Mechanics - B/Fluids
- Vol. 122
Flow analysis over a rough cylinder under vortex-induced vibration at a low Reynolds number
- 27 October 2026
- Published by Elsevier in European Journal of Mechanics - B/Fluids
- Vol. 122
Flow analysis over a rough cylinder under vortex-induced vibration at a low Reynolds number
- Open Access
- 26 June 2026
- Published by MDPI AG in Journal of Manufacturing and Materials Processing
- Vol. 10 (7) , 219
Neural Network Enabled Process Parameter Optimization for Laser Powder Bed Fusion of Inconel 718
Laser powder bed fusion (LPBF) is a widely utilized metal additive manufacturing (AM) process for fabricating intricate geometries with high mechanical strength. However, achieving defect-free parts remains challenging due to complex thermodynamics and process variability. Component quality is primarily determined by mel-pool morphology, which depends on key process parameters such as laser power, scan speed, and layer thickness. Improper parameter selection causes defects like porosity (keyhole and lack of fusion), balling, and residual stresses, compromising structural integrity. Optimizing these parameters is crucial but difficult due to the multi-scale, multi-physics nature of the process, which traditionally relies on costly, time-intensive experimental trials. We present results from a data-driven approach using machine learning (ML) models to predict and optimize LPBF melt-pool characteristics, reducing reliance on trial-and-error experimentation. We find that laser power and scan speed predominantly influence the melt-pool formation. Higher scan speeds produce more favorable melt pools, whereas excessive laser power at low scan speeds leads to deep keyhole defects. To predict and classify melt pools efficiently, several ML models are deployed, including logistic regression, decision trees, ensemble learning, and fully connected neural networks. The standard neural network achieved the highest cross-validated macro-F1 score of 0.978 ± 0.014, while the weighted neural network achieved the highest recall for the rare optimal melt-pool class, 0.967 ± 0.050. These findings show that class-weighted learning provides a recall-oriented strategy for identifying suitable LPBF process windows, while avoiding overreliance on single train-test split performance. The findings underscore the effectiveness of ML in accurately classifying LPBF melt pools to rapidly identify optimal process parameters.
- 26 June 2026
- Published by MDPI AG in Journal of Manufacturing and Materials Processing
- Vol. 10 (7) , 219
Open AccessNeural Network Enabled Process Parameter Optimization for Laser Powder Bed Fusion of Inconel 718
Laser powder bed fusion (LPBF) is a widely utilized metal additive manufacturing (AM) process for fabricating intricate geometries with high mechanical strength. However, achieving defect-free parts remains challenging due to complex thermodynamics and process variability. Component quality is primarily determined by mel-pool morphology, which depends on key process parameters such as laser power, scan speed, and layer thickness. Improper parameter selection causes defects like porosity (keyhole and lack of fusion), balling, and residual stresses, compromising structural integrity. Optimizing these parameters is crucial but difficult due to the multi-scale, multi-physics nature of the process, which traditionally relies on costly, time-intensive experimental trials. We present results from a data-driven approach using machine learning (ML) models to predict and optimize LPBF melt-pool characteristics, reducing reliance on trial-and-error experimentation. We find that laser power and scan speed predominantly influence the melt-pool formation. Higher scan speeds produce more favorable melt pools, whereas excessive laser power at low scan speeds leads to deep keyhole defects. To predict and classify melt pools efficiently, several ML models are deployed, including logistic regression, decision trees, ensemble learning, and fully connected neural networks. The standard neural network achieved the highest cross-validated macro-F1 score of 0.978 ± 0.014, while the weighted neural network achieved the highest recall for the rare optimal melt-pool class, 0.967 ± 0.050. These findings show that class-weighted learning provides a recall-oriented strategy for identifying suitable LPBF process windows, while avoiding overreliance on single train-test split performance. The findings underscore the effectiveness of ML in accurately classifying LPBF melt pools to rapidly identify optimal process parameters.
- Open Access
- 20 February 2026
- Published by Elsevier in SSRN Electronic Journal
GPU optimized integration of immersed boundary method and overset mesh framework for moving boundary problems
This paper focuses on a GPU-optimized overset grid framework coupled with a sharp-interface Immersed Boundary Method (IBM) for high-fidelity simulation of incompressible flows involving moving bodies, multi-body interactions, and complex geometries. The method addresses long-standing limitations of overset IBM, such as expensive hole-cutting, donor-search operations, and poor scalability, by leveraging a lightweight sliding-window block reallocation strategy, minimalistic interface maps, and flux-preserving interpolation between coarse and refined meshes. A combined Array-of-Structures / Structure-of-Arrays data structure is devised for efficient management and storage of multiple blocks of the domain. Explicit OpenACC directives leveraging the unified memory architecture enable GPU acceleration. At the same time, a dual-level OpenMP–OpenACC parallelism assigns each overset block to a dedicated GPU, enabling multi-GPU execution without message-passing overhead. The solver is validated and verified across a wide range of canonical and complex cases, including flow past cylinders and spheres, sedimentation of smooth and rough particles, and fluttering plates, with results demonstrating second-order accuracy, local and global mass conservation, and excellent agreement with experimental and numerical benchmarks. High-Reynolds-number propeller simulations confirm DNS-level resolution, with the overset mesh adequately capturing Kolmogorov scales. Multi-body demonstrations involving two flapping robotic butterflies following independent sinusoidal trajectories highlight robust block autonomy and accurate vortex-interaction dynamics. Overall, the proposed framework achieves close to 200× speedup relative to single-core CPU execution, retains accuracy while using significantly fewer grid points than uniform DNS meshes, scales efficiently across multiple GPUs, and establishes a versatile, scalable methodology for simulating turbulent, bio-inspired, and strongly unsteady moving-body flows.
- 20 February 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessGPU optimized integration of immersed boundary method and overset mesh framework for moving boundary problems
This paper focuses on a GPU-optimized overset grid framework coupled with a sharp-interface Immersed Boundary Method (IBM) for high-fidelity simulation of incompressible flows involving moving bodies, multi-body interactions, and complex geometries. The method addresses long-standing limitations of overset IBM, such as expensive hole-cutting, donor-search operations, and poor scalability, by leveraging a lightweight sliding-window block reallocation strategy, minimalistic interface maps, and flux-preserving interpolation between coarse and refined meshes. A combined Array-of-Structures / Structure-of-Arrays data structure is devised for efficient management and storage of multiple blocks of the domain. Explicit OpenACC directives leveraging the unified memory architecture enable GPU acceleration. At the same time, a dual-level OpenMP–OpenACC parallelism assigns each overset block to a dedicated GPU, enabling multi-GPU execution without message-passing overhead. The solver is validated and verified across a wide range of canonical and complex cases, including flow past cylinders and spheres, sedimentation of smooth and rough particles, and fluttering plates, with results demonstrating second-order accuracy, local and global mass conservation, and excellent agreement with experimental and numerical benchmarks. High-Reynolds-number propeller simulations confirm DNS-level resolution, with the overset mesh adequately capturing Kolmogorov scales. Multi-body demonstrations involving two flapping robotic butterflies following independent sinusoidal trajectories highlight robust block autonomy and accurate vortex-interaction dynamics. Overall, the proposed framework achieves close to 200× speedup relative to single-core CPU execution, retains accuracy while using significantly fewer grid points than uniform DNS meshes, scales efficiently across multiple GPUs, and establishes a versatile, scalable methodology for simulating turbulent, bio-inspired, and strongly unsteady moving-body flows.
- 8 October 2025
- Published by Elsevier in Computers & Fluids
- Vol. 303
Application of a high-order meshless method to study mixed convection heat transfer in a cavity with rotating circular cylinders
- 8 October 2025
- Published by Elsevier in Computers & Fluids
- Vol. 303
Application of a high-order meshless method to study mixed convection heat transfer in a cavity with rotating circular cylinders
- 29 September 2025
- Published by Springer Nature in The European Physical Journal Special Topics
- Vol. 234 (23) , 7361-7387
Simulation of mixed and natural convection in power-law fluids inside square enclosure with circular cylinder using sharp interface immersed boundary method
- 29 September 2025
- Published by Springer Nature in The European Physical Journal Special Topics
- Vol. 234 (23) , 7361-7387
Simulation of mixed and natural convection in power-law fluids inside square enclosure with circular cylinder using sharp interface immersed boundary method
- 6 September 2025
- Published by Springer Nature in The European Physical Journal Special Topics
- Vol. 234 (23) , 7123-7149
Effects of increased heart rates on the hemodynamic features and arterial stress in a realistic left anterior descending coronary artery model of a young human with different degrees of stenosis
- 6 September 2025
- Published by Springer Nature in The European Physical Journal Special Topics
- Vol. 234 (23) , 7123-7149
Effects of increased heart rates on the hemodynamic features and arterial stress in a realistic left anterior descending coronary artery model of a young human with different degrees of stenosis
- Open Access
- 5 September 2025
- Published by Cambridge University Press (CUP) in Journal of Fluid Mechanics
- Vol. 1018, A43
Direct numerical simulations reveal vortex stabilisation through streamlined leading and trailing edges of a bileaflet mechanical heart valve
A mechanical heart valve is a durable device used to replace damaged ones inside a living heart, aiming for regulated blood flow to avoid the risks of cardiac failure or stroke. The modern bileaflet designs, featuring two semicircular leaflets, aim to improve blood flow control and minimise turbulence as compared to the older models. However, these valves require lifelong anticoagulation therapy to prevent blood clots, increasing bleeding risks and necessitating regular monitoring. Turbulence within the valve can lead to complications such as haemolysis (damage to red blood cells), thrombosis, platelet activation and valve dysfunction. It also contributes to energy loss, increased cardiac workload, and endothelial damage, potentially impairing the valve efficiency and increasing the risk of infective endocarditis. To address these challenges, a design-modified St Jude Medical (SJM) valve with streamlined edges was conceptualised and assessed using direct numerical simulations. Results show that the streamlined design minimises abrupt blood flow alterations and reduces turbulence-inducing vortices. Compared to existing SJM valves, the new design ensures smoother flow transitions, reduces flow disturbances, and reduces pressure drop. It significantly decreases shear stress, drag and downstream turbulence, enhancing haemodynamic efficiency. These improvements lower the risk of complications such as haemolysis and thrombosis, offering a safer and more efficient option for valve replacement, establishing the potential of edge streamlining in advancing mechanical heart valve technology, and favouring patient outcomes.
- 5 September 2025
- Published by Cambridge University Press (CUP) in Journal of Fluid Mechanics
- Vol. 1018, A43
Open AccessDirect numerical simulations reveal vortex stabilisation through streamlined leading and trailing edges of a bileaflet mechanical heart valve
A mechanical heart valve is a durable device used to replace damaged ones inside a living heart, aiming for regulated blood flow to avoid the risks of cardiac failure or stroke. The modern bileaflet designs, featuring two semicircular leaflets, aim to improve blood flow control and minimise turbulence as compared to the older models. However, these valves require lifelong anticoagulation therapy to prevent blood clots, increasing bleeding risks and necessitating regular monitoring. Turbulence within the valve can lead to complications such as haemolysis (damage to red blood cells), thrombosis, platelet activation and valve dysfunction. It also contributes to energy loss, increased cardiac workload, and endothelial damage, potentially impairing the valve efficiency and increasing the risk of infective endocarditis. To address these challenges, a design-modified St Jude Medical (SJM) valve with streamlined edges was conceptualised and assessed using direct numerical simulations. Results show that the streamlined design minimises abrupt blood flow alterations and reduces turbulence-inducing vortices. Compared to existing SJM valves, the new design ensures smoother flow transitions, reduces flow disturbances, and reduces pressure drop. It significantly decreases shear stress, drag and downstream turbulence, enhancing haemodynamic efficiency. These improvements lower the risk of complications such as haemolysis and thrombosis, offering a safer and more efficient option for valve replacement, establishing the potential of edge streamlining in advancing mechanical heart valve technology, and favouring patient outcomes.
- 5 May 2025
- Published by American Physical Society (APS) in Physical Review Fluids
- Vol. 10 (5)
Probing quasigeostrophic turbulence via complex networks
- 5 May 2025
- Published by American Physical Society (APS) in Physical Review Fluids
- Vol. 10 (5)
Probing quasigeostrophic turbulence via complex networks