From 973677708d5b53727c989412ce1a9513879ac2c9 Mon Sep 17 00:00:00 2001 From: ram-cherukuri <104155145+ram-cherukuri@users.noreply.github.com> Date: Fri, 20 Dec 2024 12:03:48 -0500 Subject: [PATCH 1/6] Create Readme.md --- examples/cfd/external_aerodynamics/Readme.md | 12 ++++++++++++ 1 file changed, 12 insertions(+) create mode 100644 examples/cfd/external_aerodynamics/Readme.md diff --git a/examples/cfd/external_aerodynamics/Readme.md b/examples/cfd/external_aerodynamics/Readme.md new file mode 100644 index 000000000..47c29f0be --- /dev/null +++ b/examples/cfd/external_aerodynamics/Readme.md @@ -0,0 +1,12 @@ +# Training recipes for External Aerodynamics + +![Results from DoMINO for RTWT SC demo](../../../../docs/img/domino_result_rtwt.jpg) + +We have a set of model training recipes that highlight key architectures for external automotive aerodynamics use case. +The architectures are: +- DoMINO is a local, multi-scale, point-cloud based model architecture and the recipe trains a model to take STL +geometries as input and evaluates flow quantities such as pressure and +wall shear stress on the surface of the car as well as velocity fields and pressure +in the volume around it. +- XAeroNet is a collection of scalable models for large-scale external aerodynamic evaluations. It consists of two models, XAeroNet-S and XAeroNet-V for surface and volume predictions, respectively. +- FIGConvUNet [1], a novel architecture that can efficiently solve CFD problems for large 3D meshes and arbitrary input and output geometries. FIGConvUNet efficiently combines U-shaped architecture, graph information gathering, and integration, learning efficient latent representation through the representation graph voxel layer. From 813b180516978ff37e80a4c8e1cf0696e262bdb4 Mon Sep 17 00:00:00 2001 From: ram-cherukuri <104155145+ram-cherukuri@users.noreply.github.com> Date: Fri, 20 Dec 2024 12:44:30 -0500 Subject: [PATCH 2/6] Update Readme.md --- examples/cfd/external_aerodynamics/Readme.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/examples/cfd/external_aerodynamics/Readme.md b/examples/cfd/external_aerodynamics/Readme.md index 47c29f0be..b1656adf3 100644 --- a/examples/cfd/external_aerodynamics/Readme.md +++ b/examples/cfd/external_aerodynamics/Readme.md @@ -9,4 +9,4 @@ geometries as input and evaluates flow quantities such as pressure and wall shear stress on the surface of the car as well as velocity fields and pressure in the volume around it. - XAeroNet is a collection of scalable models for large-scale external aerodynamic evaluations. It consists of two models, XAeroNet-S and XAeroNet-V for surface and volume predictions, respectively. -- FIGConvUNet [1], a novel architecture that can efficiently solve CFD problems for large 3D meshes and arbitrary input and output geometries. FIGConvUNet efficiently combines U-shaped architecture, graph information gathering, and integration, learning efficient latent representation through the representation graph voxel layer. +- FIGConvUNet, a novel architecture that can efficiently scale for large 3D meshes and arbitrary input and output geometries. FIGConvUNet efficiently combines U-shaped architecture, graph information gathering, and integration, learning efficient latent representation through the representation graph voxel layer. From a32746de3015319b573e8e0408fbc7e35b23936b Mon Sep 17 00:00:00 2001 From: ram-cherukuri <104155145+ram-cherukuri@users.noreply.github.com> Date: Fri, 20 Dec 2024 12:49:41 -0500 Subject: [PATCH 3/6] Update Readme.md --- examples/cfd/external_aerodynamics/Readme.md | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) diff --git a/examples/cfd/external_aerodynamics/Readme.md b/examples/cfd/external_aerodynamics/Readme.md index b1656adf3..e9c592f52 100644 --- a/examples/cfd/external_aerodynamics/Readme.md +++ b/examples/cfd/external_aerodynamics/Readme.md @@ -5,8 +5,6 @@ We have a set of model training recipes that highlight key architectures for external automotive aerodynamics use case. The architectures are: - DoMINO is a local, multi-scale, point-cloud based model architecture and the recipe trains a model to take STL -geometries as input and evaluates flow quantities such as pressure and -wall shear stress on the surface of the car as well as velocity fields and pressure -in the volume around it. -- XAeroNet is a collection of scalable models for large-scale external aerodynamic evaluations. It consists of two models, XAeroNet-S and XAeroNet-V for surface and volume predictions, respectively. -- FIGConvUNet, a novel architecture that can efficiently scale for large 3D meshes and arbitrary input and output geometries. FIGConvUNet efficiently combines U-shaped architecture, graph information gathering, and integration, learning efficient latent representation through the representation graph voxel layer. +geometries as input and evaluates flow quantities such as pressure and wall shear stress on the surface of the car as well as velocity fields and pressure in the volume around it. Learn more about training recipe with DoMINO here (https://github.com/NVIDIA/modulus/blob/ram-cherukuri-patch-1/examples/cfd/external_aerodynamics/domino/README.md)). +- XAeroNet is a collection of scalable models for large-scale external aerodynamic evaluations. It consists of two models, XAeroNet-S and XAeroNet-V for surface and volume predictions, respectively. Learn more about training recipe with XAeroNet here (https://github.com/NVIDIA/modulus/blob/ram-cherukuri-patch-1/examples/cfd/external_aerodynamics/xaeronet/README.md) +- FIGConvUNet, a novel architecture that can efficiently scale for large 3D meshes and arbitrary input and output geometries. FIGConvUNet efficiently combines U-shaped architecture, graph information gathering, and integration, learning efficient latent representation through the representation graph voxel layer. Learn more about training recipe with FIGConvUNet here (https://github.com/NVIDIA/modulus/blob/ram-cherukuri-patch-1/examples/cfd/external_aerodynamics/figconvnet/README.md) From 78c4beae1532b898b305efd6463e2919e9cade4f Mon Sep 17 00:00:00 2001 From: ram-cherukuri <104155145+ram-cherukuri@users.noreply.github.com> Date: Fri, 20 Dec 2024 12:51:19 -0500 Subject: [PATCH 4/6] Update Readme.md --- examples/cfd/external_aerodynamics/Readme.md | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/examples/cfd/external_aerodynamics/Readme.md b/examples/cfd/external_aerodynamics/Readme.md index e9c592f52..9cb6569e8 100644 --- a/examples/cfd/external_aerodynamics/Readme.md +++ b/examples/cfd/external_aerodynamics/Readme.md @@ -5,6 +5,6 @@ We have a set of model training recipes that highlight key architectures for external automotive aerodynamics use case. The architectures are: - DoMINO is a local, multi-scale, point-cloud based model architecture and the recipe trains a model to take STL -geometries as input and evaluates flow quantities such as pressure and wall shear stress on the surface of the car as well as velocity fields and pressure in the volume around it. Learn more about training recipe with DoMINO here (https://github.com/NVIDIA/modulus/blob/ram-cherukuri-patch-1/examples/cfd/external_aerodynamics/domino/README.md)). -- XAeroNet is a collection of scalable models for large-scale external aerodynamic evaluations. It consists of two models, XAeroNet-S and XAeroNet-V for surface and volume predictions, respectively. Learn more about training recipe with XAeroNet here (https://github.com/NVIDIA/modulus/blob/ram-cherukuri-patch-1/examples/cfd/external_aerodynamics/xaeronet/README.md) -- FIGConvUNet, a novel architecture that can efficiently scale for large 3D meshes and arbitrary input and output geometries. FIGConvUNet efficiently combines U-shaped architecture, graph information gathering, and integration, learning efficient latent representation through the representation graph voxel layer. Learn more about training recipe with FIGConvUNet here (https://github.com/NVIDIA/modulus/blob/ram-cherukuri-patch-1/examples/cfd/external_aerodynamics/figconvnet/README.md) +geometries as input and evaluates flow quantities such as pressure and wall shear stress on the surface of the car as well as velocity fields and pressure in the volume around it. Learn more about training recipe with DoMINO [here](https://github.com/NVIDIA/modulus/blob/ram-cherukuri-patch-1/examples/cfd/external_aerodynamics/domino/README.md)). +- XAeroNet is a collection of scalable models for large-scale external aerodynamic evaluations. It consists of two models, XAeroNet-S and XAeroNet-V for surface and volume predictions, respectively. Learn more about training recipe with XAeroNet [here](https://github.com/NVIDIA/modulus/blob/ram-cherukuri-patch-1/examples/cfd/external_aerodynamics/xaeronet/README.md) +- FIGConvUNet, a novel architecture that can efficiently scale for large 3D meshes and arbitrary input and output geometries. FIGConvUNet efficiently combines U-shaped architecture, graph information gathering, and integration, learning efficient latent representation through the representation graph voxel layer. Learn more about training recipe with FIGConvUNet [here](https://github.com/NVIDIA/modulus/blob/ram-cherukuri-patch-1/examples/cfd/external_aerodynamics/figconvnet/README.md) From 75d1c09a2e19a7d8c2146a0c85e92cc6818759eb Mon Sep 17 00:00:00 2001 From: ram-cherukuri <104155145+ram-cherukuri@users.noreply.github.com> Date: Fri, 20 Dec 2024 12:51:33 -0500 Subject: [PATCH 5/6] Update Readme.md --- examples/cfd/external_aerodynamics/Readme.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/examples/cfd/external_aerodynamics/Readme.md b/examples/cfd/external_aerodynamics/Readme.md index 9cb6569e8..00e429855 100644 --- a/examples/cfd/external_aerodynamics/Readme.md +++ b/examples/cfd/external_aerodynamics/Readme.md @@ -5,6 +5,6 @@ We have a set of model training recipes that highlight key architectures for external automotive aerodynamics use case. The architectures are: - DoMINO is a local, multi-scale, point-cloud based model architecture and the recipe trains a model to take STL -geometries as input and evaluates flow quantities such as pressure and wall shear stress on the surface of the car as well as velocity fields and pressure in the volume around it. Learn more about training recipe with DoMINO [here](https://github.com/NVIDIA/modulus/blob/ram-cherukuri-patch-1/examples/cfd/external_aerodynamics/domino/README.md)). +geometries as input and evaluates flow quantities such as pressure and wall shear stress on the surface of the car as well as velocity fields and pressure in the volume around it. Learn more about training recipe with DoMINO [here](https://github.com/NVIDIA/modulus/blob/ram-cherukuri-patch-1/examples/cfd/external_aerodynamics/domino/README.md) - XAeroNet is a collection of scalable models for large-scale external aerodynamic evaluations. It consists of two models, XAeroNet-S and XAeroNet-V for surface and volume predictions, respectively. Learn more about training recipe with XAeroNet [here](https://github.com/NVIDIA/modulus/blob/ram-cherukuri-patch-1/examples/cfd/external_aerodynamics/xaeronet/README.md) - FIGConvUNet, a novel architecture that can efficiently scale for large 3D meshes and arbitrary input and output geometries. FIGConvUNet efficiently combines U-shaped architecture, graph information gathering, and integration, learning efficient latent representation through the representation graph voxel layer. Learn more about training recipe with FIGConvUNet [here](https://github.com/NVIDIA/modulus/blob/ram-cherukuri-patch-1/examples/cfd/external_aerodynamics/figconvnet/README.md) From c1727eb5fd55e6da349641dcb1f4d779bacc5391 Mon Sep 17 00:00:00 2001 From: ram-cherukuri <104155145+ram-cherukuri@users.noreply.github.com> Date: Fri, 20 Dec 2024 17:18:07 -0500 Subject: [PATCH 6/6] Update Readme.md --- examples/cfd/external_aerodynamics/Readme.md | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/examples/cfd/external_aerodynamics/Readme.md b/examples/cfd/external_aerodynamics/Readme.md index 00e429855..f1cbed5ff 100644 --- a/examples/cfd/external_aerodynamics/Readme.md +++ b/examples/cfd/external_aerodynamics/Readme.md @@ -5,6 +5,6 @@ We have a set of model training recipes that highlight key architectures for external automotive aerodynamics use case. The architectures are: - DoMINO is a local, multi-scale, point-cloud based model architecture and the recipe trains a model to take STL -geometries as input and evaluates flow quantities such as pressure and wall shear stress on the surface of the car as well as velocity fields and pressure in the volume around it. Learn more about training recipe with DoMINO [here](https://github.com/NVIDIA/modulus/blob/ram-cherukuri-patch-1/examples/cfd/external_aerodynamics/domino/README.md) -- XAeroNet is a collection of scalable models for large-scale external aerodynamic evaluations. It consists of two models, XAeroNet-S and XAeroNet-V for surface and volume predictions, respectively. Learn more about training recipe with XAeroNet [here](https://github.com/NVIDIA/modulus/blob/ram-cherukuri-patch-1/examples/cfd/external_aerodynamics/xaeronet/README.md) -- FIGConvUNet, a novel architecture that can efficiently scale for large 3D meshes and arbitrary input and output geometries. FIGConvUNet efficiently combines U-shaped architecture, graph information gathering, and integration, learning efficient latent representation through the representation graph voxel layer. Learn more about training recipe with FIGConvUNet [here](https://github.com/NVIDIA/modulus/blob/ram-cherukuri-patch-1/examples/cfd/external_aerodynamics/figconvnet/README.md) +geometries as input and evaluates flow quantities such as pressure and wall shear stress on the surface of the car as well as velocity fields and pressure in the volume around it. Learn more about training recipe with DoMINO [here](https://github.com/NVIDIA/modulus/tree/main/examples/cfd/external_aerodynamics/domino/README.md) +- XAeroNet is a collection of scalable models for large-scale external aerodynamic evaluations. It consists of two models, XAeroNet-S and XAeroNet-V for surface and volume predictions, respectively. Learn more about training recipe with XAeroNet [here](https://github.com/NVIDIA/modulus/tree/main/examples/cfd/external_aerodynamics/xaeronet/README.md) +- FIGConvUNet, a novel architecture that can efficiently scale for large 3D meshes and arbitrary input and output geometries. FIGConvUNet efficiently combines U-shaped architecture, graph information gathering, and integration, learning efficient latent representation through the representation graph voxel layer. Learn more about training recipe with FIGConvUNet [here](https://github.com/NVIDIA/modulus/tree/main/examples/cfd/external_aerodynamics/figconvnet/README.md)