gemini_radioss.md

If you want to bypass Kitware's animtovtkhdf translation layer entirely, you can read OpenRadioss ANIM files directly into Python memory using Vortex-Radioss, an open-source Python library developed by the OpenRadioss community specifically for post-processing.

Vortex-Radioss unpacks the compressed binary/ASCII arrays inside A00x files into standard NumPy arrays. Once the data is in NumPy, it is straightforward to construct a native pyvista.UnstructuredGrid.

Step 1: Install Dependencies

Vortex-Radioss relies on lasso-python for CAE array handling.

pip install lasso-python

Next, download or clone the Vortex-Radioss library from their GitHub repository: Vortex-CAE/Vortex-Radioss. Ensure the module is in your Python path.

Step 2: The Python Pipeline

Here is a template on how to bridge the RadiossReader output natively into PyVista. Because Vortex-Radioss returns raw arrays, we must manually format the element connectivity into VTK's expected format (where each cell is prefixed by its number of points).

import numpy as np
import pyvista as pv

# Import the native Radioss reader from Vortex-CAE
from vortex_radioss.animtod3plot.RadiossReader import RadiossReader

def read_anim_to_pyvista(filepath):
    # 1. Parse the OpenRadioss ANIM file
    anim = RadiossReader(filepath)
    
    # You can inspect the available arrays using:
    # print(anim.raw_header.keys())
    
    # 2. Extract Geometry (Adjust dictionary keys based on your specific ANIM output)
    # Assuming 'nodes' is an (N, 3) array of XYZ coordinates
    nodes = anim.raw_header['node_coordinates'] 
    
    # Assuming 'shells' is an (M, 4) array of 0-indexed node IDs for Quad elements
    shell_connectivity = anim.raw_header['shell_connectivity'] 
    
    # 3. Format Connectivity for PyVista / VTK
    # VTK requires connectivity arrays to be padded with the number of points per cell.
    # For example, a quad is: [4, node0, node1, node2, node3]
    num_cells = shell_connectivity.shape[0]
    points_per_cell = shell_connectivity.shape[1]
    
    # Create the padding column [4, 4, 4, ...]
    padding = np.full((num_cells, 1), points_per_cell, dtype=np.int64)
    
    # Horizontally stack the padding and the connectivity, then flatten it
    vtk_connectivity = np.hstack((padding, shell_connectivity)).flatten()
    
    # Specify the VTK cell type (VTK_QUAD = 9, VTK_HEXAHEDRON = 12, etc.)
    # PyVista stores these in pv.CellType
    cell_types = np.full(num_cells, pv.CellType.QUAD, dtype=np.uint8)
    
    # 4. Construct the PyVista UnstructuredGrid
    grid = pv.UnstructuredGrid(vtk_connectivity, cell_types, nodes)
    
    # 5. Attach Results Data
    # For example, if you requested velocities (/ANIM/VECT/VEL)
    if 'nodal_velocities' in anim.raw_header:
        grid.point_data['Velocity'] = anim.raw_header['nodal_velocities']
        
    # If you requested Von Mises stress (/ANIM/ELEM/VONM)
    if 'element_von_mises' in anim.raw_header:
        grid.cell_data['Von Mises Stress'] = anim.raw_header['element_von_mises']
        
    return grid

# --- Usage ---
mesh = read_anim_to_pyvista("data/cube_TYPE7A001")
mesh.plot(scalars="Von Mises Stress", cmap="turbo", show_edges=True)

Important Nuances for this Approach:

• Array Keys: Depending on what you requested in your Engine file (e.g., /ANIM/VECT/VEL, /ANIM/ELEM/VONM), the exact string keys inside anim.raw_header will vary. You will need to print anim.raw_header.keys() on your specific file to map the names correctly.
• Mixed Element Types: If your model contains multiple element types (e.g., hexas, shells, and beams), you will need to concatenate their connectivity arrays and cell_types arrays together before passing them into pv.UnstructuredGrid.
• Zero-Indexing: Ensure your connectivity arrays are 0-indexed. If Radioss outputs 1-indexed node IDs, subtract 1 from the connectivity array before stacking it with the VTK padding.