Export L-PBF Scan Vectors to VTK/Paraview via PySLM

The in-built visualisation for scan paths in PySLM leverages matplotlib – refer to a previous post. This is sufficient for most user’s needs when attempting to interpret and visualise the scan paths generated in PySLM, or those imported from a slice taken from an existing machine build files. Extending this beyond multiple layers or large parts becomes more tricky when factoring in visualisation of some parameters (e.g. Laser Power, effective scan speed). Admittedly, the performance of Matplotlib becomes limited to explore the intricacies and complexities embedded within the scan vectors. 

For scientific research, the fusion of scan vector geometry with volumetric datasets such as X-Ray CT during post-inspection of parts/samples, or those generated within the build process including pyrometry data, thermal-imaging offer the ability to increase our understanding and insight to observations of the effect of process on the material produced using L-PBF.  GPU based visualisation libraries such (vispy) would offer the possibility to accelerate the performance, but are not user-friendly nor offer interactivity when manipulating views and the data and are often cumbersome when processing volumetric datasets often encountered in Additive Manufacturing. Paraview is a cross-platform open-source scientific visualisation tool that is especially powerful for processing, interaction and visualisation of large-scale scientific datasets.

Paraview and the underlying VTK library offers an alternative ready-made solution to visualise this information, and are most importantly hardware accelerated with the option for raytracing provided by OSPRay and OptiX for latest RTX NVIDIA cards that include Raytracing (RT) cores. Additionally, the data can be augmented and processed using parallelised filters and tools in Paraview.

VTK File Format

Ignoring the HDF5 variations that are most useful for structured data, the underlying format within vtk that used for storing vector based data and point cloud data is the .vtp file format. The modern VTK file formats use an XML schema – unlike the legacy format, to store a structured series of geometry (volumetric data, lines, polygons, 3D elements and point clouds). The internal data format can be stored using ascii encoding or binary. Binary data can be incorporated directly within a parsable .xml format using a Base64 encoding and may additional incorporate internal compression. Alternatively data can be stored in an appended data section located at the footer of the file, which treats data section as a contiguous block of raw data. Different sub-formats exist, that are appropriate for different types of data e.g. volumetric, element based (Finite Volume / Finite Element derived) or polygon based. An approach relevant to export scan vector geometry the .vtp – format is most suitable.

The data stored in the VTK Point file consists of:

  • 3D points coordinates
  • Data attributes stored at each point location
  • Geometric elements (lines, polygons) defining connectivity with reference to the list of point coordinates

Paraview exporter implementation:

The Paraview exporter is simplistic, because the data compression is currently ignored. The process is similar to the technique used in the function pyslm.visualise.plotSequential, whereby hatch and contour vectors are merged and reprocessed in order that they represent always a series of lines (an n x 2 x 2 array). This is not the most efficient option for ContourGeometry (border scans) where scan vectors are continuously joined up, but simplifies the processing working with the data.  

Once the scan vector coordinates and the relevant data are packaged up into a single array, the data is wrote within the sub-sections of the XML file. Data is stored using floating points or integers accordingly in a binary representation. The data used to represent coordinates and indices for each vector, are stored with the ‘appended’ option within the <DataArray> element of each section. The raw data is stored and collected that are then written in the <AppendedData> element at the end of file with raw encoding option chosen. The byte offsets for the position of each ‘chunk’ of data that are referenced by the <DataArray> element are collected and stored incrementally.

For reference, the following information is provided for writing raw data, because this was difficult to obtain from the VTK documentation directly.

<AppendedData encoding=”raw”> Start of Raw Data Section
_ Underscore character is starting location for reading raw data
Section Size (Int32/Int64)Integer representing size of following section (include the size in bytes
with the offsets provided
). The integer type should match the size used in the header.
Raw data (e.g Int32, float32, float64)
….Repeated the above (two rows) for each referenced data section
</AppendedData>

Example Scan Vector Data exported to VTK

An example Aconity .ILT file was imported into PySLM and then exported to a .vtp VTK file that was processed in Paraview. The scan order is visualised by the colour map with each vertex assigned a global-id. The ‘Tube‘ filter was applied to each scan vector in order to improve their visibility.

Visualisation scan-vectors for L-PBF/SLM processed in PySLM and exported to VTK (.vtp) file format

The script excerpt can currently be found on a Gist. This will be later included in future versions of PySLM along with other import/exporters.