Vector Slicer is a program for the generation of print paths for 2D planar patterns in which both the shape and the preferred printing direction are defined. It allows both vector and director operation making it suitable for 3D printing of anisotropic materials such as liquid crystal elastomers (LCEs) or polymers (LCPs).
The program by default uses splay-based seeding which analyses the pattern in order to find regions where the paths have locally diverged the most, and seeds paths there. Furthermore, the filling is determined by generating parameters (GPs) which are optimised using Bayesian optimisation (Bayesopt, https://github.com/rmcantin/bayesopt) in order to minimise the user-defined penalty function.
As an ideal slicing of a pattern where the direction is prescribed is not generally possible, one has to trade the amount of holes, overlaps and director agreement in order to obtain the best slicing for the application, which can be realised by modification of the penalty function.
Please contact Michał Zmyślony at [email protected] with any comments and suggestions.
Please ensure that system specific C++ compiler, git (https://git-scm.com/) and cmake (https://cmake.org/) are installed.
In the desired parent directory (it will create a subdirectory vector_slicer) run
git clone https://github.com/Zmmyslony/vector_slicer.git
cd vector_slicer
sudo chmod +x ./install_lin.sh
./install_lin.sh
Warning: This software uses OpenMP and so is incompatible with AppleClang. By default, the installation script uses LLVM.
In the desired parent directory run
git clone https://github.com/Zmmyslony/vector_slicer.git
cd vector_slicer
sudo chmod +x ./install_mac.sh
./install_lin.sh
In the desired parent directory run in terminal
git clone https://github.com/Zmmyslony/vector_slicer.git
cd vector_slicer
install_win.bat
The program can be used either as a dynamically linked library controlled from Python or directly from C++ using configuration files (see description below).
For most users Python operation is recommended as it contains input file generation, slicing and reading the outputs, while C++ usage requires the input files to be separately generated.
Due to the variety of gcode formats in use, the slicer does NOT generate the output in gcode, but rather in abstract pixel-based coordinates. There is a separate program for translating it into gcode (https://github.com/Zmmyslony/Vector_Slicer_GCode), however, at the time of writing this, it only directly supports Hyrel printers, specifically, System 30M, but other formats can also be implemented on request.
The Python interface can be found in python subdirectory. For an example usage check python/guide.py.
The input patterns are created by combining Shape and Director into Pattern, which then can be used to generate the input files. A few commonly used shapes and directors come predefined in shapes.py and directors.py.
The program requires the input director pattern to be a directory containing two matrices - one boolean which needs to be named shape.csv which tells which pixels ought to be filled, and another called thetaField.csv which is the angle that the director makes with the x-axis. Alternatively, thetaField.csv can be replaced by two matrices called xField.csv and yField.csv which are respectively x and y components of the vector field dictating the preferred direction.
Optionally, analytically calculated splay vector
Additionally, config.txt needs to be provided in order to control the filling. Options are
Unoptimisable:
- PrintRadius (default=5) – defines the relation between grid spacing and path width. Small values (<4) as they introduce numerical errors, likewise large values (>10) offer little improvement.
- InitialSeedingMethod (optional, default=Splay) – options: Splay, Perimeter, Dual; Splay filling first searches regions of maximum divergence and seeds lines there, after that it switches to Perimeter which finds areas on the perimeter where either the splay is zero or the splay vector points outwards and then seeds them, and then dual line the approach is taken in order to ensure complete filling.
- StepLength (default=10) – maximal (and usual) length of each segment of the path. Too short segments will create patterns that are too dense for the printer buffer to handle.
Optimisable:
- Repulsion (default=0) – strength of repulsion from preexisting paths in range RepulsionRadius.
- RepulsionAngle (default=3.14) – the maximum angle between repulsed direction and preferred.
- SeedSpacing (default=2 * PrintRadius)
- Seed (default=0) – used only when using the function "recalculateBestConfig".
- TerminationRadius (default=5) – minimal distance between two separate paths for propagation to continue.
- RepulsionRadius (default=0) – legacy, allows the pattern to repulse from paths further than the print radius away from itself.
There are five text files which configure the execution of the program and they are all contained in the configuration directory:
- files_to_test.txt – here we list using either relative or absolute paths to the directories containing the shape matrix and director field. Multiple directories are allowed.
- disagreement.cfg – sets the number of threads and seeds for each step of optimisation and the final search.
- bayesian_optimisation.cfg – configures the number of iterations, relearning period and other basic configurations.
- disagreement_function.cfg – defines the disagreement function used in the optimisation.
- filling.cfg – switching between vector and director operation, control over removal of points and short lines.
The locations of each of the configuration files together with the directories used for importing and exporting can be modified in the vector_slicer_config.h.in file.
The program saves the patterns in the output directory, with the outputs of each type in their subdirectory with names corresponding to the names of the patterns used.
The main output file is the paths file which contains the slicing information. The format is as follows:
- Layer information is contained between "# Start of pattern" and "# End of pattern". Each layer corresponds to single seed.
- Each line within a layer contains information about single continuous path, where the values are flattened pixel-based coordinates, i.e. x1, y1, x2, y2, etc.
Remaining files:
- best_config - config and seeds used for the generation of paths.
- filled_matrices - contains information about how many times was each pixel fixed for the first (best) seed.
- logs - Bayesopt logs used for debugging.
- optimisation_save - Bayesopt optimisation progress for re-optimisation or optimisation monitoring.
- used_seeds - list of coordinates of seeds used for propagation of paths in the best numerical seed.
