gen_truss.md

Generate truss data automatically

After slientruss3d 1.3.x, you can use slientruss3d.generate module to generate truss data automatically. For now, only simple cube-like truss can be generated by slientruss3d, but I think this is still a helpful way for anyone who suffers from lake of truss data.

---

Introduction

The following is an example figure for how 4-cube truss looks like. And also introduce some technical terms here.

---

Generate cube-like truss

Generate cube-like trusses and save them into JSON files.

from slientruss3d.truss    import Truss
from slientruss3d.type     import LinkType, GenerateMethod
from slientruss3d.generate import GenerateRandomCubeTrusses, NoChange

GenerateRandomCubeTrusses(gridRange              = (5, 5, 5), 
                          numCubeRange           = (5, 5), 
                          numCaseRange           = (1, 10), 
                          lengthRange            = (50, 150), 
                          forceRange             = [(-30000, 30000), (-30000, 30000), (-30000, 30000)],
                          nForceRange            = None,
                          method                 = GenerateMethod.Random,
                          linkType               = LinkType.Random,
                          memberTypes            = [[1., 1e7, 0.1]],
                          isAddPinSupport        = True,
                          isAllowParallel        = False,
                          isDoStructuralAnalysis = False,
                          isPlotTruss            = False,
                          isPrintMessage         = True,
                          saveFolder             = None,
                          augmenter              = NoChange(),
                          seed                   = None  ) -> list[Truss]

• gridRange : Max range of the global cube grid.
• numCubeRange : Range of the total numbers of cube-like truss block on the global grid.
• numCaseRange : Range of the numbers of cases for each numCubeRange (only influence names of the saved JSON files).
• lengthRange : Range of the edge length (that is, height, width, deepth) for every cube-like truss block.
• forceRange : Range of the magnitudes for xyz-axis of the random forces assigned at joints.
• nForceRange : Range of the number of joints which will be assigned random external forces. If it's None, nForceRange ~ Uniform(1, Number_of_joints).
• method : Algorithm to decide positions to generate cube-like truss blocks.
• linkType : Link type of faces of every cube-like truss block.
• memberTypes : The member types in the list will be randomly assigned to each member. (list[list] or list[slientruss3d.type.MemberType])
• isAddPinSupport : Whether to add pin supports in the generated truss or not.
• isAllowParallel : Whether to allow parallel members in the generated truss or not.
• isDoStructuralAnalysis : Whether to do structral analysis after each truss be generated.
• isPlotTruss : Whether to plot the truss after each truss be generated.
• isPrintMessage : Whether to print the message for the generating progress on the screen.
• saveFolder : Folder to save the generated result (in JSON file). If it's None, this method won't save the generated result to JSON file.
• augmenter : (We will discuss it in the next section)
• seed : Set random seed.

Explanation of global cube grid ( gridRange = (3, 3, 3), numCubeRange = (2, 2), lengthRange = (50, 150) for example ) :

Options of linkType:

• LinekType.LeftBottom_RightTop
• LinekType.RightBottom_LeftTop
• LinekType.Cross
• LinekType.Random   (default)

Options of method:

• GenerateMethod.DFS
• GenerateMethod.BFS
• GenerateMethod.Random   (default)

Data Augmentation

You can do some data augmentation for the cube-like trusses generated by GenerateRandomCubeTrusses or any other 3D trusses. For example:

from slientruss3d.generate import GenerateRandomCubeTrusses
from slientruss3d.generate import MoveToCentroid, RandomTranslation, AddJointNoise, RandomResetPin, NoChange, TrussDataAugmenterList

# Some parameters for your generated cube truss:
GRID_RANGE                = (5, 5, 5)
CUBE_NUMBER               = 4
GENERATE_NUMBER           = 1
EDGE_LENGTH_RANGE         = (100, 200)
EXTERNAL_FORCE_RANGE      = [(-1000, 1000), (-1000, 1000), (-1000, 1000)]
IS_DO_STRUCTURAL_ANALYSIS = True
IS_PLOT_GENERATED_TRUSS   = True
SAVE_FOLDER               = './'

# Data augmentor:
transforms = TrussDataAugmenterList(
    NoChange(),                                                         # Do nothing
    MoveToCentroid(),                                                   # Move the centroid of the truss to [0., 0., 0.]
    RandomTranslation(translateRange=[-30., 30.]),                      # Translate the whole truss randomly
    AddJointNoise(noiseMeans=[0., 0., 0.], noiseStds=[10., 10., 10.]),  # Add guassian noise to all positions of the joints
    RandomResetPin(minNumPin=5, maxNumPinRatio=0.6)                     # Reset the positions and number of pin supports.
)

# Generate cube-like truss:
truss = GenerateRandomCubeTrusses(gridRange=GRID_RANGE,
                                  numCubeRange=(CUBE_NUMBER, CUBE_NUMBER),
                                  numEachRange=(1, GENERATE_NUMBER),
                                  lengthRange=EDGE_LENGTH_RANGE,
                                  forceRange=EXTERNAL_FORCE_RANGE,
                                  isDoStructuralAnalysis=IS_DO_STRUCTURAL_ANALYSIS,
                                  isPlotTruss=IS_PLOT_GENERATED_TRUSS,
                                  saveFolder=SAVE_FOLDER,
                                  seed=42,
                                  augmenter=transforms  # <- Do data augmentation !!!
                                )[0]

Before augmentation:

After augmentation: