slientruss3d : Python for stable truss analysis and deep learning research

Description

slientruss3d is a python package which can solve the resistances, internal forces and joint dispalcements in a stable 2D or 3D truss by direct stiffness method. And also can do truss optimization by Genetic Algorithm (GA), generate truss data and work with pytorch-geometric conveniencely.

This repo is writen by :

Taiwan                                          (臺灣)
Department of Civil Engineering                 (土木工程學系)
National Yang Ming Chiao Tung University (NYCU) (國立陽明交通大學)
Shih-Chi Cheng                                  (鄭適其)

Content

1. Installaltion
   • Install
   • Time consuming
   • Update log
2. Quick start
   • Basic example >( Just read this if you aren't familiar with coding 😎 )
   • Truss
   • Member
   • MemberType
   • SupportType
3. Combine with JSON
   • Example code
   • Embed in Web APP
   • Format of JSON
4. Plot your truss
   • Example code
   • Example figures
5. Truss optimization
   • Introduction
   • Gene
   • Fitness function
   • Crossover
   • Evolution policy
   • Example code
   • Geneic algorithm
   • Customization
6. Generate truss data automatically
   • Introduction
   • Generate cube-like truss
   • Data Augmentation
7. Convert Truss to Pytorch-Geometric HeteroData
   • Introduction
   • Installation
   • How to use it ?
   • Fields in HeteroData
   • Example code

Install

First, check your python version:

Python must >= 3.9.7

Second, download the slientruss3d package:

pip install slientruss3d 

Time consuming

The following are time consuming tests for doing structural analysis for each truss (Each testing runs for 30 times and takes average !).

• 6-bar truss   : 0.00037(s)
• 10-bar truss  : 0.00050(s)
• 25-bar truss  : 0.00126(s)
• 47-bar truss  : 0.00253(s)
• 72-bar truss  : 0.00323(s)
• 120-bar truss : 0.00557(s)
• 942-bar truss : 0.05253(s)

Testing on :

CPU: Intel(R) Core(TM) i7-10750H CPU @ 2.60GHz
RAM: 8GB DDR4 * 2

Update log

New feature in v2.0.0 update !

• Important API adjustment : We simplified the JSON format in slientruss3d, see the details in Format of JSON. You can use the v1_to_v2.py module in the root folder to convert the old JSON format to the new one rapidly.

• Data Augmentation : You can use some new method in slientruss3d.generate module to do data augmentation to generated cube-like trusses ! See more details in Data Augmentation. For example:

• Graph Deep Learning : With the increasing importance of deep learning in the field of truss design, we also provide a solution to let our users convert the Truss object to the data structure of Pytorch-Geometric conveniently. See the details in Convert Truss to Pytorch-Geometric HeteroData.

An important API adjustment after v1.3.25 !

After slientruss3d v1.3.25, the method Truss.Solve() will return None instead of the result of structural analysis. If you want to get the result of structural analysis, please use other getters defined in Truss object such as:

# Get result of structural analysis:
displace, stress, resistance = truss.GetDisplacements(), truss.GetInternalStresses(), truss.GetResistances()
return displace, stress, resistance

New feature in v1.3.0 update !

After slientruss3d v1.3.0, 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.

More details are in Generate truss data automatically

New feature in v1.2.0 update !

After slientruss3d v1.2.0, you could use slientruss3d.ga module to do member type selection optimization conveniencely with Genetic Algorithm (GA)! Just simply define the topology of the truss and what member types you want to use, and then you could start the optimization.

More details are in Truss optimization

Besides GA, there are some new useful methods in the Truss object:

class Truss:

    ...

    # Check whether all internal forces are in allowable range or not:
    def IsInternalStressAllowed(self, limit, isGetSumViolation=False) -> tuple[bool, dict | float]: 
        ...

    # Check whether all internal displacements are in allowable range or not:
    def IsDisplacementAllowed(self, limit, isGetSumViolation=False) -> tuple[bool, dict | float]:
        ...

Enjoy 😎 !