Object-Oriented Version of 99 lines Topology Optimization

General

Object-oriented and numba optimized version of "99 lines Topology optimization in Matlab" Python version. Main objective is simply is to enhance numerical experiments when changing topology optimization parameters and boundary condition and load cases.

Object-oriented version allows to force seamlessly changes in topology optimization settings and numba allows to speed up computations.

Example 1: Design Generation with one load

from SetTopol import TopolSettings  

This imports all necessary classes and modules to set and launch a new topology optimization

top = TopolSettings()
top.optimize()

This launch a topology optimization with default values as in the original version from

You should see in the terminal :

In [3]: top.optimize()
it.: 1 , obj.: 59.825 Vol.: 0.500, ch.: 0.200
it.: 2 , obj.: 24.137 Vol.: 0.500, ch.: 0.200
...
it.: 84 , obj.: 8.464 Vol.: 0.500, ch.: 0.001
it.: 85 , obj.: 8.464 Vol.: 0.500, ch.: 0.001
Elapsed time : 4.980943202972412 s

If you need to plot the evolution of the optimization then, you need to specify store=True

top.optimize(store=True)

This will store all density values for the whole optimization and allow to plot a animation of the optimization and the optimization objective at once.

To save the evolution of design generation and of its objective function (here compliance), you need to call plot function and specify the name of the .gif animation

[In 5]: top.plot(name='example')
Saving Animation ... 

To save only the final design, you need to call save_design function with a name to the .png design image

top.save_design(name='example')
Saving design ... 

Features

OO version allow to seamlessy force change in dependent parameters. Typically when changing nx (number of elements in x direction), ndofs (number of degrees of freedom) change as well. To see it :

In [7]: top
Out[7]:
Topology optimization
   50 elements in x_direction, 50 elements in y direction
   5202 total number of degrees of freedom   0.5 of total volume allowed
   5.4 radius filter

and try to change nx (number of elements in the x direction)

In [8]: top.nx = 100
Caution this will change number of dofs and hence
 the optimization problem

In [9]: top
Out[9]:
Topology optimization
   100 elements in x_direction, 50 elements in y direction
   10302 total number of degrees of freedom   0.5 of total volume allowed
   5.4 radius filter

now let us reset nx to 50 and try to change the load node, then, re-run the optimization and save the new design

top.setf(value=1, node=51*51-1, teta=270)
top.optimize(store=True)
top.plot(name="example2")
Out[10]:
it.: 1 , obj.: 122.849 Vol.: 0.500, ch.: 0.200
it.: 2 , obj.: 67.225 Vol.: 0.500, ch.: 0.200
it.: 3 , obj.: 44.583 Vol.: 0.500, ch.: 0.200
...
it.: 60 , obj.: 23.910 Vol.: 0.500, ch.: 0.001
Elapsed time : 8.248879671096802 s
Saving Animation ... 

Example 2: Design Generation with multiple loads

This design is fixed from the left side and loaded with 2 loads of $1N$ each: one upward in the top-righ corner, one downward in the bottom-left corner top-right corner load: ( top.nx \times (top.ny+1) ) + upward direction => angle = 90 bottom-left corner load: ( (top.nx+1) \times (top.ny+1)-1 ) + downward direction => angle = 180+90

its dimension is $100 \times 100$

from SetTopol_multiple_loads import TopolSettings
import numpy as np
top = TopolSettings(nx = 100, ny = 100, nbr_loads=2, vol = 0.4, rmin = 1.2, penalinit = 3.0, penalmed = 3.0, filt = 0, nu=0.3)
top.fixed= np.arange(0, top.ny+1).tolist()
top.setf(values=[1,1], nodes=[(top.nx+1)*(top.ny+1)-1, top.nx*(top.ny+1) ], tetas=[180+90, 90])
top.optimize(store=True)
top.plot(name="example3")

Requirements

Python > 3.6

numba

matplotlib

numpy