Evolutionary Topology Optimisation with GSA

Inspired by a recent Arup Thoughts article about a radical minimum weight joint, and making use of some online references (Huang and Xie, 2010), I have developed a Python routine for running an Evolutionary Topology Optimisation (ETO) on a GSA mesh.

The basic idea of this method is quite simple: remove the bits of the structure that are doing the least work, then repeat until everything is working hard. Hard work in this case is defined by the ratio of the highest stress, or strain energy, to the lowest (I chose strain energy in this case), and switch off any element that falls beneath the Rejection Ratio (RR). The trick is to ramp up the RR slowly so that the remaining structure is stable. You do this by starting the RR low, say at 1% of peak stress or 0.1% strain energy, and then increasing it by an small amount, the Evolutionary Rate (ER), whenever all elements pass the RR test. You then stop when RR reaches a set amount, say 25%.

In this Python script I make use of Analysis stages, so nothing is actually deleted from the model and you can see how the structure optimises at each step. On the other hand, I delete the analysis results after each loop to keep the file size down, so you will need to analyse all if you want to see how the stresses develop over time.

As an example, here is a solid mesh with two fixed support points that evolves into a cross between an arch and a truss.

The routine will end when either the RR reaches the max value or when the analysis fails. If the analysis fails because of bits of the mesh coming loose, try halving the evolutionary rate, as this can allow the mesh time to remove the weak areas.

Note that when you run the script you will need to change the directory to something suitable on your machine, as well as possibly the file names.

Do please share any other interesting meshes and their required ETO settings.This routine should work on either 2d mesh or 1d elements, as both give a strain energy output. If you want to change the script to work on stresses you will need to be more careful as 1D and 2D elements will have different output formats.

Reference: Huang, X. and Xie, Y. (2010). Evolutionary topology optimization of continuum structures. Chichester, West Sussex: Wiley.

Evolutionary Topology Optimisation Python files

LINK:	http://www.oasys-software.com/blog/2015/07/evolutionary-topology-optim...
	See more stories from oasys

Most recent headlines