Saturday, July 20, 2013

Topology Optimization in GE Jet Engine Bracket Challenge Entry

There is an interesting contest on GrabCAD for designing a lighter weight engine bracket for a GE engine the winner of which GE will produce and test using an additive manufacturing method (maybe DMLS). One of the contestants used PareTO software (methods based on these matlab scripts I linked previously) to design a pretty nice looking bracket.

Update: There are more entries that are using topology optimization.
  • KMWE and TU Delft Team Entry. This comment the team makes is interesting: "Since the optimised topology models are in stl we have started to first create a volume model with stp extension so we meet the competition recuirements. This takes a lot of time!" This bottle neck in the work-flow is similar to the problem CFD analysts have with structured grid generation. While many (most) 3D printers will take an stl format file (which is just a triangulate surface), you still really want the normal CAD formats (parametric) for a couple reasons. Usually for the metal printing processes you have to add support material. This is done more easily / accurately with something other than an stl. Also you want to be able to use the part in larger assemblies, and this is likely to go better using a native CAD format.
  • GE Jet Engine Bracket v1.5, Topology Optimized Bracket - V4, by Igor Lins e Silva
  • GE-jet engine bracket-opti-design-phase 1, by Cheng.Li
  • Engine Bracket V2.1, by Igor Lins e Silva
  • GE Challenge, by Charlie Pyott. I like this one because he uses a lattice, which reminds me of the octet truss things I was working on previously. Charlie also has a website with other interesting designs.


  1. Topology optimization as a 'killer app' for moving simulation sooner in the design process:
    While I’m actually a big fan of moving simulation up in the product development process, I don’t think any of today’s CFD or FEA programs are likely to be the killer-apps for upfront simulation. Rather, I think structural topology optimization might be it. If only more people knew what it was.

    The best way to understand topology optimization is to compare it to FEA. With FEA, you define the shape of a part, add loads and constraints, then run the simulation, to see if you got the shape right. With topology optimization, you define the loads and constraints, specify what space the part has to fit into, then, when you run the simulation, the software will tell you what shape the part should be in order to satisfy constraints like minimum weight or maximum stiffness. It is the opposite of FEA.

    Structural Topology Optimization: the Opposite of FEA

  2. Here's how the entries in the contest will be evaluated:

    1. GE will compute the weight of each entry and the entries will be ordered
    from lightest to heaviest.

    2. Starting with the lightest entry and proceeding until we have ten
    a. We will ensure that the entry fits within the exact faces of the original
    GE bracket design. Failing this test will disqualify the entry. Please note:
    Due to differences in CAD software, they will allow deviations up to 0.010"
    outside of the original part envelope.
    b. We will perform four separate finite element analyses on each geometry
    using the loads described in the rules. We will use ANSYS, and we will
    assume an isotropic linear-elastic material model with Young's Modulus = 110
    GPa = 15,950 ksi and Poisson's Ratio = 0.31. The bolts, pin and mounting
    surface will be treated as rigid bodies.
    c. The maximum vonMises stress will be calculated; if this stress exceeds
    the yield stress (903 MPa = 131 ksi) the entry will be disqualified.