Optimisation in the FEA sense of the word is using a series of iterative simulations to achieve an improved design. The most common and visible example of this is Topology Optimisation, where a mesh of the available volume for a concept component is eaten away against a mass target or against a target performance characteristic.
Topology optimisation has been around for over 20 years and has a resurgence of interest as it lends itself to lightweighting in combination with 3D printing. This is a good thing for manufacturers, but we shouldn’t lose sight of the other methods of optimisation for they can all add value to our engineering.
Several other types of optimisation are available in MSC Nastran. These can be used separately or in combination to optimise several components of a system using different techniques in the same simulation.
Topography – for sheet bodies, introducing raised areas into a panel to stiffen it against a target
Topometry - sizing the local thickness on an element-by-element basis to achieve a target performance
Shape – assigning design variables to abstracted properties such as thickness, fibre orientation and beam section dimensions and adjusting them to achieve a target
Basis Vector – modifying the location of nodes using a field to achieve a target
How does it work?
All these methods are based on the same principle. We have a number of design variables and an objective. We test the sensitivity of the objective to perturbations in each design variable and then adjust those that drive the objective towards our goal. The objective is most often “minimise the mass”.
We can apply constraints to the design to achieve our engineering goals. Targets can be expressed as “maximum deflection of 1.25mm”, “maximum stress of 145MPa” or “first mode frequency above 50Hz” for example and are used to avoid concepts that will not satisfy our performance criteria.
The optimisation will progress, altering design variables towards the target while obeying the constraints you have. It keeps going until there are no possible changes that will not violate the constraints or that any change in design variable results in a move away from the objective. We can improve the success rate of this using Global Optimisation where we optimise from multiple starting points to avoid settling for a local minimum of the objective.
When do we use it?
The most benefit from optimisation comes at the very beginning of the design process. If your product design is an evolution of previous versions, it may be that backing off and asking the question “what’s the best use of mass I can make in this available space?” via a topology optimisation might give a new, improved, direction. Or adapting topometry optimisation to derive concept layups for a composite structure by asking Nastran to select the thickness of different plies across a part to arrive at a concept zone layup.
There’s also value in exploring alternative manufacturing techniques. An expensive fabricated panel using bonded/welded ribs for reinforcement could be replaced by a single operation, net shape, pressing designed to meet your criteria using topography optimisation.
What else can it do?
The objective doesn’t have to be mass reduction, it can be constructed from any of the available responses (or customised to use external responses). One example where a non-mass objective was used was for a customer wanting to improve their model correlation. A motorsport team had a vehicle model that wasn’t correlating well with physical tests of stiffness. The powertrain was part of the overall structure and the bolted joint between the gearbox housing and engine block was represented by bushing elements at each bolt.
There was evidence that the stiffness assumed for the bushing was not correct. An objective function was written that represented the error term between measured and predicted deflection under torsional loads, the bushing stiffnesses assigned design variables and the optimisation was used to minimise the error term, arriving at a better model for the bolts.
Similarly, another Nastran user was able to automate the selection of dampers to eliminate unwanted sway response in a bridge structure – a discrete variable option in Nastran allows the selection of variable values from a pre-defined list that might represent, for example, commercially available dampers.
I’m interested, what’s the next step?
If you think you might have a problem that can be addressed, or think you might want to investigate what this technology could do you for please get in touch. We have lots of experience with this and can discuss your requirements and give you a demonstration of how this could be implemented in your environment.