What is FEA? How can I harness its power?
Finite Element Analysis (FEA) in its most basic form is a computational tool which is used to understand how materials and structures behave under loads. The nuts and bolts of the process are extremely complex but essentially boil down to breaking a large problem into many smaller ones which are addressed individually to provide an answer to the entire problem.
For instance, when faced with a structural plate which has a force applied at a given point, instead of treating the plate as a single entity, it is broken up into many smaller pieces known as ‘elements’ which are all interconnected to build up a ‘mesh’ to represent the plate. To understand the behaviour of the plate under the load, the behaviour of each individual element is analysed. This is subsequently used to understand the behaviour of the neighbouring elements until the behaviour of the entire mesh has been solved. At this point the individual results of each element are summed together to produce a solution for the entire plate. In such an example, the ‘solution’ can mean a number of things such as, the deflection of the plate or the stresses within it.
While the intricate process sounds, and indeed is very complex and mathematically involved, fortunately relatively recent developments in FEA software have broadened its appeal, and allowed its power to be harnessed by a much wider user base. With much of the solution process hidden deep with the software, along with welcome developments in user interfaces, FEA has become much more user friendly, and therefore applicable to a wide range of applications – it has finally escaped from its deep seated roots in accademia, and found its way to designers and engineering across many different sectors. And the impacts have been significant. Large complex products comprising many different geometrically intricate parts can now be analysed with a desktop PC with accurate results available within hours, and sometimes seconds.
More recently still, analysis of flexible parts such as rubbers and plastics has become commonplace with FEA software. Previously, most analysis was confined to stiff metals where up to the failure point, the material behaved in a linear manner – that is to say as the relationship between the stress and strain of the material is consistent. With flexible materials such as plastics, this assumption is not always true, and a more complicated relationship between stress and strain must be taken into consideration for an accurate solution. Such developments mean that it is now possible to analyse how plastic parts flex when loads are applied to them – a particularly useful application of this is in the analysis of snap fit fixtures found within almost all plastic products.
So how can the power of FEA be used to improve the efficiency of the product development process? In the traditional product development cycle, concepts are produced, parts are designed as a 3D model (CAD), prototypes are produced, designs are tested, developments are made and parts are retested. And this process continues iteratively until the desired result is achieved. Where FEA really transforms this process is through drastically reducing the length and number of prototypes required to finalise a design. Such a benefit may be best illustrated through an example:
Suppose we are to design a plastic enclosure to house some electronics – the enclosure has a flexible thumb operated latch to allow the user to open and close its lid. In a traditional design process, a design would be produced with the dimensions of the key features such as the flexible latch governed mainly through experience and gut feel. The part would be prototyped, and it may be found that the latch feature cracked and broke when the lid was tested. The designer would then make modifications and tweaks to the design of the latch to thicken it to increase its strength – these modifications are likely to be guesses and it is easy to see that such a modification may lead to the part becoming too stiff to operate easily. And so the process of iterative prototyping goes on, and with it the cost and timescales of the project only increase.
Suppose now, that we have the same design brief, and have access to an FEA software system. As the first CAD designs are being developed, FEA simulations can be run in parallel to assess the behaviour of the parts in real-time. Materials are defined for each part, and assumed loads can be applied to the flexible latch to simulate the desired level of thumb pressure required to operate it. Results from the FEA analysis will then be able to show which areas of the part are likely to undergo large stresses and therefore will be likely to fail. The analysis can also show how the latch will deform when the desired force is exerted on it – such results are invaluable and by designing the parts alongside the FEA simulations we are able to produce a first prototype which is many steps further down the development cycle than would be possible with a traditional design approach. Even more importantly, when making subsequent changes to the parts, our tweaks and modifications are informed with detailed feedback – we can determine for instance how many mm’s thinner the latch would need to be to be operated under a smaller force. Such modifications are no longer left to rules of thumb and guesswork.
While the huge benefits FEA can add to product development are hopefully now clear, it is critically important also to understand its limitations, and how they can be managed to ensure it is being used as effectively as possible.
Understanding FEA’s limitations, and avoiding the common pitfalls
While the potential power of FEA in product development is huge, it is extremely important to understand where its limitations lie, and where it is dangerous to rely upon.
As discussed previously, FEA works on the fundamental approach of breaking down a large complex problem into many smaller, simpler problems. The computational resource required to solve the problem is directly linked to the degree to which it is broken down – that is the quality and resolution of the simulation mesh. Ultimately the quality and resolution of the mesh is probably the single most important feature of a simulation, and it can have a significant impact upon the final result. A coarse mesh means the total number of elements is relatively low. This means the computational resources required to perform the simulation are reduced, but so is the resolution of the mesh meaning that some areas may not accurately represent the geometry part, which may lead to inaccuracies in the simulation result. Conversely, a very fine mesh will accurately represent the part geometry but will require much more computational resource to perform. Most software packages automate the vast majority of the meshing process which simplifies things considerably, and reduces the likelihood of creating an inaccurate simulation, however control of the level of resolution of the mesh is key to efficiently performing a simulation. For instance, in areas away from high levels of stress it may not be necessary to have a particularly refined mesh. On the other hand, in critical areas or where there is complex geometry, a fine mesh is likely to be very important. Managing these conditions together is key to producing a mesh which is refined enough in the critical areas to represent the part accurately, while minimising the total number of elements in the simulation to maximise the efficiency of the process.
FEA analysis is often found to be able to accurately simulate complicated problems where the results can be validated through experimentation – this does however rely on the analysis being set up correctly with valid assumptions made throughout. Where FEA results are found to be inaccurate, and not inline with physical evidence, the problem more often than not lies within the setup of the analysis which has made one or more invalid assumptions about the behaviour of the system. It is therefore imperative that care is taken when setting up an analysis, and we must continually sanity check if the assumptions being made are valid. Assumptions such as where the load is being applied, where the part is held fixed, how two parts interact with each other are all examples of potential pitfalls which must be carefully considered.
How do these limitations affect the application of FEA in the design process? While simulations provide invaluable insights into performance characteristics of parts under predicted loads, what should not be forgotten, is that this doesn’t not remove the need for prototyping entirely. The prototyping process still forms a key part of the design validation, where the assumptions and setup of the simulation can be tested to ensure they accurately reflect the real situation. That being said, FEA can be used as a tool to enter the prototyping stages with a much fuller understanding of the parts and how they can be tweaked to achieve the desired result in a minimum number of iterations.