I have often been described as a talkative person. It’s not that I love my fits of verbal diarrhoea, but the need to be as specific as possible and to give as many details as I can leads me to either speak in paragraphs or use a particular vocabulary. In the meantime, I require precision when listening or taking in information.
It’s mainly because I’m aware of the error range within transmission and exchange. It’s inescapable. In my mind’s eye therefore, the narrower I can make the range, the better it is. This goes for any kind of transmission, from a simple “how are you?” to a more sophisticated “can you explain this to me?”
Being an Orwellian fan in the old days, I’ve always realised the strong hold language has on the thoughts’ process and data intake/output. Therefore, whenever I’m spoken to, my verbal diarrhoea takes the lead. While I don’t always have the time or will to talk in details or write paragraphs, I do it anyway to reduce, as much as possible, the error range of an interaction.
Finite Elements Analysis
During my engineering student years, I discovered the Finite Elements Analysis and had a stimulating overdrive and a deep fascination/affection for the simple yet powerful practice. Here I am wondering how computers could manage the integral definition of surfaces and every microscopic bit of matter. Then at the same time apply loads and merge this information in order to know where and when stresses and strains will happen and when they may be critical. Here I am lost in how such studies could ever be achieved since every microscopic bit is made of even tinier blocks and so on and so on… I am left facing the idea that it is impossible to solve constraints/flow equations at a very specific point or area. Then, FEA comes to focus as an answer.
Picture a surface, a simple limited surface of 1mm thickness, like a piece of paper. If I ask you how many points are in the surface, you’d probably say infinite, or you would require an estimate of the dimensions to estimate a rough number. When a surface is submitted to a force, a motion for example, our mechanical models suggest that every point is submitted to its share of the force. Therefore, to know how the surface will behave you have to not only know how every point will behave, but how they will interact with each other.
Finite Elements Analysis steps in and suggests the following:
1. Let’s not take into account all the points of the surface and instead take a finite group of points spread through it.
2. Let’s connect those points with squares preferably (or triangles as a second preference) so their interaction can be taken into account and accurately calculated.
3. Let’s distribute the conditions on these points and calculate the outcome.
4. We can then create an oversight of how the surface will behave under specific conditions by comparing how the state of the finite group of points evolved.
Nothing is ever perfect
There is no exactitude in the world, even less in measures and mechanics. Mechanical design engineers are especially aware of such facts and hope to just converge towards the Northern Star which is the theoretically most accurate value. So, if we are to live with such incertitude, why not use a simple model that will give us an estimate and which we can make more or less accurate depending on our needs?
Through experience I came to the realisation that refining the mesh and using a smaller precision does not help in the calculation or the precision of results anyway. The main point of FEM was to deliver optimal results in optimal time depending on the needs of the engineer. By increasing the number of points from thousands to millions and millions, FEA starts requiring more and more time just like theoretical equations. This process then ends up confronting unsolvable matrices and major problems, the least of them being a lack of convergence.
The benefits of local refining
When I read more about the convergence of the mechanical models and simulation, I noticed a certain expression that was brought up every now and then: local refining.
If I come back to the previous example of the surface, let’s say that I know I will apply a certain stress in the middle of it. This place will be a particular source of much needed data since I’d like to know if it’s going to be able to hold the stress and not be plastically deformed and so on and so forth… Consequently, when choosing the set of points where the study will be performed, I can choose to have many localised ones in that critical region, as well as other at a greater distance from each other.
The outcome is a balance between resources, time and analysis needs. Processing doesn’t take too long and the results are generally good everywhere but more accurate when focused on the critical regions: refining according to the needs of the designer is more sensible than refining the overall part.
The simple art of conversation
Now, when it comes to speech, this had a crucial impact on my speech patterns: I’m able to answer with a conscious state. I’m aware of the time imparted and whether I’m required to refine or not, and if so what my interlocutor’s need is exactly. Granted that it’s not always as clear and precise as I’m describing it but now at work, I’m able to not precipitate everything at one answer and point out answers or questions in a more precise way to get a more refined exchange. A refined speech is an interactive one where you interpret the input data (intonations, haste, vocabulary, conditions, emphasis…etc) and come up with an answer that meets basic requirements (the answer and the level of detail required for certain points).
From there on discussion can be further developed depending on the exchange. There is no need to hold a panel, to give a lecture or to be barely able to give the main topic. In the trade of mechanical designing, it is tricky enough to understand what a customer is picturing exactly in his/her head. Therefore, to breed a habit of engaging an interactive discussion succinct enough on several points, and refined on others, is a great start at making a first draft towards the final desired model. Details are much appreciated but their directed flow is even better.