Como você pode acelerar as ferramentas de software de engenharia aeroespacial?

Parallelizing your code is an excellent way to speed up your simulations without sacrificing any high-fidelity parts of your simulation. However, it must be known that parallel computing often gives diminishing returns. Running your code across 10,000 processors will most likely not be much faster than 8,000 processors. Running your code across 8 processors, however, will be orders of magnitude faster than running your code in serial.

This in my opinion is the most important aspect to be considered as majority of tools involve complex computations requiring huge time, efforts and resources to have substantial results. While distributed computing could optimize time, cloud computing addresses the challenges towards the memory requirements. An example in my experience was to have computational results across different geometrical configurations of an aircraft wing. With parallel computation of this analysis, the time required was reduced by 10x the initial time of multiple weeks.

Optimization of code involving mathematical computations could be two-fold. While the right programming techniques should definitely be used, more important is to implement the most optimized methods for computations. This should be thoroughly considered as any redundant calculation could not only be unnecessary but also would lead to an increased computational time. The scope and final outcome of any analysis should always be thought about while considering what computational technique to be used.

While verification and validation is necessary, it may not always be of a substantial relevance when it comes to validating the tools used for aerospace or mechanical analysis. This is especially true for the tools used for complex calculations. For such cases the scenarios might not always be consistent and well-defined to start with. The same tools are used to perform calculations for multiple cases. What helps in such cases is while the tool development is in progress, to have the test conducted with expected and actual outcome. This in turn could help in doing any corrections needed at an earlier stage itself and to have some initially defined test cases to benchmark the performance of software tools.

Automating certain work flows according to your requirement and using an AI tool to get simulations or solutions to close to your desired answer and then spending time on the final solution is a sure shot way to optimize time taken to generate results, be it CFD or FEA. There are also certain papers on how Machine learning can approximate calculations to fit a solution in much less considerable time than running the numbers. A combination of automation and AI will definitely improve software for aerospace engineering

One common mistake in design is that of design to manufacture this is one of the aspects which must be part of any initial design in mechanical engineering. A skilled engineer is aware of pitfalls that can happen down the road with manufacturing and often design his parts so that it can be easily manufactured, a build verification tool in software can help find difficult to manufacture areas and flag it early on in the design process ,thus eliminating rework.

What tools the author has been using? Any commercial tools that are prone to error will not stay in business long. With in-house/University spin-out tools where there are errors and they are cumbersome but not with commercial tools that have been on the market for some time. The home grown tools are often developed piecemeal to solve a particular problem and then updated to add in more capability. That is how they become slow, unwieldy, and prone to errors. The evolution of the tool, combined with the fact the author is often the initial user means that best practices are not followed. That is a stark difference to a software company setting out to create a product. Whilst still a generalisation it is less so than the originators

Start by developing a plug-and-play interoperability framework that lets the user decide which third party tools to use, while working with models and results in an unified format. This way, the modeling and simulation framework is scalable and different third party products can be validated against each other.