Dr Ralph Sundermeier works in the Technical Design Department of Volkswagen, where he is responsible for the interior calculation within car projects and mainly for the development of calculation methods.
Dr Sundermeier was Keynote speaker at NAFEMS recent World Congress in Salzburg, Austria. His presentation entitled 'Frontloading by calculation - Evolution or Revolution?' shared Volkswagen's experience of frontloading calculations in the early stages of product development.
Below is an extract from our interview with Dr Sundermeier at NAFEMS World Congress where we discusses his presentation in more depth and where he believes the future of simulation to be.
The risk is always that quality might not be assured. You should make sure that the right things are done at the right time. If we were to make calculations in the early phase with data of too poor quality, then make decisions based on the results of the simulation runs, and those decisions were to lead to wrong or faulty functions of the cars, then this would be a very bad thing. So I believe that front-loadingis the correct step, but in the future we have to change a lot of things to do this successfully.
If you couple methods, you get new needs. These new needs will be transferred to those who support the codes and deliver the software. This is what is changing. This initiates the changes. And when we change processes, a similar thing happens, because the usual tools on the market are made for special situations. If you change the situation by front-loading, for example, somehow you have to change the tools as well. Some tools may be changed by evolution, but I feel a lot of them have to be changed by revolution – in-depth changes need to be made.
Nowadays it's impossible to imagine car developments being completed in just one year without the use of simulation. So, simulation has had a strong impact on car development processes and on the development processes for cars that we already have in the market. I can’t imagine how that could succeed without this change driven by simulation.
I expect to see thisimpact to a greater extent . This will result in a series of changes. I think that in the future we will certainly see progressive changes to methods and processes.
This depends very much on the product you plan to design. When it comes to cars, we have continuous changes driven by laws, global laws or local laws, laws related to car safety, fuel economy and energy consumption, for example. From this perspective, there is no special technique that is being under-utilised; all are used in development,and they must develop in parallel.
In my opinion, the biggest area in need of revolutionising is the attempt to utilise over a hundred thousand cores in parallel per simulation run. The technology of a lot of software codes on the market is twenty to thirty years old; they were developed under completely different circumstances, and they have been updated several times. In my opinion, however, it is time to ask the question of whether to programme the same functionality in a completely new way by looking forward, so that a lot of cores can be used in parallel in each simulation run. In the past,the entire set of input data was always available. When you do front-loading,if you want to make a decision about a car project without having the car data completely designed, you have a chicken and the egg problem. Here, you must have approach esthat work even with incomplete data or that work with data of low quality, and once again this is new. These things have to be developed and tools are needed to fill the gap where data are missing. That's new, and that's caused by process changes. In the past, no one would try to perform calculations without data, but since time to market is a driving factor today, there's no other way to do it – today or in the future.
The biggest emerging topic,in my opinion, is simulation coupling. This is becoming an increasingly bigger issue. People doing a calculation job have more and more problems in reviewing whether their calculations and whether their process chain that they are currently initiating are actually working. If you consider a shoe maker for example, he's able to make a complete shoe because he's doing every step needed to make a shoe with his own hands. At a car company, the production process is distributed among some thousand hands, and we have a lot of quality assurance systems to assure the quality of the car at the end of the process. In the virtual world, in the past the behaviour of the people was like that of the shoemakers, it's all about setting up a job, looking at how it works and looking at the results, all in the hands of one person. There were no types of security mechanisms or quality assurance systems. For the future, if we make the process chains increasingly longer, and if people do not have the opportunities to look deeply into the process because they do not understand all of its details, then we have to develop assurance systems, self-monitoring systems in programme loops that change the behaviour of those loops so that individual engineers can be assured that they are getting some help, that in the end the results of their calculations runs are credible, and the results are free of computer-related or statistical problems.