Computer Modelling Pitfall

I encountered some incident about computer modelling recently, which made me re-think the reliability of our support teams in the headquarters.

It somehow gave me a chill down the backbones, as if we were having the blind leading the blind.
I disagree the way our engineers trusted the numerical modellings without much scrutiny.

Incident 1:
When I performed dynamic balancing of the equipment, I discovered a natural frequency of the equipment casing, they (the engineers from structural design group) denied it by saying their simulation doesn't show a natural frequency.

They then told me to remove some of the temporary supports. Though disagree, I went ahead do as they said, I didn't find any significant improvements with or without those extra supports. Few days after that incident, apparently they did a vibration test on the same equipment themselves, and came to me claimed that removing the supports reduces the vibration, and I didn't know how to do it properly.

I looked at the spectrum, it only reduces by 1 mils. The natural frequency spikes still clearly shown in the spectrum, stood high and proud, among all other frequencies.

Incident 2:
As I have started my training with the engineering design group, I sat through with some senior engineers to see how they design the structures and perform the calculations.

Listening to these people boasting how powerful their numerical modelling software is, how complicated their model is -- taking more than 24 hours to calculate etc... made me smile. It is similar to listening to a teenage boy boasting how powerful his calculator is, and how much time he spent using that calculator.

Well, it is not surprising when a simple equipment, which can be well represented by beam elements, was modelled in shell elements. I can see that some of the frame works, which are plain metal sheets, can be modelled with shell elements. But when it comes to support beams, if modelled using shell elements, it would require more work compared to beam elements. Reason being although the support beams are formed by metal sheets, the structural behaviour of that formed metal sheet is no longer demonstrating same behaviour as the plain metal sheets.

I don't even know if the constraints for the shell elements were implemented properly or not. Besides, there is no need to include the plain metal sheets in that particular structural calculation. It doesn't give much effect to the overall structural strength. Having developed such bulky model, do you still expect your super computer to run as fast as it is supposed to be? I could already hear the computer groans.

When I asked about the variables and assumptions that were taken into account, when they develop the models. They said there is nothing, the models they developed are accurately representing the real world scenarios. Even the most powerful computer with gigantic models in Los Alamos National Lab dare not claim that.

When I asked about the potential issues that discovered so far in this product series, which they would like to solve before the next upgraded series. They didn't know what I am talking about. The facial expressions I got from them were priceless -- some offended because I questioned their almighty numerical models, some puzzled because they don't see the need of improvements, because the model tells them this product series is great.

I feel scared, when they told me they would be going to other branch offices around the globe to give training. When they can't even answer my simple and politically correct questions properly.

I know these are senior engineers, there are various other things I could still learn from them. But I somehow starting to realise my strength. Just because I don't speak a word of how much I know, it doesn't mean I don't know a thing. I also understand that it is not my job to criticise /question their work. I could only be more careful next time, when requesting for their assistance in the future.

It is a pity when people forgot that the computer is just a tool, numerical modelling is just a calculation program. That is why people need to verify their models with experiments. Even with experiments, we still need to question both the experiments and computer models. But then I guess, that is too much of a research work for many? Then what are these engineers suppose to do?

It is an even larger pity, when the corporate drove the engineers too hard, that they don't have time to stop and think. Because engineers are the ground level you need to help build your corporate kingdom, you need to make sure they steer your chariot in the right direction, rather than letting them run blindly.