5 Replies Latest reply on Mar 8, 2011 1:02 PM by Chris Michalski

    Confidence in results

    John Sutherland

      I admit to being surprised by this marketing blurb for newly released Caterpillar road prime movers:-

       

      "Computational Fluid Dynamics, a computerised application used to optimised design, was used in the early stages of the process; 1/8 scale wind tunnel testing was used for basic shape development; as well as full scale wind tunnel testing."

       

      Is this typical in high tech engineering industries?

        • Re: Confidence in results
          Bill McEachern

          Not exactly sure what you are asking or commenting on. 

            • Re: Confidence in results
              John Sutherland

              The belt and braces approach to design; CFD virtual prototype + 1/8 scale wind tunnel physical prototype + full size wind tunnel physical prototype.

                • Re: Confidence in results
                  Dan Cook

                  It should be. For anything where a human life is at risk, it better be.

                   

                  CFD is only one tool in an engineers kit, and in many ways it is a very dangerous tool. You can make very pretty pictures

                  that "look OK" that are far from physical reality.

                   

                  Scale models are just that, models. So much can be learned from them, but they also aren't "real" in certain ways.

                   

                  Full scale testing has always been part of the product development that I have worked on at Reynolds Metals, Apple, and

                  Cirque du Soleil.

                  • Re: Confidence in results
                    Bill McEachern

                    In my view it comes down to risk and uncertainty. Getting close enough with CFD depends on exactly what the probelm is and ones validated experience with that class of problem and what the risks are of missing  and by how much. If there is a lot of money at stake or safety risks you obtain better validation depending on how much it costs relative to the risks of course. A wind tunnel or other experimental apparatus has its place for resolving certain things and as does a CFD simulation. Using both is a good way to ensure you have a simulation model that has all the adequate physics as oppose to one that you assume has all the significant physics and/or geometery included. Just because it makes a pretty picture doesn't mean it is a decent or adequate reflection of what you are aiming at. Individuals and companies need a process to ensure that simulation results are relied upon to the degree to which they are fully understood and are supplmented with experimental data as required to ensure that the confidence level achieved is appropraite for the risks being assumed. Sometime belt and suspenders is the only way to go, sometime just the belt and some time just the suspenders works.

                    • Re: Confidence in results
                      Chris Michalski

                      Every simulation is an approximation of a physical system.  The software capabilities have increased, but if you've ever watched the Tacoama Narrows Bridge collapse you realize that an inaccurate simulation may have rounded the decimals that added up to those harmonics.

                       

                      As Bill said, if there is a danger of physical harm or loss of life I don't know a PE who's going to sign off on something with only digital simulations, some things require real models and empirical data.  Many can get away with either a scale or a full size mockup, but I've been using Flow since 2004 and I still find peculiarities that keep you from trusting everything it says.