x
Our website uses cookies. By using the website you agree ot its use. More information can be found in our privacy policy.

Diffcrash

DiffCrash_klein_02.jpg

Diffcrash is a tool for analyzing huge datasets and identifying non-robust design regions. With its help, you can suggest design changes that ultimately lead to a more robust model and/or product.

Predictability in simulation results

Even very small changes applied to a simulation model may cause a large variation in the calculated results. This is particularly true for numerically challenging problems like, e.g., car crash simulations that include many buckling bifurcation points. The reasons for result dispersions occurring within a simulation model could be physical and/or numerical, but how do we know which one it is? 

The way forward

Independent of model size, Diffcrash can analyze a set of simulations and highlight the areas with a large scatter of the simulation results in the model. With its advanced data reduction tools, it is possible to correlate the behavior between different regions in the model, i.e., to find the root cause of the observed variations. With this information, it becomes much easier to come up with an appropriate design change that would increase the overall robustness of the model and/or product.

A typical flow chart when working with Diffcrash is illustrated below.

workflow.png

You can also visit SIDACTs web page to read more about Diffcrash.