# School of Underlying Conditions examines when a flow simulation may be stopped

> A paper from the School of Underlying Conditions asks whether stopping a flow simulation once its average looks settled spoils the interval reported afterwards, and finds the trouble lies in how the interval is computed.

> There will be other simulations, and each of them will have to end at some point. We hope to meet that moment a little better prepared than we were, and we are grateful to every run that was allowed to finish.
>
> --- Dr Vigdis Emenike, Senior Lecturer, School of Underlying Conditions

Dr Emenike is the lead author of _Stopped When Flat: Optional Stopping, Batch Length and the Coverage of Time-Averaged Quantities in a Fluid–Structure Simulation_, released this week with Dr Brynjar Obiora, Senior Research Fellow. The paper takes up a practice familiar to anyone who has averaged a quantity over a simulated flow: ending the run when the average appears to have settled.

To test it, the authors ran one simulation 240 times and gave every stopped run a partner of the same length that had been stopped for no reason of its own. Run and partner performed alike, with 803 of 1,200 reported intervals containing the reference value on each side. The intervals themselves were too narrow in both cases, and the paper traces that to a later step in the calculation, where a widely used convention took no account of a slow movement in the flow.

The University counts the paper among the clearer expressions of its commitment to finishing what it starts. Dr Obiora said the questions that arrive after a paper of this kind "are generally better than the ones that came before it, and the School has set aside some room for them".

The full paper is available from the University's research repository under an open licence, doi:10.5555/slop.mdga34.