The importance of uncertainty quantification (UQ) has been emphasized for simulation validation. However, UQ itself may just confirm that the validation is not conclusive due to large uncertainty. Thus, uncertainty reduction based on uncertainty investigation is important. However, uncertainty investigation and uncertainty reduction can serve as well to expose hidden problems in simulation. This paper aims to share lessons learned from the uncertainty investigation of a horizontal shock tube simulation with three findings. Firstly, UQ of the simulation identified the presence of a large canceling error in the numerical flux scheme AUSM+ as it revealed inconsistencies in the simulation runs. The method of groups of runs was applied to more precisely point out the inconsistency. Secondly, the investigation about the uncertainty due to the inconsistency between the experiment and the one-dimensional simulation exposed lack of knowledge about the problem. It turned out that the influence of the inconsistency was small, which has been unexpected based on expert opinion. Thirdly, the uncertainty in the initial volume fraction was investigated based on rigorous UQ and it reduced the uncertainty and revealed a hidden modeling error in the simulation. With all the efforts, the uncertainty in model error estimation was significantly reduced. That is, the error in the shock tube simulation was clearly seen.
展开▼
