Speaker
Description
Self-pressurization phenomenon due to heat leakage into liquid hydrogen (LH2) tanks, is considered the primary challenge and dominating factor for its long-term storage and safe operation. Until now, there is still a difficulty in accurately predicting the pressure evolution in such tanks. Despite the wide-spread use of CFD as a mature-developed non-equilibrium modeling approach, intrinsic complex nonlinear phenomena are still hindering its applicability to real practical situations due to assumptions and limitations. In light of this, artificial neural network (ANN) as an intelligent modeling approach is first hybridized with CFD model to provide a new synergistic approach called (ANN-CFD) hybrid model for the sake of improving the predictions of the single CFD model. This hybridization has been implemented through integrating the CFD model with distributed neural networks for improving the CFD model outputs. The hybrid synergistic approach has been applied to the multi-purpose hydrogen test bed (MHTB) as one of the most important LH2 test facility available from literature work. Here, the ullage pressure would be predicted as a response to the hold time under given operating conditions of heat flux, initial filling ratio and initial operating pressure. Compared to experimental results, the predictions of the pressure evolution from the hybrid synergistic model has been proven to be desirable in its accuracy with an average error of 0.5 % and a maximum error of 1 % better than the CFD model solely which presented a significant deviation. The present work has revealed the strong capability of the new synergistic ANN-CFD hybrid model in improving the predictions of CFD model significantly with robust, accurate and consistent results.
| Submitters Country | China |
|---|
