Selective removal of sulphur in liquid fuels for fuel cell applications

Fuels like diesel and gasoil must be desulphurised to extremely low levels before being used as hydrogen source for modern fuel cell applications and to avoid sulphur poisoning of therein used catalysts. A commercial Ni/NiO-sorbent has been identified as being able to remove even refractory sulphur species like 4,6-dimethyldibenzothiophene and the total sulphur concentration is lowered to below 0.2 ppm. The influence of temperature, residence time and level of the sulphur content in the untreated fuel has been investigated in parametric studies. Gas chromatography with mass spectrometric detection (GC–MS) of treated gasoils and diesels reveal which sulphur species are selectively removed and which are left in the fuel. The selectivity and activity of the sorbent can be influenced by the operating temperature. Moreover, GC–MS chromatograms of the breakthrough curves reveal that the sorbent capacity is related to specific sulphur species. Their molecular structure and the alkyl groups at the 4- and 6-positions of dibenzothiophene as well as the C₃-benzothiophenes influence the adsorption and the sorbent capacity significantly.     

Artificial neural network to predict the power number of agitated tanks fed by CFD simulations

The power consumption of the agitator is a critical variable to consider in the design of a mixing system. It is generally evaluated through a dimensionless number known as the power number $N_p$. Multiple empirical equations exist to calculate the power number based on the Reynolds number $Re$ and dimensionless geometrical variables that characterize the tank, the impeller, and the height of the fluid. However, correlations perform poorly outside of the conditions in which they were established. We create a rich database of 100 k computational fluid dynamics (CFD) simulations. We simulate paddle and pitched blade turbines in unbaffled tanks from $Re$ 1 to 100 and use an artificial neural network (ANN) to create a robust and accurate predictor of the power number. We perform a mesh sensitivity analysis to verify the precision of the $N_p$ values given by the CFD simulations. To sample the 100 k mixers by their geometrical and physical properties, we use the Latin hypercube sampling (LHS) method. We then normalize the data with a MinMax transformation to put all features in the same scale and thus avoid bias during the ANN's training. Using a grid search cross-validation, we find the best architecture of the ANN that prevents overfitting and underfitting. Finally, we quantify the performance of the ANN by extracting 30% of the database, predicting the $N_p$ using the ANN, and evaluating the mean absolute percentage error. The mean absolute error in the ANN prediction is 0.5%, and its accuracy surpasses correlations even for untrained geometries.