A fully coupled mass, momentum, and heat balance model was constructed in axisymmetric 2D for an ammonia decomposition packed bed membrane reactor using computational fluid dynamics and compared to experimental results for a Pd/YSZ membrane surrounded by Ru/Al2O3 catalyst. In addition to the complex geometry of the fittings used to seal the membrane, a total of 16 design variables were found to be required to fully specify the tubular integrated membrane reactor. Aside from emissivity, all input variables were experimentally determined or obtained from tabulated material property data, and no correlation fitting was performed. Experimental data were obtained for temperatures of 450 °C–530 °C, trans-membrane pressures of 0 MPa–0.4 MPa, and gas hourly space velocity feed rates of 375 h−1 to 620 h−1, chosen such that the ammonia conversion and hydrogen recovery spanned a broad range with maximum values of 99.9 % conversion or 88 % recovery with greater than 99.8 % purity. The model was able to successfully determine the ammonia conversion and hydrogen recovery of the experimental results to within an average error of −0.31 % and −2.4 %, respectively. The purity was determined within +0.0071 %, but experimental detection limitations of the impurities affected the fit between experiment and model of the N2 and NH3 in the permeate. The sensitivity of the model to various assumptions and input variables was explored. Notably, the inclusion of purity is a rarity in published models and this study adds to the discussion on the ability of modeling to predict high purity results. Compared to the relatively minimal effects of using the full geometry, the inclusion of heat transfer and radial diffusion were critical in successfully predicting the experimental results.
Read full abstract