Mechanism-guided multi-objective optimization of ammonia decomposition in a packed bed membrane reactor

Ammonia (NH 3 ) is a promising hydrogen (H 2 ) carrier, but thermodynamic equilibrium and H 2 -induced product inhibition constrain cracking. We developed a one dimensional packed bed membrane reactor (PBMR) model coupling nickel catalyzed NH 3 decomposition with H 2 -selective separation. The catalytic and membrane submodels were assessed separately against packed bed experiments (500-700 °C, 1-3 bar) and existed H 2 /N 2 permeation data, respectively. Ten thousand Latin hypercube samples supported artificial neural network training, sensitivity analysis and multi-objective genetic optimization of NH 3 conversion, H 2 recovery and output. Temperature dominated performance, whereas improving membrane properties yielded diminishing returns beyond adequate permeation capacity. The model predicted that continuous H 2 removal lowered surface H* coverage, consistent with reduced hydrogen inhibition and improved PBMR performance. At the selected compromise condition, NH 3 conversion, H 2 recovery and recovered H 2 flow rate reached 98.79%, 98.39% and 2.07 × 10 −4 mol s −1 , respectively. This mechanistic and data-driven framework balances competing reactor objectives.