Helium, a non-renewable strategic resource, demands efficient purification technologies to safeguard national resource 

security. To address the technical challenges of helium extraction from liquefied natural gas boil-off gas (BOG), this 

study developed a novel pressure swing adsorption (PSA) process using Aspen Adsorption software. Through systematic

adsorbent screening, the CMS-5000-A carbon molecular sieves (CMS) with a pore size rangez,2 of 0.5~0.7 nm was 

selected, whose surface oxygen-containing polar functional groups enhanced interactions with the quadrupole moment of 

N2, achieving an exceptional N2/He adsorption selectivity of 87.5:1. A multi-scale coupled model integrating mass, energy, 

and momentum conservation equations was established to systematically investigate the effects of operational parameters 

on separation performance. An extended two-column PSA scheme incorporating pressure equalization and vacuum desorption

steps was proposed. For the first time, Bayesian algorithm-enabled multi-parameter co-optimization of six key operational

parameters (feed flow rate, pressure, bed height, porosity, packing density, temperature) was realized, overcoming 

the limitations of traditional single-factor optimization in handling parameter coupling effects. Bayesian algorithm-enabled 

multi-parameter optimization revealed that an adsorption pressure of 0.3~0.4 MPa increased helium recovery by 15%, 

while a 1.0~1.1 m adsorption bed height ensured product purity≥99.999% with energy consumption maintained at 0.36 

kWh/m3

. Compared to the single-column system, the two-column configuration improved helium recovery by 5.06%. 

Notably, the novel PSA process reduces energy consumption by up to 60% compared to conventional methods, providing 

a data-driven and quantifiable design paradigm for lean helium purification and presenting an efficient and energy-saving 

technical solution for industrial-scale helium purification.