.png){kind=icon}
.png){kind=icon}
Return
CFD-ML integrated multi-objective optimization for n-butane partial oxidation reactor
.png){kind=icon}
PRE
AI
DOI：10.1016/j.cjche.2025.06.019.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This study develops a CFD-ML integrated framework to achieve multi-objective optimization for the partial oxidation of n-butane to maleic anhydride (MA). A reactor-pellet coupled model was established to investigate the effects of four key operating parameters, revealing that inlet temperature dominates reactor performance by increasing MA yield from 35.0% to 37.6% while sharply raising hotspot temperatures by 42 K. The coupled model was then employed to generate 621 cases for training machine learning models, among which the Gaussian Process Regression (GPR) model exhibits superior accuracy. The GPR model was further integrated with the genetic algorithm to generate Pareto-optimal sets. The results indicate that a critical inflection point is identified on the Pareto front, and once this point is exceeded, even a slight increase in MA yield could lead to a sharp rise in the reactor hotspot temperature, thereby increasing the risk of thermal runaway.
Keywords:
CFD-ML integration
multi-objective optimization
n-butane partial oxidation
maleic anhydride
Gaussian Process Regression
Pareto-optimal solutions
Journal
IF:
3.7
Papers:
5.2K
Citations:
1.1W