The selective oxidation of alcohols to aldehydes is a vital reaction in organic synthesis and chemical production. However, this reaction exhibits a complex reaction network influenced by temperature, time, and pressure conditions. A study published in Frontiers of Chemical Science and Engineering presents a hybrid modeling strategy that addresses optimization challenges through artificial intelligence-assisted process optimization.
The research team focused on the iso-octanol oxidation process, where iso-octanol mixed with unreacted iso-octanol enters a fixed-bed reactor to produce iso-octanal and hydrogen byproduct. The gas-liquid mixture is cooled and separated in two flash tanks. The liquid products enter a vacuum distillation tower, where 99.5 percent pure iso-octanal is obtained at the top, while unreacted iso-octanol is recycled back to the reactor.
To overcome the scarcity of experimental data due to long operating cycles and hydrogen safety concerns, the researchers constructed a precise mechanistic model using Aspen Plus based on existing research data. The PENG-ROB equation of state was used for property calculations, and a plug flow reactor was selected as the reaction unit. When the mechanism model converges, the calculated data satisfy material and energy balance. This mechanism model generated 1,845 sets of reliable training data.
Four machine learning models were investigated as surrogate models: deep neural network, deep belief network, extreme gradient boosting, and random forest. These surrogate models were integrated with four multi-objective optimization algorithms including NSGA-II, NSGA-III, MOEA/D, and RVEA to identify optimal operating conditions balancing production costs and carbon dioxide emissions.
The results demonstrated remarkable efficiency improvements. Surrogate models achieved computational speeds exceeding 400 times those of traditional direct simulation methods. Among the 16 combined optimization models tested, the XGBoost-RVEA combination exhibited the fastest computational speed, achieving solution times as low as 14.3 seconds.
The optimization outcomes revealed significant environmental benefits. Primary energy demand was reduced by 10 percent to 5.09 gigajoules per hour after optimization. Greenhouse gas emissions achieved a substantial 12 percent reduction, leading to 0.29 tons of carbon dioxide equivalent per hour.
Further analysis indicated that the dehydrogenation reactor and distillation separation unit were the primary energy-consuming subprocesses after optimization, accounting for 54 percent and 46 percent respectively. Carbon dioxide equivalents dominated greenhouse gas emissions, contributing 64 percent and 36 percent respectively.
Correlation analysis using SHAP interpretability methods revealed that reaction temperature significantly influences both total production value and greenhouse gas emissions more than other variables including residence time, reaction pressure, distillation tray numbers, and feedstock positions. Notably, reaction temperature exhibited a strong negative correlation with production costs, suggesting that increasing temperature reduces separation costs by enhancing feed vaporization rates to the distillation column.
This work provides impetus for the engineering application of octanol dehydrogenation to octanal production and serves as a reference for the industrial application of alcohol dehydrogenation to high-value aldehydes.
DOI
10.1007/s11705-026-2630-1