Abstract
Hypergravity significantly enhances gas–solid mass transfer, yet systematic equilibrium data and reliable predictive models remain scarce. To address this gap, we conducted a comprehensive set of 256 adsorption experiments for silica gel and water vapor under a full-factorial design, spanning hypergravity factors (β) from 0 to 40, temperatures (T) from 298.2 to 308.2 K, and relative humidities (φ) from 20% to 95%. Compared with the fixed bed (β = 0), hypergravity at β = 16 increased the adsorption capacity by 39%.
Six classical adsorption isotherm models were evaluated, among which the Sips and Dubinin–Radushkevich (DR) equations delivered the best fits. Based on adsorption potential theory, we developed a modified DR model by correlating the affinity coefficient with T, β, and φ, yielding an explicit functional relation q = f(T, β, φ). This model preserves clear mechanistic interpretability while achieving high accuracy (R2 = 0.992–0.999) and stable performance within a practical hypergravity range of β = 10–50, making it particularly suitable for mechanistic analysis and process simulation under moderate-to-high hypergravity conditions.
Complementing this theoretical approach, we also proposed a dimensionless correlation and two machine learning models—support vector machine (SVM) and backpropagation (BP) neural network. The BP model, trained on a combined dataset comprising experimental data, modified DR predictions, and characteristic-curve extrapolation results, attained the highest predictive accuracy (test R2 = 0.9994).
Collectively, the integrated framework—combining the modified DR, dimensionless, and machine learning models—offers a reliable suite of tools for predicting hypergravity adsorption isotherms and provides a methodological blueprint extendable to other adsorbent–adsorbate gas–solid systems.
扫码关注我们
求助内容:
应助结果提醒方式:
