Iranian Journal of Chemical Engineering (IJChE)

Abstract The shell of walnut fruit is an agricultural byproduct generated in considerable amounts upon the processing of fruit to obtain its valuable kernel. Walnut is recognized as an appreciated tree nut belonging to the Juglandaceae family. The fruit is mainly composed of a kernel, seed coat, shell, and husk. In recent years, the shell portion of walnut fruit has been widely used in the development of different high–value materials. In this regard, high carbon and low ash contents of the walnut shell (WS) make it a suitable material and inexpensive waste agricultural product for the activated carbon (AC) preparation. In this review, recent developments in the preparation of AC from the shell of walnut fruit and the characterization methods are highlighted. Additionally, the applications of AC mainly in the removal of different hazardous materials such as heavy metals (HMs) ions, dyes as well as pharmaceuticals and other dangerous materials are comprehensively discussed.

Abstract Hydrocarbon reforming processes are important routes for the production of synthesis gas and hydrogen. However, the complex interactions among operating conditions, reaction kinetics, transport phenomena, and catalyst properties make process prediction challenging. This review examines the applications of machine learning (ML) and artificial neural networks (ANNs) in the modeling, prediction, and optimization of hydrocarbon reforming processes, with emphasis on methane reforming. Studies employing ANN, Random Forest, Support Vector Machine, Gradient Boosting, XGBoost, and CatBoost were reviewed and compared based on statistical metrics, validation strategies, dataset size and quality, overfitting, and model generalizability. The results indicate that model performance depends on the reforming process, target output, data structure, and validation strategy, and no single algorithm is universally superior. Although MLP has shown better statistical performance than RBF in some dry methane reforming studies, independent test datasets may reveal limitations in ANN generalizability. Ensemble methods, including Random Forest, CatBoost, and Gradient Boosting, have demonstrated competitive performance, although the best model varies with the predicted output. Furthermore, SHAP and Partial Dependence Plots can identify influential factors such as reaction temperature, feed ratio, and metal loading. Overall, rigorous validation, independent testing, overfitting control, and interpretability are essential for reliable ML applications in reforming processes.