MACHINE LEARNING AND CLUSTERING-BASED ANALYSIS OF METHANE REFORMING PROCESSES UNDER DIFFERENT OPERATING CONDITIONS

Abstract

This study presents a data-driven framework for analyzing methane reforming processes under different operating conditions by integrating supervised machine learning with unsupervised clustering. Methane reforming is governed by complex nonlinear relationships among temperature, pressure, and reactant feed ratios, which significantly influence methane conversion, hydrogen yield, carbon yield, and product gas composition. To address this complexity, the proposed approach combines predictive modeling with operating-regime discovery in order to provide both quantitative accuracy and qualitative process insight. In the predictive stage, machine learning models were employed to estimate key methane reforming outputs from process operating variables. Comparative evaluation showed that the artificial neural network achieved the best overall performance, with high coefficient of determination and low prediction errors relative to alternative models. In the clustering stage, K-means was applied to classify operating conditions into distinct groups, and the optimal number of clusters was identified using the elbow criterion. The clustering results revealed clearly separated operating regimes corresponding to low-, moderate-, and high-performance process behavior, particularly in relation to hydrogen yield and methane conversion. Correlation analysis further confirmed strong positive effects of temperature on conversion and hydrogen production, while pressure exhibited comparatively adverse influence under several operating conditions. The integrated results demonstrate that the proposed framework is effective not only for accurate prediction of methane reforming responses but also for uncovering hidden patterns within multidimensional process data. Overall, the study establishes that combining machine learning and clustering offers a practical and interpretable methodology for methane reforming analysis, with potential value for process screening, optimization, and intelligent operational decision-making in hydrogen and syngas production systems.