Қазақстанның мұнай-газ саласының хабаршысыҚазақстанның мұнай-газ саласының хабаршысы2707-42262957-806XKMG Engineering10901510.54859/kjogi109015Research ArticleDesign of a combined isomerization-catalytic reforming unit for increasing gasoline octane number using an AI-based quality prediction modelAbdigaliAruzhanabdigali_aa_1@enu.kzhttps://orcid.org/0009-0005-9636-622XAikenovaNuriya Ye.<p>PhD</p>aikenova_nye@enu.kzhttps://orcid.org/0000-0002-1144-4008ZhakhanshayevaZhansayazhakhanshayeva_zhn@enu.kzhttps://orcid.org/0009-0009-1804-0549Eurasian National University30092026831061181106202621092026Copyright © 2026, Abdigali A., Aikenova N.Y., Zhakhanshayeva Z.2026<p><strong>Background:</strong> The production of high-octane gasoline meeting stringent environmental requirements remains an important task for petroleum-refining industries, including Kazakhstan. At the same time, artificial intelligence and digitalization are increasingly applied to support process design and fuel-quality prediction.</p> <p><strong>Aim:</strong> To design a combined isomerization-catalytic reforming system for producing AI-92, AI-95, and AI-98 gasoline grades using an AI-based model for gasoline-quality prediction.</p> <p><strong>Materials and Methods:</strong> Three commercial gasoline grades (AI-92, AI-95, and AI-98) were characterized by pycnometric density measurement, capillary viscometry, and thermogravimetric analysis with differential thermal analysis. Linear Regression, k-Nearest Neighbors, and Decision Tree models were used to predict research octane number (RON). Kinetic models were developed for light-naphtha isomerization and catalytic reforming, together with material and energy balance calculations, DIP-column design, and reformate-yield estimation.</p> <p><strong>Results:</strong> The Decision Tree model provided exact RON predictions for the three tested grades (MAE = 0.00; R² = 1.00). The isomerization unit with DIP recycle produced an isomerate with RON 87.4, while the catalytic reforming unit produced reformate with RON 98.5 at 500°C and a liquid yield of 85.4 vol%. The calculated blended RON values were 92.1, 95.3, and 98.2 for AI-92, AI-95, and AI-98, respectively.</p> <p><strong>Conclusions: </strong>The combined process configuration enables the production of the full range of studied gasoline grades by adjusting the proportions of isomerate and reformate in the blending pool. The AI-based quality model can be used as a supporting tool for defining target octane specifications during process design.</p>supporting tool for defining target octane specifications during process design. Keywords: gasoline octane numberartificial intelligencemachine learningisomerizationcatalytic reforminglight naphthakinetic modelmaterial balanceenergy balancerefinery designбензиннің октан саныжасанды интеллектмашиналық оқытуизомеризациякаталитикалық риформингжеңіл нафтакинетикалық модельматериалдық балансэнергетикалық балансмұнай өңдеу қондырғысын жобалауоктановое число бензинаискусственный интеллектмашинное обучениеизомеризациякаталитический риформинглёгкая нафтакинетическая модельматериальный балансэнергетический баланспроектирование нефтеперерабатывающей установки1.iea.org [интернет]. IEA. Kazakhstan 2022: Energy Sector Review [дата обращения: 05.05.2026]. Режим доступа: www.iea.org/reports/kazakhstan-2022.2.akorda.kz [интернет]. Ойициальный сайт Президента Республики Казахстан. Послание Главы государства Касым-Жомарта Токаева народу Казахстана «Казахстан в эпоху искусственного интеллекта: актуальные задачи и их решения через цифровую трансформацию» [дата обращения: 05.05.2026]. Режим доступа: node2.akorda.kz/en/president-kassym-jomart-tokayevs-state-of-the-nation-address-to-the-people-of-kazakhstan-kazakhstan-in-the-era-of-artificial-intelligence-current-challenges-and-solutions-through-digital-transformation-1083029.3.Osman W.S., Fadel A.E., Salem S.M., et al. Optimum Design of Naphtha Recycle Isomerization Unit with Modification by Adding De-Isopentanizer // Processes. 2023. Vol. 11, Issue 12. doi: 10.3390/pr11123406.4.Jechura J. Gasoline Upgrading: Reforming, Isomerization, & Alkylation. Chapter 10&11. In: CBEN 409 – Petroleum Refining. Golden : Colorado School of Mines, 2019.5.blog.axens.net [интернет]. Axens. How to Achieve Gasoline Pool Octane Requirement and Boost Light Cut Naphtha Contribution with an Environmental-friendly Process? [дата обращения: 05.04.2026]. Режим доступа: blog.axens.net/how-to-achieve-gasoline-pool-octane-requirement.6.kbr.com [интернет]. KBR. MAX-ISOMTM Maximize octane using innovative catalytic distillation principles [дата обращения: 05.04.2026]. Режим доступа: www.kbr.com/sites/default/files/documents/2023-09/refining_maxisom_2022apr.pdf.7.Lin X., Yin X., Shitao Y. A Trustworthy Artificial Intelligence Framework for Predicting Gasoline Octane Loss Using Sparse Autoencoder and Stacking Ensemble Learning in Petrochemical Processes principles [дата обращения: 05.04.2026]. PrePrint, 2025. doi: 10.20944/preprints202504.2593.v1.8.Bounaceur R., Barthélemy N., Delort N., et al. A Multimodal Learning Model based on a QSPR approach for the estimation of RON, MON and CN, for any C, H, O hydrocarbons // Fuel. 2025. Vol. 381, Part C. doi: 10.1016/j.fuel.2024.133438.9.Chen Zh., vom Lehn F., Pitsch H., Cai L. Design of novel high-performance fuels with artificial intelligence: Case study for spark-ignition engine applications // Applications in Energy and Combustion Science. 2025. Vol. 23. doi: 10.1016/j.jaecs.2025.100341.10.Pedregosa F., Varoquaux G., Gramfort A., et al. Scikit-learn: Machine Learning in Python // Journal of Machine Learning Research. 2011. Vol. 12. P. 2825–2830.11.Fix E., Hodges J.L. Discriminatory Analysis. Nonparametric Discrimination: Consistency Properties // International Statistical Review. 1989. Vol. 57. P. 238–247. doi: 10.2307/1403797.12.Breiman L., Friedman J., Olshen R.A., Stone C.J. Classification and Regression Trees, 1st ed. New York : Chapman and Hall/CRC,1984.13.Mohamed M.F., Shehata W.M., Abdel Halim A.A., Gad F.K. Improving gasoline quality produced from MIDOR light naphtha isomerization unit // Egyptian Journal of Petroleum. 2017. Vol. 26, Issue 1. P. 111–124. doi: 10.1016/j.ejpe.2016.02.009.14.Benavides A., Zapata C., Benjumea P., et al. Predicting Octane Number of Petroleum-Derived Gasoline Fuels from MIR Spectra, GC-MS, and Routine Test Data // Processes. 2023, Vol. 11, Issue 5. doi: 10.3390/pr11051437.15.Volkova G.G., Reshetnikov S.I., Shkuratova L.N., et al. n-Hexane skeletal isomerization over sulfated zirconia catalysts with different Lewis acidity // Chemical Engineering Journal. 2007. Vol. 134, Issues 1–3. P. 106–110. doi: 10.1016/j.cej.2007.03.079.16.Ivanchina ED, Chuzlov VA, Ivanchin NR, et al. Mathematical modeling of the process catalytic isomerization of light naphtha // Petroleum and Coal. 2019. Vol. 61, Issue 2. P. 413–417.17.Seader J.D., Henley E.J., Keith Roper D. Separation Process Principles. 3rd ed. Wiley, 2010. 848 p.18.Smith R. Kinetic analysis of naphtha reforming with platinum catalyst // Chemical Engineering Progress. 1959. Vol. 55. P. 76–80.19.Ancheyta-Juárez J., Villafuerte-Macías E. Kinetic Modeling of Naphtha Catalytic Reforming Reactions // Energy Fuels. 2000. Vol. 14, Issue 5. P. 1032–1037. doi: 10.1021/ef0000274.20.Rahimpour M.R., Esmaili S., Ghalehghazi N.B. A kinetic and deactivation model for industrial catalytic naphtha reforming // Iranian Journal of Science and Technology, Transaction B: Engineering. 2003;27(B2):279–290.