Sulfur recovery units (SRUs) at oil refineries are complex, multi-parameter systems whose Cold Bed Absorption (CBA) reactors operate under incomplete and uncertain process information, which limits conventional control. This study develops an engineering neuro-fuzzy decision support system (DSS) for optimizing sulfur production at the Atyrau oil refinery (Kazakhstan) under uncertainty. The system integrates a dynamic mathematical model of the process, an expert-derived fuzzy rule base, a neuro-fuzzy estimator of sulfur yield and quality, and a multi-criteria fuzzy optimization block based on the Bellman–Zadeh approach, together with an explicit operator-facing decision algorithm and a software interface. The models were identified and validated on 200 operating-log records of the unit. These records show that the effect of reactor temperature on the sulfur quality index reverses sign with reactor pressure — the correlation is −0.57 below 0.79 kgf/cm² and +0.75 above it — so that a model additive in the two variables cannot represent the relationship: linear regression on temperature and pressure attains a cross-validated R2 of only 0.06 for the quality index, whereas the neuro-fuzzy estimator attains 0.83 (RMSE 0.72 points, MAPE 0.74%). For sulfur yield, where the relationship is close to linear once pressure and throughput are accounted for, the corresponding figures are 0.71 and 0.85 (RMSE 0.33 t/day, MAPE 1.35%). The same ordering is reproduced on an independent hold-out test set and after removal of the records lying at the unit’s capacity and regulatory bounds, and the expert rule base was revised into an asymmetric form accordingly. A worked application of the optimization block to three operating situations raised the degree of joint satisfaction of the yield, quality and energy criteria from 0.08–0.25 to 0.49–0.57; in one of the three the recommended action is to reduce process severity and save a third of the reheat duty rather than to increase output. The resulting DSS supports the operator in analyzing the current process state, evaluating alternative operating modes and forming recommendations, while the final decision remains with the operator.

This article presents a hybrid methodology for the intelligent control of the reactor-regenerator unit of a residual fluid catalytic cracking (RFCC) plant under conditions of uncertain input information. The main challenge in catalytic cracking process control is the instability of feedstock composition and the delay in laboratory data on product quality, which complicates the operational management of the RFCC process while maintaining the required gasoline quality. The objective of this study is to support effective management of the gasoline production process by maximizing the yield of high-quality gasoline while keeping its density within specified limits. The proposed architecture combines a regression model built on six dominant parameters selected through Pearson’s correlation analysis, a Mamdani-type fuzzy inference system with a database of 26 expert rules, and a machine-learning module (Random Forest and gradient boosting) that corrects the residual error of the combined regression-fuzzy model, allowing the control system to adapt to changing process conditions. Testing the developed models on real data from the Shymkent Oil Refinery shows that the hybrid model reduces the root-mean-square error from 0.1973 (regression model alone) to 0.0161 and the mean absolute percentage error from 0.328% to 0.028%, while the coefficient of determination increases from 0.9895 to 0.9999; this improved accuracy is estimated to correspond to an approximate 3.2% increase in achievable gasoline yield and a 0.4% improvement in density stability. The developed decision support system for catalytic cracking process control, based on the proposed methodology, acts as a “virtual analyzer”, ensuring effective control in real time without the use of expensive in-line quality control devices.