This project aims to develop a soft analyzer for the analysis of BTEX emissions from the thermal section exit of sulfur recovery units (SRU). The ultimate goal is to develop a soft analyzer that will assist in optimizing the SRU and decreasing fuel gas consumption significantly. This analyzer will be developed based on the most important reactions governing aromatics destruction/formation in Claus furnace. The reactions of lower hydrocarbons, PAH, and BTEX in our previously developed kinetic model for the Claus furnace will be critically re-examined with the goal of reducing the number of reaction pathways. The gas phase kinetics of other chemically-active constituents that include CO, COS, and SO2 will also be re-evaluated to identify reactions that do not contribute significantly to sulfur production and BTEX destruction.
The reduced reaction mechanism will be used to simulate the Claus furnace and WHB. The predicted data will be validated against experimental data to ensure the high predictability and reliability of the reduced kinetic model. The robust Chemkin Pro software will be systematically coupled with HYSYS and MATLAB or HEEDS (an optimization software) to develop a user-friendly soft analyzer. The coupling of two process simulators offers a novel means of enhancing the computational accuracy of SRU simulations. Chemkin Pro, while handling the thermal section accurately, could not be used to model the catalytic section due to its complexity, and Aspen HYSYS (an equilibrium based process simulator) could not be used to capture the detailed chemistry of kinetically-limited reactions of BTEX destruction. Thus, the use of two process simulators is necessary for a reliable and complete evaluation of the SRU performance and fuel gas consumption in different Habshan processing units. This will assist in evaluating the reliability and potential benefit of the proposed soft analyzer that will be developed. An effort will be made to minimize the computational time of the soft analyzer to the minimum possible through the use of machine learning algorithms.
