The use of sensitivity analysis and genetic algorithms for the management of catalyst emissions from oil refineries

Excessive catalyst emissions from Fluidized Catalytic Cracking Units (FCCU) during start-up situations are common, and have been deemed ‘normal’, with little research conducted on determining their causes. A MATLAB model found to predict trends in emission rates under normal conditions has been expanded to better represent the actual processes inside an FCCU. First and second order sensitivity analysis techniques are used to assess the interactions between various operational parameters, with a genetic algorithm used to optimize the operating conditions to minimize air emissions. These ‘key’ parameters may then be altered to help manage both normal and start-up emissions through operational changes. It was also found that significant scale-up issues arise with the use of the attrition models found in the literature.