The understanding of the dynamic behaviour of industrial processes is of fundamentalimportance to obtain products with desired specifications. The faster and less expensiveway to carry out this task is through the formulation of deterministic mathematicalmodels. However, these models include in most cases ordinary differential equationsand also partial differential equations in other ones. The solution of such modelsinvolves a considerable mathematical effort, requiring simplifications which may affectthe quality of the predictions. A booming alternative way of modeling is the use ofartificial intelligence techniques, specifically the application of fuzzy logic. Thisapproach has the great advantage of not requiring system fundamental knowledge, butonly input-output data, which makes it widely applicable in complex processes. Thiswork presents the fuzzy modeling of a process of molecular distillation heavy liquidpetroleum residue 400 °C+. In molecular distillation processes, the model is formed bypartial differential equations involving temperature and composition, which must benumerically solved due to its complexity. Here the process was simulatedcomputationally through the development of a software in Fortran 90/95, using finitedifferencemethods. The program is regarded as a virtual plant for the generation ofdynamic data required to build the fuzzy model. Furthermore, experimental data from amolecular distillation process of residue 400 °C+ of J-ES-110 crude oil at differentoperating conditions were used in the generation of the cognitive model. In the fuzzymodel, the distillation temperature and the feed flow rate were considered as inputvariables, while the liquid interface temperature, the film thickness, the concentrationprofiles, and the distillate flow rate were considered as the model output responses. Thefuzzy model obtained was compared with the results generated from the deterministmodel, showing an excellent agreement.
展开▼
