Real Component Mixture of Petroleum Cuts to beintroduced to a Neural Network Reactor model

摘要

Petroleum cuts are very complex mixtures which contain very large number of components,from which only few can be usually identified. In this paper, a method ispresented to introduce a range of petroleum fractions to the neural network model of areactor. To do so, a "model" of the mixture should be used. Reactor feed type plays anessential role on the reactor product qualities. In order to introduce petroleum cuts withfinal boiling points of 865°F maximum to the neural network, a real componentsubstitute mixture is made from the original mixture. Such substitute mixture is fullydefined, it has a chemical character, and physical properties can be simply retrievedfrom databases. The mixture compositions are defined with the aid of optimization. Theobtained TBP curves of several substitute mixtures are in good agreement with theexperimentally obtained curves. Nine single carbon structural increments will be therepresentatives of 93 real component compositions in order to make the topology of theneural network smaller and hence to have less complex model. An NN model was also