On-line prediction of naphtha cracker product yields

An on-line model for the radiant section of a naphtha cracking furnace is built based on the molecular kinetics of naphtha pyrolysis. To predict the main product yields on-line, the average properties of naphtha (e.g. molecular formula, form-heat, heat conductivity coefficient and thermal capacity) and the selectivity coefficients of the primary reaction should be adjusted on-line to adapt to the gradual change of feed-in naphtha. An estimation algorithm is introduced here to calculate the average properties of naphtha based on normal physical characteristics of naphtha. The normal characteristics include ASTM D86 values (IBP, 10%, 30%, 50%, 70%, 90% and 95%), PONA values (n-paraffines, i-paraffines, naphthenes and aromatics) and density. In addition, a stochastic method is brought forward to adjust the selectivity coefficients of primary reaction. The on-line application showed that the main product yields predicted agree well with measured data from laboratory. The predicted yields also show 10 minutes ahead of chromatogram equipment. It indicates that this on-line model is effective.