EVALUATION AND MODELING OF FOULING CHARACTERISTICS OFPETROLEUM CRUDE OILS AND CRUDE BLENDS

Fouling of preheat exchangers in refinery crude distillation unit is a complex phenomena and identified to be the major energy consuming source in petroleum refineries. The cost of fouling could be substantial where it comprises the economics and environmental aspect. In this research work, four Malaysian crude oils and a condensate were selected for fouling experiments using Hot Liquid Process Simulator (HLPS). The experiments were conducted to determine the effect of surface temperature and crude blending on the fouling characteristics of the selected crudes and crude blends. A method published in the literature is used to analyze the raw data into a meaningful fouling resistance data. Initial fouling rates are then determined by taking the slope for the linear portion of the fouling resistance versus time curve. Arrhenius plot was used to obtain the true activation energy, E so that the fouling propensity could be determined. Crude ranking in term of fouling propensity for the neat crude is in the order of Crude C > Crude D > Crude B > Crude A and for the crude blend, it is in the order of A-C blend> A-D blend> A-B blend. The effect of adding Condensate E to Crude C has resulted in the lowest activation energy in comparison with the other crudes and crude blends. Four threshold fouling models were validated with the experimental data established. The models evaluated are Ebert and Panchal model, Panchal eta!. model, Polley eta!. model and Nasr and Givi model. Furthermore, three estimation methods were used for each model which are (i) estimation method I - physical properties estimated at inlet bulk temperature, (ii) estimation method 2 - physical properties estimated at film temperature or surface temperature (for Polley eta!. model only) and (iii) estimation method 3 - physical properties estimated at film temperature or surface temperature (for Polley eta!. model only) plus the exclusion of removal term. Model parameters were estimated using least square technique to minimize the error between the predicted and experimental data.