Parametric regression of a multiparameter thixotropic model for waxy crude oil based on multiobjective strategy

Abstract

The waxy crude oil exhibits a time- and shear-dependent non-Newtonian behavior called thixotropic behavior below the gelation temperature. This behavior can be quantitatively described by using a thixotropic model. However, many parameters are commonly contained in a thixotropic model with good-description capacity, and it still remains challenging to obtain high fitting precision and reasonable physical interpretation simultaneously in the process of parametric regression. In this study, based on a representative multiparameter thixotropic model, a multiobjective regression strategy is put forward. The fitting precision and physical interpretation are taken into account in this regression strategy, and their importance can be adjusted dynamically with the structure breakdown of waxy crude oil. In order to successfully achieve the multiobjective regression strategy, a self-adaptive nondominated sorting differential evolution algorithm (self-adaptive NSDE) is proposed. This regression algorithm is independent on the expressions of the regression model, and thus it can also be applied to the multiobjective parametric regression of other complex models including thixotropic models in rheology field and even other models in different fields. Based on experimental data, the parametric regression of a thixotropic model is performed, and the obtained results are analyzed and discussed in detail. It is found that the regressive data exhibit a small deviation at the starting stage of structure breakdown and the equilibrium flow curve can be ensured at the ending stage of structure breakdown. This finding well verifies considerations of multiobjective regression strategy. Compared with previous three regression strategies in the literature reports, the multiobjective regression strategy exhibits advantages in the fitting precision and physical interpretation.