Abstract
AbstractRepresenting the first-generation high-performance water-based drilling fluid, KCl/Polymer drilling fluids are widely used to drill troublesome shale formations containing water-sensitive clay minerals. In addition to maintaining wellbore stability, its rheological properties also play a crucial role in enhancing overall drilling performance. An accurate description of the rheological behavior of drilling fluids is essential in optimizing drilling fluid hydraulics. This study evaluates traditional and novel optimization algorithms for the parameterization of rheological models using an extensive field rheological database of KCl/Polymer drilling fluids. An objective function based on a symmetric mean absolute percent error is used in parameterizing rheological models. Golden Section Search (GSS), Generalized Reduced Gradient (GRG), and Trust Region (TR) methods are used as new alternatives to traditional Gaussian-Newton (GN) and linear/semi-linear regression (LR/QLR) methods. As a more statistically plausible criterion, the symmetric mean absolute percentage error is also used to measure the goodness of fit of rheological models with datasets. It has been shown that GRG and TR algorithms outperform conventional methods in finding optimal model parameters. The three- and four-parameter models fitted the rheological data best, with a more uniform symmetrical error distribution than the two-parameter models.