<p>Mixture experiments are widely used to model the relationship between the proportions of formulation components and product quality attributes; however, the presence of component restrictions and nonlinear response behavior may introduce multicollinearity and numerical instability in traditional Scheffé models. This study proposes an alternative mixture modeling approach that incorporates square root transformation terms to improve predictive flexibility while preserving interpretability. The proposed formulation is evaluated through experimental examples involving constrained mixture regions and through a Monte Carlo simulation study comparing the quadratic Scheffé model with the square root model under multiple stability and predictive performance criteria, including adjusted coefficient of determination, mean square error, condition number, and variance inflation factor. Results from empirical case studies show that the inclusion of square root terms, particularly when combined with pseudocomponent transformations, can substantially improve model fit and reduce residual variability in restricted mixture regions. However, the simulation analysis indicates that when the true data-generating mechanism follows a quadratic Scheffé structure, the classical Scheffé formulation generally exhibits superior numerical conditioning and lower multicollinearity, whereas the square root model provides competitive predictive accuracy with slightly higher instability measures. These findings highlight that the proposed model should be interpreted as a complementary alternative rather than a universal replacement for traditional mixture models. Overall, the combined use of pseudocomponent transformations and nonlinear terms offers a flexible framework for modeling complex mixture systems, especially in applications where boundary effects or strong nonlinear component behavior are present.</p>