Artificial neural network as a strategy to predict rheological properties in emulgel formulations

This study presents an innovative approach that combines quality by design (QbD) principles with artificial neural networks (ANNs) to predict and optimize the formulation of carbopol-based emulsions. By integrating these two strategies, we enhance our understanding of formulations by linking critical material attributes and process parameters to critical quality attributes, such as viscosity. The predictive model was refined by selecting key variables: mixing time, mixing speed, and viscosity. These variables were used to estimate carbopol concentration and to capture the nonlinear relationships that influence emulsion behavior. Experimental data were employed to train, validate, and test the ANN model, which was then compared with four commercial formulations to evaluate its predictive accuracy and practical relevance. Notably, the model demonstrated excellent predictive performance for systems with viscosities exceeding 50,000 mPas, underscoring its applicability to high-viscosity pharmaceutical products. This integrated QbD-ANN framework offers a systematic and effective method for formulation optimization, reducing experimental workloads while improving process understanding. The findings indicate a strong correlation between predicted and experimental values, confirming the robustness and reliability of the QbD-ANN approach. Integrating the three key variables enables a more in-depth examination of the interactions between process and formulation, providing a comprehensive tool for understanding and controlling emulsion viscosity. In conclusion, this study establishes a data-driven methodology that facilitates rational pharmaceutical development, ensuring product quality, reproducibility, and innovation in alignment with modern pharmaceutical quality management principles.