Understanding the Shear Rate-Dependent Viscosity of Complex Dispersions: A Focus on the Quemada Model and Neural Networks

Researchers have developed a new model, known as the Quemada model, to describe the viscosity of complex dispersions such as food products, paints, coatings, and cosmetics. The model takes into account various factors such as particle interactions, cluster sizes, and shear rate to predict the viscosity of these complex fluids accurately.

To further enhance the predictive power of the Quemada model, researchers have utilized neural networks (NNs) and physics-encoded neural networks (PeNNs). By training these networks with data generated from the Quemada model, researchers aim to predict the shear rate-dependent viscosity of complex dispersions more effectively.

Using Python and TensorFlow, researchers built and tested various NNs and PeNNs with different architectures. The performance of these networks was evaluated based on metrics such as coefficient of determination (R²) and mean square error (mse) between predicted and actual viscosity values.

In one case, the networks were trained to predict viscosity based on two input parameters – shear rate and particle volume fraction. In another case, the networks were trained with four input parameters, including additional factors like structure parameters at different shear rates.

The results showed that PeNNs, which incorporate physics-based activation functions, outperformed traditional NNs in predicting the viscosity of complex dispersions. By combining physics principles with neural network technology, researchers were able to achieve more accurate predictions of viscosity, essential for optimizing formulation and processing conditions in various industries.

This study highlights the potential of combining physics and machine learning techniques to enhance our understanding of complex fluid behavior and improve predictive models for viscosity in dispersions. The findings could have significant implications for industries dealing with complex fluids, leading to more efficient and effective product development processes.