(673g) An Adaptive Sampling Surrogate Model for Mixing Time Prediction and Mixing Characterization

Characterization of process tanks to determine effective mixing and homogenous conditions is a time consuming and cost intensive effort. Therefore, we have developed a comprehensive in-silico model to predict mixing time based on tank specifications, agitation rate and mixture physical properties. Our model is developed based a Computational Fluid Dynamics (CFD) model which is validated in mixing analysis of biopharma mixing vessels. The CFD model is integrated into a four-variables surrogate model in which an adaptive sampling method is applied. Our surrogate model is capable of predicting mixing time and mixing characterization with less than 25% error with only 60 model training samples. The developed surrogate model is available in as a web interface for use by domain experts.