In the industrial sector, foaming remains an inevitable side effect of mixing, shearing, powder incorporation, and the metabolic activities of microorganisms in a bioprocess. Excessive foaming can interfere with the mixing of reactants and lead to problems such as decreased effective reactor volume, microbial contamination, product loss, and increased reaction time. Physical modeling of foaming is an arduous process as it requires estimation of foam height, which is dynamic in nature and varies for different processes. This work demonstrates a novel application of ensemblebased machine learning methods for prediction of different fermenter types in a fermentation process (to allow for successful data integration) and of the onset of foaming. Ensemble-based methods are robust nonlinear modeling techniques that aggregate a set of learners to obtain better predictive performance than a single learner. We apply two ensemble frameworks, extreme gradient boosting (XGBoost) and random forest (RF), to build classification and regression models. We use real plant data for 64 batches from four fermenters with different material, geometry, and equipment specifications. Our first task is to develop an ensemble-based fermenter classification model that uses well-known fermentation independent variables for each batch alone, without having to incorporate explicitly the design specifications. The resulting fermenter classification model is able to differentiate or classify the fermenter type with an accuracy of 99.49% for our integrated data sets of over 183 000 instances. This enables us to integrate multiple plant data sets from different fermenter specifications and develop a generalized foaming prediction model. Next, we build classification and regression models for foaming prediction. The resulting models are able to predict the foaming indicator (the exhaust differential pressure) to achieve an accuracy of 82.39% and an RMSE value of ±12 mbarg, which is well within the tolerance for foaming prediction in industrial practice. These results demonstrate the effectiveness of ensemble-based machine learning models for fermenter classification, data integration, and foaming prediction involving multiple fermenter design specifications. Using these tools, we can orchestrate the addition of chemical antifoam agents (AFA) or defoamers in an ad hoc manner to mitigate the adverse effects of excessive AFA addition. Our work differentiates itself from previous work in this area through the following contributions: (1) accurate ensemble-based classification modeling to differentiate fermenter types on the basis of known independent variables alone, without prior knowledge of fermenter design specifications, thus allowing for data integration of multiple plant data sets to build better prediction models; (2) accurate prediction of foaming based on exhaust differential pressure using both classification and regression models; and (3) usage of a large, industrial, multivariate fermenter data set.
In industrial fermenters, antifoaming agents (AFA) are often added in a reactive manner to prevent foaming in a reactor. To successfully predict the antifoam addition in a proactive manner for an industrial data set (about half a million instances for 163 batches), we use a genetic algorithm called TPOT (Tree-based Pipeline Optimization Tool), and to build antifoam profiles, we use exploratory time-series analysis. Through this work, we have successfully demonstrated the following: (1) the ability of automated machine learning (AML) to predict AFA addition for multiple strains of microorganisms using large-scale industrial fermentation data set and (2) the edge of data-guided antifoam profiles based on time series. From our initial deployment results, we see significant improvements in fermentation product yield, reflected by the significant decrease in hourly volume retention (up to 53%) and decrease in the count of exhaust differential pressure (a foaming indicator) exceeding a threshold of 100 mbarg, among other performance indicators.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.