A general deep hybrid model for bioreactor systems: Combining first principles with deep neural networks. (September 2022)

• Record Type: Journal Article
• Authors: Pinto, José; Mestre, Mykaella; Ramos, J.; Costa, Rafael S.; Striedner, Gerald; Oliveira, Rui
• Is Part Of: Computers & chemical engineering. Volume 165(2022)
• Publication Date: 2022-09
• Subjects: Hybrid semiparametric modeling -- Deep neural networks -- ADAM algorithm -- Stochastic regularization -- Bioprocess dynamics -- Pichia pastoris; Chemical engineering -- Data processing -- Periodicals; 660.0285
• DOI: 10.1016/j.compchemeng.2022.107952
• Languages: English
• ISSNs: 0098-1354

Abstract

Highlights: Hybrid semiparametric models have been widely used for bioprocess modeling. Shallow structures and "nondeep" training are well covered in the literature. Here, deep structures/training are investigated in a hybrid modeling context. ADAM with stochastic regularization significantly reduces the training CPU. Deep hybrid models show higher predictive power than the shallow counterpart. Abstract: Numerous studies have reported the use of hybrid semiparametric systems that combine shallow neural networks with First Principles for bioprocess modeling. Here we revisit the general bioreactor hybrid model and introduce some deep learning techniques. Multi-layer networks with varying depths were combined with First Principles equations in the form of deep hybrid models. Deep learning techniques, namely the adaptive moment estimation method (ADAM), stochastic regularization and depth-dependent weights initialization were evaluated in a hybrid modeling context. Modified sensitivity equations are proposed for the computation of gradients in order to reduce CPU time for the training of deep hybrid models. The methods are illustrated with applications to a synthetic dataset and a pilot 50 L MUT+ Pichia pastoris process expressing a single chain antibody fragment. All in all, the results point to a systematic generalization improvement of deep hybrid models over its shallow counterpart. Moreover, the CPU cost to train the deep hybrid models is shown to be lower than for the shallow counterpart. In the pilot 50L MUT+ Pichia pastoris data set, the prediction accuracy was increased by 18.4% and the CPU decreased by 43.4%.