Perspectives on the integration between first-principles and data-driven modeling

Outlook: How I Learned to Love Machine Learning (A Personal Perspective on Machine Learning in Process Systems Engineering)

“…45,66,81 Moreover, the community has extensively explored the use of ML models as surrogates of complex models. 68,82,83 Along these lines, I would like to highlight the development of OMLT, which is a software package for optimization modeling that automates the conversion of ML models (e.g., nonlinear neural networks) into tractable, mixedinteger linear representations. 84 I believe that this work can be highly impactful, as it can help standardize modeling environments (e.g., every unit operation in a chemical process is a neural network).…”

“…Analogous to the story on how capabilities of optimization were extended by the PSE community to tackle complex problems, I would like to highlight some of the work that the PSE community has been doing in extending the capabilities of ML. The PSE community has developed powerful global optimization algorithms and software that can handle formulations that have embedded neural network and GP models. , The community has also recently developed BO architectures that combine data-driven and physics models (and models of different levels of resolution) to guide experimental design. ,, Moreover, the community has extensively explored the use of ML models as surrogates of complex models. ,, Along these lines, I would like to highlight the development of , which is a software package for optimization modeling that automates the conversion of ML models (e.g., nonlinear neural networks) into tractable, mixed-integer linear representations . I believe that this work can be highly impactful, as it can help standardize modeling environments (e.g., every unit operation in a chemical process is a neural network).…”

The Application of Physics-Informed Machine Learning in Multiphysics Modeling in Chemical Engineering

“…Insufficient or sparse data can lead to the inadequate predictability of datadriven models, thereby impeding the progress and optimization of surrogate modeling. Additionally, data-driven models are mostly regarded as black-box or gray-box models, 17 which lack interpretability and do not reflect domain knowledge of physics. This lack of interpretability makes it challenging to extract insights and understand the mechanisms that govern the system being modeled, hindering the development of more accurate and efficient models.…”

Modeling and Control of a Chemical Process Network Using Physics-Informed Transfer Learning

“…Recent advances in physics-informed RNN (PIRNN) were summarized in ref . Different structures of the integrated first-principles and data-driven models were reviewed in ref , in which the physics-informed machine learning method was highlighted. Furthermore, physics-informed machine learning has been integrated with transfer learning in some recent works.…”

“…To address the problem of data scarcity, physics-informed machine learning is shown to have the potential to improve the prediction performance by embedding domain knowledge into the training process. 17 Although an accurate first-principles model is difficult (even impossible) to develop for a complex chemical process network (e.g., a chemical plant), a priori process knowledge, such as mass and energy balance equations, as well as process structure knowledge, can be utilized to improve the development of RNN models. Hybrid modeling approaches have also been proposed to enhance the prediction accuracy by combining the first-principles model and the deep neural network.…”