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

William Bradley, Jinhyeun Kim, Zachary A. Kilwein, Logan Blakely, Michael Eydenberg, Jordan Jalvin, Carl D. Laird, Fani Boukouvala

Abstract: Highlights•Strengths and limits of modeling frameworks merging first-principles and data are evaluated.•Hybrid first-principles, data-driven methods generally outperform purely data-driven models.•Recent advances in gradient evaluation and sampling methods have led to mature software implementations of hybrid modeling (HM) frameworks.•Care should be taken when extrapolating/interpreting data-driven component of HMs.•An assessment is given for improving current methods through uncertainty quantification, constraint handling and open-source publication of novel hybrid implementations.