Data-driven framework for forecasting sedimentation at culverts

Xu, Haowen

doi:10.17077/etd.jj4u-kfv0

Cited by 2 publications

(1 citation statement)

References 78 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recently, several papers have considered the use of constraints (based on physical knowledge) on the output of a machine learning model so that these mod- els can be trained even with unlabeled data by relying on physical principles [10,18,22]. Another very common approach is residual modeling, where an ML model is used to predict the errors made by a physics-based model [2,3,24]. In addition to generating results that may be inconsistent with physical laws, these residual modeling approaches do not have the capability to train ML models with unlabeled data by relying on physical principles.…”

Section: Related Workmentioning

confidence: 99%

Physics Guided RNNs for Modeling Dynamical Systems: A Case Study in Simulating Lake Temperature Profiles

Jia

Willard

Karpatne

et al. 2019

Proceedings of the 2019 SIAM International Conference on Data Mining

180

175

View full text Add to dashboard Cite

This paper proposes a physics-guided recurrent neural network model (PGRNN) that combines RNNs and physicsbased models to leverage their complementary strengths and improve the modeling of physical processes. Specifically, we show that a PGRNN can improve prediction accuracy over that of physical models, while generating outputs consistent with physical laws, and achieving good generalizability. Standard RNNs, even when producing superior prediction accuracy, often produce physically inconsistent results and lack generalizability. We further enhance this approach by using a pre-training method that leverages the simulated data from a physics-based model to address the scarcity of observed data. Although we present and evaluate this methodology in the context of modeling the dynamics of temperature in lakes, it is applicable more widely to a range of scientific and engineering disciplines where mechanistic (also known as process-based) models are used, e.g., power engineering, climate science, materials science, computational chemistry, and biomedicine.

show abstract

Section: Related Workmentioning

confidence: 99%