Physics-informed machine learning-aided framework for prediction of minimum film boiling temperature

Kim, Kyung Mo; Hurley, Paul; Duarte, Juliana P.

doi:10.1016/j.ijheatmasstransfer.2022.122839

Cited by 26 publications

(9 citation statements)

References 50 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Similar research developed a PINN model to solve stiff chemical kinetic problems . Kim et al suggest the PINN approach to predict the minimum film boiling temperature. In other studies, PINN was applied to predict the cetane number .…”

Section: Introductionmentioning

confidence: 99%

Physics-Informed Neural Networks with Group Contribution Methods

Babaei

Stone

Knotts

et al. 2023

J. Chem. Theory Comput.

View full text Add to dashboard Cite

Thermophysical properties of organic compounds are used in countless scientific, engineering, and industrial settings in developing theories, designing new systems and devices, analyzing costs and risks, and improving existing infrastructure. Often, due to costs, safety, prior interest, or procedural difficulties, experimental values for desired properties are not available and must be predicted. The literature is filled with prediction techniques, but even the best traditional methods have significant errors compared to what is possible considering experimental uncertainty. Recently, machine learning and artificial intelligence techniques have been applied to the property prediction problem, but the examples to date do not extrapolate well outside the data set used for training the model. This work demonstrates a solution to this problem by combining chemistry and physics when training the model and builds upon prior traditional and machine learning methods. Two case studies are presented. The first is for parachor which is used for surface tension prediction. Surface tensions are needed to design distillation columns, adsorption processes, gas−liquid reactors, liquid−liquid extractors, improve oil reservoir recovery, and undertake environmental impact studies or remediation actions. A set of 277 compounds is divided into training, validation, and test sets, and a multilayered physics-informed neural network (PINN) is developed. The results demonstrate that better extrapolation by deep learning models can be developed by adding in physics-based constraints. Second, a set of 1600 compounds is utilized for training, validating, and testing a PINN to improve normal boiling point predictions based on group contribution methods and physics-based constraints. The results show that the PINN performs better than any other method with a normal boiling point mean absolute error of 6.95 °C on training and 11.2 °C on test data. Key observations are that (1) a balanced split by compound type is important to have representative compound families in each of the train, validation, and test sets and (2) constraining group contributions being positive improves predictions on the test set. While this work demonstrates improvements for only surface tension and normal boiling point, the results offer significant hope that PINNs can improve prediction of other relevant thermophysical properties over existing approaches.

show abstract

Section: Introductionmentioning

confidence: 99%

Physics-Informed Neural Networks with Group Contribution Methods

Babaei

Stone

Knotts

et al. 2023

J. Chem. Theory Comput.

View full text Add to dashboard Cite

show abstract

“…Physics-informed approaches, such as predicting an error between a physical model's prediction and the true value, can facilitate extrapolation [30,31]. Another popular approach to handling extrapolation are physics-informed neural networks, where the derivatives obtained from automatic differentiation are incorporated into the loss function via a partial differential equation [32][33][34][35].…”

Section: Introductionmentioning

confidence: 99%

Physics‐informed Transfer Learning for Out‐of‐sample Vapor Pressure Predictions

Lansford

Jensen

Barnes

2023

Propellants Explo Pyrotec

View full text Add to dashboard Cite

Recent advances have enabled machine learning methodologies developed for large datasets to be applied to the small experimental datasets typically available for chemical systems. Such advances typically involve a databased approach to transfer learning, where a portion of the experimental data for the property of interest is used to fine-tune a model that is pre-trained on computationally generated data. This transfer learning approach does not work for very small experimental datasets, where there are only enough data for model validation. Here, we develop a physics-informed transfer learning strategy to train a directed-message passing neural network (D-MPNN) model, enabling extrapolation outside of the training domain. We demonstrate this approach by training a D-MPNN model on interpolated vapor pressures and validate the model on an out-of-sample test set of energetic molecule vapor pressures, achieving accuracy comparable to those of experiments.

show abstract

“…They can be divided into two types according to whether considering the conditions of the heating surface or not. Recent studies ,, indicated that the previous MFB temperature models had failed to consolidate the wide range of data owing to the use of data in limited experimental ranges in the development of each model. Therefore, there is a need for more comprehensive and accurate models that consider a wider range of conditions, including the heating surface.…”

Section: Introductionmentioning

confidence: 99%

“…Alotaibi et al developed a model for predicting MFB temperature using the random forest method based on the same database and parameters. Kim et al built a larger database containing 399 data points from 10 research sources and combined the conventional correlations and machine learning techniques (multilayer perceptron and random forest) to predict MFB temperature. Their results also showed that the ML model had superior predictive capacity to the existing correlations.…”

Section: Introductionmentioning

confidence: 99%

“…Although ML-based methods have demonstrated superior performance in predicting MFB temperature to previous correlations, they still encounter certain limitations. One limitation is the lack of consideration for surface roughness in previous ML-based studies. ,, However, many experimental studies ,, have shown the effect of surface roughness on MFB temperature. Another is the limited interpretability and black-box nature of conventional ML models, which provide little insight into the physical relationships between inputs and outputs.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Machine Learning Approach for Predicting Minimum Film Boiling Temperature Considering the Surface Roughness Effect

Tang,

Yu,

Gong

et al. 2023

Ind. Eng. Chem. Res.

View full text Add to dashboard Cite

The minimum film boiling (MFB) temperature, a crucial industrial design parameter for nuclear reactors, cryogenic milling, and grinding, refers to the lowest sustainable temperature at which stable film boiling occurs. Despite the development of many thermodynamic and hydrodynamic models, a universal model for predicting the MFB temperature is still required. In recent years, machine learning (ML) methods have been shown to outperform previous correlations in predicting the MFB temperature. However, these ML methods still have limitations in interpretability and the inclusion of key parameters, such as surface roughness, in the database. To address these issues, this study applies six ML models and a data interpretation method (SHapley Additive exPlanations, SHAP) to investigate MFB temperature. The extreme gradient boosting (XGBoost) model performs the best among these ML models, with an MAE of 2.6% for predicting MFB temperature without considering surface roughness, significantly outperforming the existing correlations. To account for surface roughness, a reduced database excluding data points without roughness values was used to train new XGBoost models. To fully use the removed data, a strategy is developed where normalized roughness and the MFB temperature predicted by the XGBoost model trained with the complete database excluding roughness are used as input variables. This approach improves the performance and generalization capabilities under insufficient surface roughness data sources by transferring the knowledge acquired from the complete database to the ML model developed by using the reduced database. The results show that the contributions of input variables and the physical variation trend of the XGBoost models are consistent with previous experimental results and theories.

show abstract

Physics-informed machine learning-aided framework for prediction of minimum film boiling temperature

Cited by 26 publications

References 50 publications

Physics-Informed Neural Networks with Group Contribution Methods

Physics-Informed Neural Networks with Group Contribution Methods

Physics‐informed Transfer Learning for Out‐of‐sample Vapor Pressure Predictions

Machine Learning Approach for Predicting Minimum Film Boiling Temperature Considering the Surface Roughness Effect

Contact Info

Product

Resources

About