Grant: Discovery Grants Program - Individual
Carleton University
$38,000.00
Addressing overfitting, uncertainty quantification and computational challenges in machine learning with applications to engineering mechanics
EF 2024-25
Résumé
The proposed novel physics-based machine learning (ML) and neural network (NN) algorithms endowed with more robust, generalizable and computationally efficient predictive capabilities, being informed by observational data either real time or offline, can enhance confidence in designing and operating high consequence engineering systems in uncertain environments. This emergent and efficient sampling-free ML framework is applicable to wide-range of systems such as an aircraft, urban air mobility vehicles, wind turbines, long-span bridges, dams, nuclear reactors and many battlefield applications. The versatility and wide-ranging applicability of this Bayesian ML framework rests upon its ability to bridge the gap between experimentalists and computational scientists by using HPC to integrate measurement data with the mechanics-based ML models. Due to their inherent capability of generalization (ability to predict unseen data) and uncertainty quantification, these physics-based ML framwork can help save the cost of future physical experiments through virtual testing using high performance computing by making use of available (sparse and noisy) observational data. The long-term objective of this research is (1) to build scalable algorithms for sparse Bayesian physics-based ML and NNs with applications to engineering mechanics; (2) to develop a comprehensive and fully integrated open-source software infrastructure (leveraging high performance computing) to implement this sparse Bayesian ML and AI framework to broaden the scientific contributions of my research.
This initiative will produce a leading-edge ML and AI tools for building smarter technologies that can help Canadian scientists and innovators. The Canadian fields of renewable energy, infrastructure renewal, military preparedness and decision making will benefit from the timely knowledge developed from comprehensive assessments through modelling and observations using robust open-source ML and AI tools. The proposed interpretable (explainable), generalizable and robust (having low uncertainty in predictions) ML framework will boost public confidence and regulatory acceptance of scientific predictions in the broad areas of computational science and engineering. The initiative will serve the nation by capitalizing on the investments Canada has made in high performace computing infrastucture through development of new scalable algorithms of physics-based ML and NN for credible predictions and their validation through physical experiments.
Programme
Discovery Grants Program - Individual
Bénéficiaire
Emplacement
Ontario, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data
Tel que publié
Application id
RGPIN-2024-06483
Name nom
Sarkar, Abhijit
Department département
Civil and Environmental Engineering
Organization id
23
Institution établissement
Carleton University
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2024
Award amount
38000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
DISCOVERY RESEARCH
Group fr
RECHERCHE AXÉE SUR LA DÉCOUVERTE
Committee code
1509
Committee name en
Civil, Industrial and Systems Engineering
Committee name fr
Génie civil, industriel et des systèmes
Area of application code
1206
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Engineering
Area of application fr
Génie
Research subject code
1000
Research subject group en
Civil engineering
Research subject group fr
Génie civil
Research subject en
Civil engineering
Research subject fr
Génie civil
Field of research list names en
Not available
Field of research list names fr
Non disponible
Socioeconomic objective list names en
Not available
Socioeconomic objective list names fr
Non disponible
Application title
Addressing overfitting, uncertainty quantification and computational challenges in machine learning with applications to engineering mechanics
Keywords
bayesian inference; unicertainty quantification; computational mechanics; fluid-structure interaction; machine learning; artificial intelligence; domain decomposition; scalable algorithm; sparse learning; overfitting
Application summary
The proposed novel physics-based machine learning (ML) and neural network (NN) algorithms endowed with more robust, generalizable and computationally efficient predictive capabilities, being informed by observational data either real time or offline, can enhance confidence in designing and operating high consequence engineering systems in uncertain environments. This emergent and efficient sampling-free ML framework is applicable to wide-range of systems such as an aircraft, urban air mobility vehicles, wind turbines, long-span bridges, dams, nuclear reactors and many battlefield applications. The versatility and wide-ranging applicability of this Bayesian ML framework rests upon its ability to bridge the gap between experimentalists and computational scientists by using HPC to integrate measurement data with the mechanics-based ML models. Due to their inherent capability of generalization (ability to predict unseen data) and uncertainty quantification, these physics-based ML framwork can help save the cost of future physical experiments through virtual testing using high performance computing by making use of available (sparse and noisy) observational data. The long-term objective of this research is (1) to build scalable algorithms for sparse Bayesian physics-based ML and NNs with applications to engineering mechanics; (2) to develop a comprehensive and fully integrated open-source software infrastructure (leveraging high performance computing) to implement this sparse Bayesian ML and AI framework to broaden the scientific contributions of my research.
This initiative will produce a leading-edge ML and AI tools for building smarter technologies that can help Canadian scientists and innovators. The Canadian fields of renewable energy, infrastructure renewal, military preparedness and decision making will benefit from the timely knowledge developed from comprehensive assessments through modelling and observations using robust open-source ML and AI tools. The proposed interpretable (explainable), generalizable and robust (having low uncertainty in predictions) ML framework will boost public confidence and regulatory acceptance of scientific predictions in the broad areas of computational science and engineering. The initiative will serve the nation by capitalizing on the investments Canada has made in high performace computing infrastucture through development of new scalable algorithms of physics-based ML and NN for credible predictions and their validation through physical experiments.