Grant: Discovery Grants Program - Individual
University of Calgary
$32,000.00
Decision Making in High-Dimensional Engineering Systems through Adaptive Data Acquisition and Low-Rank Surrogate Models
EF 2024-25
Résumé
This Discovery Grant research program investigates decision making under uncertainty in high-dimensional engineering and industrial systems through low-rank models and adaptive sampling techniques. This covers the development of next-generation surrogate models or emulators that replace an underlying high-dimensional system with an inexpensive-to-evaluate mathematical model. This occurs for many engineering systems when the number of system parameters, degrees of freedom, or stochastic inputs increase from moderate to high dimensions, placing significant computational and modelling limitations, such as in airfoil shape optimization in turbulent flows.
The challenge is that decision support tools used during the design and analysis process, e.g., multidisciplinary optimization, often require thousands of `online' evaluations of the underlying high-fidelity model, which is computationally infeasible. Statistically accurate surrogate models or emulators are essential to enable the use of such decision support tools. The utility of this approach highly depends on the accuracy of the surrogate model, however constructing surrogate models of high-dimensional systems remains an unsolved challenge.
This research program aims to develop novel algorithmic and theoretical advances in the construction of statistically accurate surrogate models of high-dimensional systems to enable improved decision-making tools. The novel approach will leverage methods based on low-rank models that utilize tensor network decompositions and nonlinear reduction techniques, along with adaptive data sampling. Unlike traditional design-of-experiment methods that a priori select samples, the strategy here will combine surrogate models with adaptive sampling to inform selection of future observations in a low-rank subspace. The purpose is to jointly discover and iteratively refine a low-dimensional representation of the high-fidelity model using nonlinear embeddings. This approach forms the novel basis to overcome the two challenges facing surrogate model construction: (1) the curse of dimensionality and (2) the computational cost associated with a model query.
The algorithmic and theoretical outcomes from this program will be an enabling technology in the engineering space and will allow practitioners to assess their engineering models more reliably in the face of a larger range of model uncertainties, design more robust structures through the expanded design space exploration opportunity and enable more accurate model prediction in digital simulations. Another key objective of the program is to develop improved emulators in two applications: (1) transport-dominated equations used in anthropogenic tracer forecasting, and (2) generative topology design for new methods in manufacturing. This program will train HQP in the transdisciplinary field of data-driven engineering to be competitive in the global workforce and especially the Canadian technology sector.
Programme
Discovery Grants Program - Individual
Bénéficiaire
Emplacement
Alberta, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data
Tel que publié
Application id
RGPIN-2023-03760
Name nom
Mohamad, Mustafa
Department département
Mechanical & Manufacturing Engineering
Organization id
11
Institution établissement
University of Calgary
Province en
Alberta
Province fr
Alberta
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2023
Award amount
32000
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
1512
Committee name en
Mechanical Engineering
Committee name fr
Génie mécanique
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
2107
Research subject group en
Mechanical engineering
Research subject group fr
Génie mécanique
Research subject en
Modelling, simulation and finite element methods
Research subject fr
Modélisation, simulation et méthodes des éléments finis
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
Decision Making in High-Dimensional Engineering Systems through Adaptive Data Acquisition and Low-Rank Surrogate Models
Keywords
bayesian modeling and computation; experimental design; uncertainty quantification; high-dimensional dynamical systems; active learning; sequential decision making; machine learning in complex physical systems; multi-scale modelling; data-driven prediction; sensor orchestration
Application summary
This Discovery Grant research program investigates decision making under uncertainty in high-dimensional engineering and industrial systems through low-rank models and adaptive sampling techniques. This covers the development of next-generation surrogate models or emulators that replace an underlying high-dimensional system with an inexpensive-to-evaluate mathematical model. This occurs for many engineering systems when the number of system parameters, degrees of freedom, or stochastic inputs increase from moderate to high dimensions, placing significant computational and modelling limitations, such as in airfoil shape optimization in turbulent flows.
The challenge is that decision support tools used during the design and analysis process, e.g., multidisciplinary optimization, often require thousands of `online' evaluations of the underlying high-fidelity model, which is computationally infeasible. Statistically accurate surrogate models or emulators are essential to enable the use of such decision support tools. The utility of this approach highly depends on the accuracy of the surrogate model, however constructing surrogate models of high-dimensional systems remains an unsolved challenge.
This research program aims to develop novel algorithmic and theoretical advances in the construction of statistically accurate surrogate models of high-dimensional systems to enable improved decision-making tools. The novel approach will leverage methods based on low-rank models that utilize tensor network decompositions and nonlinear reduction techniques, along with adaptive data sampling. Unlike traditional design-of-experiment methods that a priori select samples, the strategy here will combine surrogate models with adaptive sampling to inform selection of future observations in a low-rank subspace. The purpose is to jointly discover and iteratively refine a low-dimensional representation of the high-fidelity model using nonlinear embeddings. This approach forms the novel basis to overcome the two challenges facing surrogate model construction: (1) the curse of dimensionality and (2) the computational cost associated with a model query.
The algorithmic and theoretical outcomes from this program will be an enabling technology in the engineering space and will allow practitioners to assess their engineering models more reliably in the face of a larger range of model uncertainties, design more robust structures through the expanded design space exploration opportunity and enable more accurate model prediction in digital simulations. Another key objective of the program is to develop improved emulators in two applications: (1) transport-dominated equations used in anthropogenic tracer forecasting, and (2) generative topology design for new methods in manufacturing. This program will train HQP in the transdisciplinary field of data-driven engineering to be competitive in the global workforce and especially the Canadian technology sector.