Grant: Discovery Grants Program - Individual
Concordia University
$32,000.00
Data-Driven Methods for Dynamical Systems Analysis
FY 2024-25
Summary
We are currently living through the big data revolution where the volume and production of data rises with the increasing demand for it in all sectors of society. What is key to understanding this data is identifying patterns and trends that can be exploited to analyze, interpret, and forecast it. The methods and techniques that will be developed under this proposal have the power to not only identify new high-fidelity models of dynamic data sets, but materially alter our current approach to the mathematical models we already have. The goal of this proposal is to combine state-of-the-art computational methods, such as machine learning, with innovative analysis to produce new results that will advance our approach of dynamic systems and bridge the gap between theory and the understanding of specific models.
The long-term goal of my work is in furthering one of the longest standing problems in physics: understanding turbulence. Turbulence is characterized by a chaotic motion in a fluid's velocity and pressure, commonly observed in fast flowing rivers, but also present in engineering applications such as liquid flowing through a pipe and air moving around an airplane's wing. Turbulent fluids are notoriously difficult to simulate on a computer and result in high-dimensional dynamic data, often only available for short bursts of time due to the computational power required to generate it. Thus, what is needed is a shift of the current mathematical paradigm that can either speed up these computations, or provide information about the system from a limited data set.
My work provides for model-agnostic computational methods that leverage the computing power of modern neural networks. These methods work directly with the data provided to the system, with short-term goals aiming to provide precise mathematical results that describe the trade-off between the amount of data provided and the accuracy of the models. The result will be simplified models of complex dynamics that allows one to analyze, forecast, and control expensive-to-produce, high-dimensional data in a computationally efficient manner.
Since I am proposing entirely new data-driven methods that must be developed from the ground up, short term objectives primarily focus on achieving provably accurate results for relatively simple models. This includes identifying weaknesses and developing robustness to overcome them. My approach is well-founded with achievable milestones that closely integrate implementation and testing with theoretical development. This provides a pathway that gradually improves the methods to handle more complex data sets, eventually moving towards my goal of working with fluid data. This work is topical, widely applicable, and provides for cross-disciplinary collaboration with researchers throughout the physical sciences. The applicability of the methods in the long-term to fluid dynamics further makes my work optimally suited for NSERC's Aerospace strategic initiative.
Program
Discovery Grants Program - Individual
Recipient
Location
Quebec, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data
As published
Application id
RGPIN-2023-04244
Name nom
Bramburger, Jason
Department département
Mathematics and Statistics
Organization id
70
Institution établissement
Concordia University
Province en
Québec
Province fr
Québec
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
1508
Committee name en
Mathematics and Statistics
Committee name fr
Mathématiques et statistique
Area of application code
1205
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Mathematical sciences
Area of application fr
Sciences mathématiques
Research subject code
2952
Research subject group en
Applied mathematics
Research subject group fr
Mathématiques appliquées
Research subject en
Linear and non-linear systems
Research subject fr
Systèmes linéaires et non linéaires
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
Data-Driven Methods for Dynamical Systems Analysis
Keywords
machine learning; autoencoder; data-driven discovery; system identification; auxiliary function; chaos; poincare map; dynamic mode decomposition; koopman operator; polynomial optimization
Application summary
We are currently living through the big data revolution where the volume and production of data rises with the increasing demand for it in all sectors of society. What is key to understanding this data is identifying patterns and trends that can be exploited to analyze, interpret, and forecast it. The methods and techniques that will be developed under this proposal have the power to not only identify new high-fidelity models of dynamic data sets, but materially alter our current approach to the mathematical models we already have. The goal of this proposal is to combine state-of-the-art computational methods, such as machine learning, with innovative analysis to produce new results that will advance our approach of dynamic systems and bridge the gap between theory and the understanding of specific models.
The long-term goal of my work is in furthering one of the longest standing problems in physics: understanding turbulence. Turbulence is characterized by a chaotic motion in a fluid's velocity and pressure, commonly observed in fast flowing rivers, but also present in engineering applications such as liquid flowing through a pipe and air moving around an airplane's wing. Turbulent fluids are notoriously difficult to simulate on a computer and result in high-dimensional dynamic data, often only available for short bursts of time due to the computational power required to generate it. Thus, what is needed is a shift of the current mathematical paradigm that can either speed up these computations, or provide information about the system from a limited data set.
My work provides for model-agnostic computational methods that leverage the computing power of modern neural networks. These methods work directly with the data provided to the system, with short-term goals aiming to provide precise mathematical results that describe the trade-off between the amount of data provided and the accuracy of the models. The result will be simplified models of complex dynamics that allows one to analyze, forecast, and control expensive-to-produce, high-dimensional data in a computationally efficient manner.
Since I am proposing entirely new data-driven methods that must be developed from the ground up, short term objectives primarily focus on achieving provably accurate results for relatively simple models. This includes identifying weaknesses and developing robustness to overcome them. My approach is well-founded with achievable milestones that closely integrate implementation and testing with theoretical development. This provides a pathway that gradually improves the methods to handle more complex data sets, eventually moving towards my goal of working with fluid data. This work is topical, widely applicable, and provides for cross-disciplinary collaboration with researchers throughout the physical sciences. The applicability of the methods in the long-term to fluid dynamics further makes my work optimally suited for NSERC's Aerospace strategic initiative.