國立虎尾科技大學 |

Efficient Learning of the State in Dynamic Systems From Model-Based to Data-Based.

Record Type:	Language materials, manuscript : Monograph/item
Title/Author:	Efficient Learning of the State in Dynamic Systems From Model-Based to Data-Based./
Author:	Sun, Jieming.
Description:	1 online resource (121 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 85-07, Section: A.
Contained By:	Dissertations Abstracts International85-07A.
Subject:	Industrial engineering. -
Online resource:	click for full text (PQDT)
ISBN:	9798381406566

Efficient Learning of the State in Dynamic Systems From Model-Based to Data-Based.
Sun, Jieming.

Efficient Learning of the State in Dynamic Systems From Model-Based to Data-Based. - 1 online resource (121 pages)

Source: Dissertations Abstracts International, Volume: 85-07, Section: A.

Thesis (Ph.D.)--The Florida State University, 2023.

Includes bibliographical references

Understanding and predicting states in dynamic systems is a core challenge in many areas of study. This involves trying to estimate an evolving signal based on past observations or the understood rules of the system. The main obstacles include errors in observations, the unknowns, and the complexity of these rules.This dissertation addresses the intricate problem of state estimation in dynamic systems under constraints of limited information. Central to our investigation is the pivotal balance between the amount of available data and the accuracy of state predictions. Broadly, the information about dynamic systems falls into two camps: measurements, which are historical observations, and models, the theoretical frameworks defining system evolution. While dense and precise measurements are desirable for accuracy, they are often constrained by technological limitations. On the flip side, while detailed models can guide precise predictions, they are sometimes elusive for complex systems.Initially, the research delves into model-based estimation, where system models are employed to calculate system states. However, uncertainties in various factors, such as initial conditions or the model itself, introduce errors. A unique scenario within model-based estimation is explored where humans function as sensors in a cyber-physical-human system, sending binary signals based on system performance, shedding light on the challenges of relying on such sparse information.The research then shifts to data-based estimation in scenarios where system models are undefined or unknown. The primary challenge here is predicting system states with scarce data and minimal understanding of the system's dynamics. Two significant techniques, physics-informed neural networks (PINN) and deep operator networks (DeepONet), are investigated to approach this problem. The proposed method attempts to combine the strengths of both techniques and also integrate findings from prior research on short-term dependencies in systems. Our methodology is bifurcated into two primary stages: discerning system dynamics, with a focus on uncovering short-term dependencies, and honing the accuracy of estimations. In the initial phase, we strive to pinpoint pivotal influences on the present state through an in-depth examination of historical data, emphasizing the role of short-term dependencies. The utility of this exploration is exemplified through a case study on the ultra-high magnetic response system. Subsequently, the latter phase prioritizes mitigating error propagation by establishing a direct correlation between initial states and inputs to prevailing states. This is achieved through a specialized DeepONet framework enriched with knowledge from short-term dependencies.In summary, this dissertation offers novel insights into the challenges and solutions for state estimation in dynamic systems with limited information. The presented methods and findings have broad implications for fields reliant on accurate state estimations.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2024

Mode of access: World Wide Web

ISBN: 9798381406566Subjects--Topical Terms:

679492
Industrial engineering.
Subjects--Index Terms:

Dynamic systemsIndex Terms--Genre/Form:

554714
Electronic books.

Efficient Learning of the State in Dynamic Systems From Model-Based to Data-Based.
LDR:04408ntm a22003857 4500 001 1147305
005 20240909100752.5
006 m o d
007 cr bn ---uuuuu
008 250605s2023 xx obm 000 0 eng d
020 $a 9798381406566
035 $a (MiAaPQ)AAI30691868
035 $a AAI30691868
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Sun, Jieming. $3 1473003
245 1 0 $a Efficient Learning of the State in Dynamic Systems From Model-Based to Data-Based.
264 0 $c 2023
300 $a 1 online resource (121 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 85-07, Section: A.
500 $a Advisor: Li, Lichun.
502 $a Thesis (Ph.D.)--The Florida State University, 2023.
504 $a Includes bibliographical references
520 $a Understanding and predicting states in dynamic systems is a core challenge in many areas of study. This involves trying to estimate an evolving signal based on past observations or the understood rules of the system. The main obstacles include errors in observations, the unknowns, and the complexity of these rules.This dissertation addresses the intricate problem of state estimation in dynamic systems under constraints of limited information. Central to our investigation is the pivotal balance between the amount of available data and the accuracy of state predictions. Broadly, the information about dynamic systems falls into two camps: measurements, which are historical observations, and models, the theoretical frameworks defining system evolution. While dense and precise measurements are desirable for accuracy, they are often constrained by technological limitations. On the flip side, while detailed models can guide precise predictions, they are sometimes elusive for complex systems.Initially, the research delves into model-based estimation, where system models are employed to calculate system states. However, uncertainties in various factors, such as initial conditions or the model itself, introduce errors. A unique scenario within model-based estimation is explored where humans function as sensors in a cyber-physical-human system, sending binary signals based on system performance, shedding light on the challenges of relying on such sparse information.The research then shifts to data-based estimation in scenarios where system models are undefined or unknown. The primary challenge here is predicting system states with scarce data and minimal understanding of the system's dynamics. Two significant techniques, physics-informed neural networks (PINN) and deep operator networks (DeepONet), are investigated to approach this problem. The proposed method attempts to combine the strengths of both techniques and also integrate findings from prior research on short-term dependencies in systems. Our methodology is bifurcated into two primary stages: discerning system dynamics, with a focus on uncovering short-term dependencies, and honing the accuracy of estimations. In the initial phase, we strive to pinpoint pivotal influences on the present state through an in-depth examination of historical data, emphasizing the role of short-term dependencies. The utility of this exploration is exemplified through a case study on the ultra-high magnetic response system. Subsequently, the latter phase prioritizes mitigating error propagation by establishing a direct correlation between initial states and inputs to prevailing states. This is achieved through a specialized DeepONet framework enriched with knowledge from short-term dependencies.In summary, this dissertation offers novel insights into the challenges and solutions for state estimation in dynamic systems with limited information. The presented methods and findings have broad implications for fields reliant on accurate state estimations.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2024
538 $a Mode of access: World Wide Web
650 4 $a Industrial engineering. $3 679492
650 4 $a Engineering. $3 561152
650 4 $a Information science. $3 561178
653 $a Dynamic systems
653 $a Physics-informed neural networks
653 $a Cyber-physical-human system
653 $a Efficient learning
655 7 $a Electronic books. $2 local $3 554714
690 $a 0546
690 $a 0723
690 $a 0537
710 2 $a ProQuest Information and Learning Co. $3 1178819
710 2 $a The Florida State University. $b Industrial and Manufacturing Engineering. $3 1187117
773 0 $t Dissertations Abstracts International $g 85-07A.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30691868 $z click for full text (PQDT)