Smart Aggregate's Impedance Signals and CNN Deep Learning-based Concrete Structural Health Monitoring Technique
- Abstract
- This study aims to develop a concrete structural health monitoring (SHM) method using smart aggregate’s impedance responses and deep learning. The following approaches are implemented to achieve the goal. Firstly, theories of impedance monitoring and deep learning techniques are presented. Secondly, smart aggregate sensors for the impedance method of concrete structure are presented. The feasibility of smart aggregate sensors is numerically and experimentally evaluated for EMI monitoring under different applied stresses. Thirdly, a deep regression learning method is developed to process smart aggregate’ impedance signals for concrete stress monitoring. The performance of the proposed method is extensively evaluated by investigating the effects of signal noises and untrained data on the accuracy of stress estimation in smart aggregate-embedded concrete cylinders. Then, cross-evaluation of trained deep learning models is conducted to verify the practicality of smart aggregate sensors embedded in concrete structures. Fifthly, an integrated stress and damage estimation method in concrete structures via deep regression and classification learning is proposed. The robustness and generalization of the proposed model are investigated under noise data, untrained cases, and cross-evaluation. Sixthly, numerical and experimental analysis of a smart aggregate-embedded prestressed concrete (PSC) anchorage under various prestressing (PS) forces is investigated to identify the relation between impedance responses and inner damage occurrence. Then, pre-trained deep learning models are practically implemented to predict the stress and damage existing in a PSC anchorage via processing smart aggregate’s EMI data under PS force variations.|본 연구는 스마트 골재의 임피던스 응답과 딥러닝을 이용한 콘크리트 구조 건전성 모니터링(SHM) 방법을 개발하는 것을 목표로 한다. 이러한 목표를 달성하기 위해 다음과 같은 접근 방식을 구현한다. 첫째, 임피던스 모니터링 이론과 딥러닝 기법을 제시한다. 둘째, 콘크리트 구조물의 임피던스 방법을 위한 스마트 골재 센서를 제시한다. 다양한 적용 응력 하에서 EMI 모니터링을 위해 스마트 골재 센서의 타당성을 수치적, 실험적으로 평가한다. 셋째, 콘크리트 응력 모니터링을 위해 스마트 골재의 임피던스 신호를 처리하기 위한 딥 회귀 학습 방법을 개발한다. 제안된 방법의 성능은 신호 노이즈와 훈련되지 않은 데이터가 스마트 골재 매립 콘크리트 실린더에서 응력 추정의 정확도에 미치는 영향을 조사하여 광범위하게 평가한다. 그런 다음 훈련된 딥러닝 모델의 교차 평가를 수행하여 콘크리트 구조물에 매립된 스마트 골재 센서의 실용성을 검증한다. 다섯째, 딥 회귀와 분류 학습을 통한 콘크리트 구조물의 통합 응력 및 손상 추정 방법을 제안한다. 제안된 모델의 견고성과 일반화는 노이즈 데이터, 훈련되지 않은 사례, 교차 평가 하에서 조사됩니다. 여섯째, 다양한 프리스트레싱(PS) 힘 하에서 스마트 골재 매립 프리스트레스 콘크리트(PSC) 앵커리지의 수치적 및 실험적 분석을 조사하여 임피던스 응답과 내부 손상 발생 간의 관계를 파악합니다. 그런 다음 사전 훈련된 딥 러닝 모델을 실제적으로 구현하여 PS 힘 변화 하에서 스마트 골재의 EMI 데이터를 처리하여 PSC 앵커리지에 존재하는 응력과 손상을 예측합니다.
- Author(s)
- TA QUOC BAO
- Issued Date
- 2025
- Awarded Date
- 2025-02
- Type
- Dissertation
- Keyword
- structural heath monitoring, concrete structure, impedance, PZT, smart aggregate, deep learning
- Publisher
- 국립부경대학교 대학원
- URI
- https://repository.pknu.ac.kr:8443/handle/2021.oak/34028
http://pknu.dcollection.net/common/orgView/200000868023
- Alternative Author(s)
- Quoc-Bao Ta
- Affiliation
- 국립부경대학교 대학원
- Department
- 대학원 해양공학과
- Advisor
- Kim Jeong Tae
- Table Of Contents
- I. INTRODUCTION 1
1.1 Overview 1
1.2 SHM in Concrete Structures 2
1.3 Research Needs on Concrete SHM Using Impedance and Deep Learning Techniques 8
1.4 Objective and Approaches 10
1.5 Thesis Organization 12
II. THEORY OF IMPEDANCE MONITORING AND DEEP LEARNING 14
2.1 Overview 14
2.2 Theory of Impedance Monitoring 14
2.3 Theory of Deep Learning for Concrete SHM 26
III. SMART AGGREGATE SENSORS FOR IMPEDANCE METHOD CONCRETE STRUCTURE 36
3.1 Overview 36
3.2 Solid Aggregate Sensor 36
3.3 Capsule Aggregate Sensor 49
3.4 Discussion 63
IV. CONCRETE STRESS ESTIMATION METHOD USING 1D CNN DEEP LEARNING OF SMART AGGREGATE’S SIGNALS 66
4.1 Overview 66
4.2 1D CNN Deep Regression Learning Model for Concrete Stress Monitoring 66
4.3 1D CNN Deep Regression Learning of SA’s Impedance Signals 72
4.4 1D CNN Deep Regression Learning of CA’s Impedance Signals 86
4.5 Summary and Remarks 101
V. INTEGRATED ESTIMATION METHOD OF STRESS AND DAMAGE IN CONCRETE STRUCTURE USING 1D CNN DEEP LEARNING OF SMART AGGREGATE’S SIGNALS 103
5.1 Overview 103
5.2 1D CNN-based Integrated Stress and Damage Method for Concrete Stress Monitoring 103
5.3 1D CNN Deep Regression and Classification Learning of CA’s Impedance Signals 109
5.4 Cross Evaluation of Trained CNN_RC Models 126
5.5 Summary and Concluding Remarks 129
VI. DAMAGE MONITORING IN PSC ANCHORAGE USING DEEP LEARNING OF SMART AGGREGATE’S SIGNALS 130
6.1 Overview 130
6.2 Scheme of Damage Monitoring for PSC Anchorage 130
6.3 Numerical Analysis of Impedance-Damage Relation in PSC Anchorage 132
6.4 Experimental Verification of PSC Anchorage Zone 142
6.5 Implementation of Pre-trained 1D CNN Models for Damage Monitoring in CA-embedded PSC Anchorage 152
6.6 Analysis and Discussion 162
6.7 Summary and Remarks 166
VII. SUMMARY AND CONCLUSIONS 168
APPENDIX A. LEARNING PERFORMANCE OF 1D CNN REGRESSION AND CLASSIFICATION MODEL 171
REFERENCES 174
CURRICULUM VITAE 179
ACKNOWLEDGMENTS 182
- Degree
- Doctor
-
Appears in Collections:
- 대학원 > 해양공학과
- Authorize & License
-
- Authorize공개
- Embargo2026-01-01
- Files in This Item:
-
Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.