CNT-시멘트 나노복합체의 특성에 대한 트리기반 앙상블 모델과 인공 신경망의 우열평가 = The Relative Merits of Tree-Based Ensemble Models and Artificial Neural Networks for CNT-Cement Nanocomposite Characteristics

국문 초록 (Abstract)

본 연구에서는 스마트시티의 실시간 구조물 안전진단을 위한 자가감지 형 구조재료로서 CNT-시멘트 나노복합체의 활용 가능성을 확인하기 위 해, 탄소나노튜브(CNT)의 혼입률에 따른 기계적·전기적 특성을 검토하고 머신러닝 모델을 활용한 하중 예측 성능을 평가하였다. 압전저항 특성을 이용하여 외부 하중을 정량적으로 예측하기 위해 인공 신경망(ANN)과 트 리 기반 앙상블 모델인 Random Forest, XGBoost, CatBoost를 적용하여 모 델의 성능을 비교 분석하였다. 반복 하중 실험을 통해 수집된 저항 변형 률 데이터를 학습시킨 결과, 모든 하중 조건에서 트리 기반 앙상블 모델 이 인공 신경망보다 우수한 예측 정확도와 안정성을 나타냈다. 특히 인장 하중 조건에서 Random Forest 모델은 결정계수( ) 0.98이라는 매우 높은 신뢰성을 기록하며 비선형적인 저항 변화와 하중 사이의 상관관계를 가장 효과적으로 학습하는 것으로 나타났다.결론적으로 본 연구를 통해 CNT- 시멘트 나노복합체의 자가감지 성능을 정량적으로 입증하였으며, 제안된 트리 기반 앙상블 모델이 구조물의 응력 상태를 실시간으로 모니터링하기 위한 지능형 센서 시스템 구축에 효과적으로 기여할 수 있음을 확인하였 다. 이는 향후 스마트시티의 주요 인프라 시설물에 적용되어 별도의 외부 센서 없이도 구조체 스스로 위험을 감지하는 자가 진단 기술로 활용될 수 있을 것으로 기대된다.

목차 (Table of Contents)

• I. 서론
• 1. 연구 배경 및 목적 ······························································ 1
• 2. 연구 내용 및 범위 ······························································ 4
• II. 이론적 배경 및 연구동향
• 1. CNT-시멘트 나노복합체 ······················································ 5
• 가. CNT 정의 및 특성 ·························································· 5
• 나. CNT-시멘트 나노복합체 특성 ··········································· 9
• 다. CNT-시멘트 나노복합체 압전저항 연구 동향 ····················· 12
• 2. 압전저항 구조건전성 모니터링 ············································ 14
• 가. 압전저항 구조건전성 모니터링 특성 ································· 14
• 나. 압전저항 구조건전성 모니터링 적용 사례 ·························· 19
• 3. 머신러닝 적용 기술 ·························································· 21
• 가. 머신러닝 활용 가능 알고리즘 요약 ··································· 21
• III. 실험 개요
• 1. 사용 재료 ········································································ 24
• 가. 시멘트 ········································································ 24
• 나. 잔골재 ········································································ 25
• 다. CNT ··········································································· 25
• 라. PCE ·········································································· 27
• 2. 실험 계획 ········································································ 28
• 가. 나노복합체 플로우 측정 ················································· 29
• 나. 나노복합체 제작 방식 ···················································· 31
• 다. 나노복합체 강도 측정 ···················································· 31
• 라. 나노복합체 전기전도도 및 저항 측정 ································ 32
• 마. 나노복합체 압전저항 측정 ·············································· 33
• 3. 실험 방법 ········································································ 34
• 가. 시험체 제작 ································································· 34
• 나. 강도 특성 ···································································· 40
• 다. 전기적 특성 ································································· 43
• 라. 압전저항 특성 ······························································ 46
• IV. 실험결과 및 고찰
• 1. CNT-시멘트 나노복합체 제조 특성 ······································· 48
• 가. CNT 혼합량에 따른 유동 및 강도 특성 ····························· 48
• 나. CNT 혼합량에 따른 전기적 특성 ···································· 56
• 2. CNT-시멘트 나노복합체 압전저항 특성 ································· 58
• 가. 압축, 인장, 휨에 따른 압전저항 특성 ································ 58
• 나. 압전저항 데이터를 기반으로 한 모델별 예측 특성 ·············· 82
• 3. 인공지능 기반 응력-압전저항 예측 모델 ······························· 85
• 가. 인공지능 기반 응력-압전저항 예측 모델 학습 ···················· 85
• 나. 교차검증 기반 응력-압전저항 모델 성능 비교 ···················· 90
• V. 결론 ················································································ 111
• 참고문헌 ············································································· 113
• ABSTRACT ·········································································· 117