Earticle

본 연구는 인공지능(AI) 시대에 확산되는 조직 내 감시·검열 메커니즘을 분석하고, 국내외 실제 사례를 통해 사회적·윤리적 함의를 도출한다. DLP 시스템, EDR 솔루션, AI 기반 생산성 모니터링, 자동 검열 메커니즘을 포함한 네 가지 감시 체계를 분석하고, 프랑스 아마존, 스타벅스 등 국내외 사례를 비교 검토하였다. 연구 결과, 과도한 AI 기반 감시·검열은 근로자의 프라이버시 침해, 표현의 자유 위축, 데이터 권력 불균형 등 윤리적 문제를 야기하며, 현행 법제의 미비점이 나타났다. 이에 GDPR의 DPIA 제도 도입, 기업 감시 최소화 원칙, 근로자 참여 보장 등 법적·제도적 개선 방안과 함께 국제 AI 윤리 원칙의 조직내 적용 방안을 제시한다. 또한 프라이버시 중심 설계(Privacy by Design)와 직장 내 AI 감시 영향평가(WASIA) 기술 규격을 제안하여 규범적 균형의 기술적 구현 방안을 함께 제시한다.

This study analyzes the mechanisms of organizational surveillance and censorship expanding in the age of artificial intelligence (AI), and derives socio-ethical implications through domestic and international cases. Four surveillance mechanisms—DLP systems, EDR solutions, AI-based productivity monitoring, and automated censorship—are analyzed, and cases from Amazon France, Starbucks, and Korean companies are comparatively reviewed. The findings indicate that excessive AI-based surveillance and censorship raises ethical concerns including privacy violations, suppression of freedom of expression, and data power imbalances, while inadequacies in current legal frameworks are identified. This paper proposes legal and institutional improvements including the introduction of GDPR-style DPIA, adoption of a monitoring minimization principle, and guarantees of worker participation, along with recommendations for applying international AI ethics principles within organizations. Additionally, this study proposes a technical implementation framework based on Privacy by Design principles and introduces the Workplace AI Surveillance Impact Assessment (WASIA) as a structured technical specification for achieving normative balance in practice.

본 연구는 건설 현장의 CCTV를 활용한 실시간 침수 예측 시스템을 개발하는 것을 목표로 한다. CCTV로부터 침수 깊이를 추정하기 위한 모델을 구축하고, 추정된 침수 깊이를 관측 데이터와 비교하 여 성능을 검증하였다. 분석 결과 높은 상관관계(R2 = 0.99)와 약 3cm의 RMSE를 보여 정량적 침수 모니터링에 CCTV를 활용할 수 있는 가능성을 확인했다. 또한, 취약 지역에 대한 실시간 침수 예측 방법을 제안했으며 XP SWMM 시뮬레이션에서 도출된 침수 특성을 활용하여 ANN(인공 신경망)과 CNN(합성곱 신경망)을 기반으로 한 모델을 학습시켰다. 모델의 성능은 평균 절대 백분율 오차 (MAPE)를 사용하여 평가되었으며, 전체 평균 오차율은 침수면적 예측에서 8.89%, 그리드 기반 침수 깊이 예측에서 19.49%가 나타났다. 향후 연구는 실시간 CCTV 침수 모니터링과 AI 모델을 통합하여 예측 정확도를 높일 수 있을 것이라 판단된다.

This study aims to develop a real-time inundation prediction system using CCTV at construction sites. A model was established to estimate inundation depth from CCTV, and its performance was validated by comparing the estimated inundation depth with observed data. The results demonstrated a high correlation (R2 = 0.99) and an RMSE of approximately 3 cm, confirming the feasibility of using CCTV for quantitative inundation monitoring. Furthermore, a real-time inundation prediction method for vulnerable areas was proposed. inundation characteristics derived from XP SWMM simulations were used to train a model based on ANN (Artificial Neural Networks) and CNN (Convolutional Neural Networks). The model's performance was evaluated using the Mean Absolute Percentage Error (MAPE), with overall average error rates of 8.89% for inundation area predictions and 19.49% for grid-based inundation depth predictions. Future efforts will focus on integrating real-time CCTV inundation monitoring with the AI model to enhance its predictive accuracy.

As Industry 4.0 revolutionizes manufacturing, protecting worker safety, particularly in high-risk, multicultural environments, has become crucial. This study presents a conceptual model for an AI-driven safety monitoring system that combines biometric wearables, environmental sensors, behavior recognition, and multilingual alerts. It allows for real-time hazard detection and inclusive communication, all managed through a centralized dashboard. A simulation of oxygen depletion in a confined space demonstrates the system's ability to respond quickly. The model is compatible with Industry 4.0 platforms like MES, ERP, and Digital Twins, and supports migrant workers by breaking down language barriers with multilingual alerts and pictograms. The framework aligns with ISO 45001 and OSHA standards, and its modular, scalable design enables predictive risk mitigation and smart safety innovation. This framework sets the stage for future prototyping and field implementation.

This study introduces an innovative, integrated safety monitoring and rescue management system designed specifically for high-risk confined spaces in industrial environments. The system utilizes wearable health sensors, environmental detection modules, AI-driven predictive analytics, and multilingual alert mechanisms to deliver real-time hazard detection and automatic emergency responses. With a strong focus on inclusivity, it provides communication in native languages, greatly aiding foreign workers who may encounter language barriers during emergencies. Its core features include automatic ventilation management, toxic gas suppression, and mechanical extraction of workers, ensuring quick action in critical situations. The ergonomic design combined with AI-based logic allows for early risk detection before issues escalate. Additionally, the system produces comprehensive incident reports for analysis and safety audits. Based on occupational safety research and aligned with international safety standards, it aims to overcome existing limitations in industrial emergency protocols and foster a more inclusive, proactive safety culture in diverse workplaces.

최근 대형 언어 모델(LLM)의 성능이 전문가 수준으로 발전하고 있다. 본 연구는 고객서비스 도메인에서 기존 인간 중심 평가 방식의 한계를 극복하고자 Expert AI(평가 전문가로서의 역할을 하는 AI)를 활용한 품질 모니터링 시스템 구축 방안을 새롭게 제시하고자 한다. 본 연구에서는 고객 질문에 대해 “고객서비스 도메인 특화된 Local LLM 기반 AI”로부터 생성된 응답 데이터 결과물의 품질을 인간 평가자와 Expert AI(AI 평가자)가 평가하도록 하였다. 단, Local LLM의 품질 평가 시, 다른 LLM AI(ChatGPT와 Perplexity)로부터 답변된 응답과 비교하여 평가되도록 하여, 도메인 특화로 구축된 Local LLM의 현재 품질 상황과 개선할 점을 파악할 수 있도록 하였다. 평가 수행 이후, 인간 평가자와 Expert AI의 품질 평가에 대한 일치도를 분석한 결과, 92.4%에 달하는 높은 일치율을 확인하였다. 특히 Local LLM의 문제점 탐지(응답 품질이 나쁜 경우)에서는 98.9%의 일치율을 보여, Expert AI가 핵심 품질 관리 역할을 수행할 수 있음을 입증하였다. 본 연구는, 지속적이면서 실시간 품질 관리가 필요한 고객서비스 분야에서 Expert AI의 적용 가능성을 실증적으로 검증함으로써, 실무 환경에서 효율성을 극대화할 수 있는 AI 기반 품질 모니터링 시스템 구축 방안을 제안한다.

Recently, the performance of large language models (LLMs) has improved to expert levels. This study proposes a novel approach to building a quality monitoring system utilizing Expert AI (AI acting as an evaluation expert) to overcome the limitations of existing human-centered methods in the customer service (CS) domain. In this study, the quality of responses to customer inquiries generated by a “CS domain-specific Local LLM-based AI” was assessed by human evaluator and Expert AI. For the quality assessment of the Local LLM, its responses were compared with those generated by other LLM-based AIs (ChatGPT and Perplexity) to evaluate the current quality and identify areas for improvement in the Local LLM. After the evaluation, we analyzed the agreement between human and AI evaluators in quality assessment, confirming a high agreement rate of 92.4%. In particular, the agreement rate for detecting Local LLM issues (e.g., poor response quality) was 98.9%, demonstrating that Expert AI can play a key quality control role. This study empirically verified the applicability of Expert AI in the CS field, which requires continuous and real-time quality management. It proposes a method for building an AI-based quality monitoring system that can maximize efficiency in corporate environments.

본 논문은 가스 누출 감지와 거주자의 비정상 행동 분석을 결합한 AI 기반 스마트 안전 모 니터링 시스템을 제안한다. 기존 가스 감지 시스템의 오경보 문제와 행동 감지 시스템의 한계를 극 복하기 위해 LSTM, CNN, Random Forest 모델을 결합하여 위험 예측의 정확도를 향상시켰다. 실험 결과, 최종 위험 감지 정확도는 94.1%, 오경보 발생률은 22.3%에서 7.8%로 감소되었다. 특히, AI 모 델을 통해 가스 농도의 급격한 변화와 거주자의 비정상 행동을 종합적으로 분석하여 실제 위험 상 황을 보다 정확하게 판단하고, 실시간으로 위험을 감지하고 자동으로 긴급 구조 요청이 가능함을 입 증했다. 또한, 가스 누출 및 행동 패턴에 따른 위험도를 다단계로 평가하여 대응의 신속성과 정확성 을 높였다. 본 연구는 가정, 산업 현장, 스마트홈 환경에서 실용적으로 적용 가능하며, 향후 다양한 센서 추가, 맞춤형 AI 모델 적용, 실제 환경에서의 장기적 성능 평가를 통한 지속적인 개선 방향을 논의한다. 이를 통해 인명 피해를 최소화하고 더욱 안전한 생활 환경 조성에 기여할 수 있을 것으로 기대된다.

This paper proposes an AI-based smart safety monitoring system that integrates gas leak detection and abnormal behavior analysis to enhance risk prediction accuracy. To address the limitations of traditional gas detection systems with high false alarm rates and the restricted capabilities of behavior detection systems, the study combines LSTM, CNN, and Random Forest models. Experimental results show that the proposed system achieves a final risk detection accuracy of 94.1% and reduces the false alarm rate from 22.3% to 7.8%. Notably, the AI models enable a comprehensive analysis of rapid changes in gas concentration and abnormal behavior patterns, allowing for more accurate risk assessment and real-time emergency response automation. Additionally, the system evaluates risk levels in multiple stages based on gas leaks and behavior patterns, enhancing the speed and accuracy of response. The findings suggest that the proposed system can be practically applied in homes, industrial sites, and smart home environments. Future research will focus on integrating various sensors, developing personalized AI models, and conducting long-term performance evaluations in real-world settings. This approach is expected to minimize casualties and contribute to creating safer living environments.

본 연구는 상처 부위의 온도·습도·압력 변화와 상처 이미지 정보를 함께 활용하여 상처 상태를 보조적으로 모니터링할 수 있는 AI 기반 스마트 거즈 시스템을 개발하고자 하였다. Arduino UNO, DHT22, FSR406, HC-05 블루투스 모듈과 App Inventor 기반 모바일 앱을 이용하여 생체환경 데이터를 수집하였고, Kaggle 상처 분할 데이터와 DFUTissueSegNet 데이터를 활용하여 U-Net 기반 상처 영역 및 조직 분할 모델을 학습하였다. 분석 결과, 상처 영역 분할 모델은 Dice score 0.792를 보였으며, 조직 분할 모델은 Class 1 0.492, Class 2 0.916, Class 3 0.665, 평균 0.691의 Dice score를 나타냈다. 또한 이미지 분석 결과와 센서 요약값을 결합한 rule-based 판단 과정을 통해 정상 회복 경향, 관리 필요, 감염 의심 상태를 구분하였다. 본 연구는 상처 상태 변화 확인과 의료진 확인 필요성을 보조하는 초기 시스템으로 활용 가능성이 있다.

This study aimed to develop an AI-based smart gauze system that assists wound monitoring by combining temperature, humidity, pressure, and wound image information. Biological environmental data were collected using Arduino UNO, DHT22, FSR406, HC-05 Bluetooth modules, and an App Inventor-based mobile application. In addition, U-Net-based wound area and tissue segmentation models were trained using Kaggle wound segmentation data and DFUTissueSegNet data. The wound area segmentation model achieved a Dice score of 0.792, while the tissue segmentation model showed Dice scores of 0.492 for Class 1, 0.916 for Class 2, 0.665 for Class 3, and an average Dice score of 0.691. A rule-based decision process combining image analysis results and sensor summary values classified normal recovery tendency, management-needed status, and suspected infection status. This system may support wound status monitoring and medical staff confirmation.

본 연구는 AI 기술을 활용한 배터리 재활용 이력 추적 및 탄소 배출 모니터링 시스템 개발을 목표로 한다. 주요 과제는 재활용 공정의 탄소 배출량을 효율적으로 측정하고 관리하기 위한 데이터 분석 체계의 확립이다. 연구 결과, 제안 된 LSTM-XGBoost 하이브리드 모델은 기존 정적 LCA 방식 대비 RMSE를 68.9% 감소시키고 결정계수(  ) 0.94를 달성 하여 산정 정밀도를 획기적으로 향상시켰다. 또한, 하이퍼레저 패브릭 기반 블록체인은 초당 150건 이상의 트랜잭션(TPS) 처리 속도와 2초 이내 지연 시간을 기록해 실시간 모니터링의 기술적 타당성을 입증하였다. 이는 배터리 재활용 분야 내 AI 적용 가능성을 높이고 환경 문제 해결의 새로운 접근법을 제시한다. 본 연구는 재활용 산업의 지속 가능성을 촉진하고 향후 탄소 감축 전략 수립의 기반 자료로 활용되어, 투명하고 친환경적인 리사이클링 시스템 구축에 기여할 것으로 기대 된다.

This study aims to develop an AI-based system for tracking battery recycling history and monitoring carbon emissions. The primary challenge is establishing efficient data analysis methods to manage emissions throughout the recycling process. Experimental results show that the proposed LSTM-XGBoost hybrid model significantly improved estimation precision, achieving an $R^2$ of 0.94 and reducing the Root Mean Square Error (RMSE) by 68.9% compared to static LCA methods. Furthermore, the Hyperledger Fabric-based blockchain verified its feasibility for real-time monitoring, recording a throughput of over 150 transactions per second (TPS) with a latency under 2 seconds. These findings demonstrate the applicability of AI in the battery recycling sector, offering a novel approach to environmental challenges. This study is expected to promote sustainable development in the recycling industry and serve as foundational data for future carbon reduction strategies, marking a significant milestone for transparent and eco-friendly recycling systems.

다수의 어르신이 생활하는 노인복지시설에서는 고령으로 인한 인지 및 운동 능력 저하로 낙상사고가 종종 발생한다. 중증 및 사망으로 이어질 수 있는 낙상사고가 발생한 경우, 골든아워를 지키기 위해 신속하게 사고 발생 여부를 파악하고 조치를 취해야 한다. 이를 위해 일반적인 RGB 카메라를 이용한 CCTV를 사용하고 이를 모니터링하는 전담인력을 배치할 수 있으나, 환자의 사생활 침해 문제 및 전담 인력의 피로 문제가 발생한다. 이러한 문제를 해결하기 위해 본 논문에서는 개인의 식별이 불가능한 열화상 영상 및 음향 정보에서 낙상을 탐지하는 인공지능을 이용하여 실시간으로 낙상사고 발생 여부를 판단한다. 또한, 낙상사고가 발생한 경우 신속한 조치를 위해 의료 관계자 및 보호자에게 자동으로 신고하는 플랫폼을 구현하였다.

In elderly care facilities where many seniors reside, safety accidents frequently occur due to age-related cognitive and motor impairments. In cases of serious incidents such as falls—which can lead to severe injury or death—it is critical to detect aㄸnd respond promptly within the golden hour. While standard CCTV systems using RGB cameras and dedicated monitoring personnel can be employed to identify such incidents, they raise concerns regarding patient privacy and staff fatigue. To address these issues, this paper proposes a real-time AI-based fall detection system that utilizes thermal imaging and audio data, which do not reveal personal identity. Furthermore, we detail the implementation of a platform that automatically notifies medical staff and guardians in the event of an accident to enable swift intervention.

연구목적: 본 연구는 작업자 안전을 고려하고 제조기업의 생산성 향상, 품질 향상을 위해 디지털 트윈 기반의 AI 비전, 제조 실행 시스템, 위치 측위 기술을 활용한 지능형 제조공정 통합 모니터링 시스템을 개발하는데 목적이 있다. 연구방법: 제조공정 특성상 사출성형기 기반의 생산 작업을 고려하여 생산라 인 구조에 맞는 설계 및 적용을 통해 불량품 검출에 특화되도록 실시간 영상 기반의 AI 감지 기술을 적 용하였다. 연구결과: 제안된 방법을 통해 시스템 개발 및 적용하였고, 시스템 도입 전/후 생산 PPM 비교 하여 불량률 감소에 따른 생산성 향상을 확인하였다. 또한, 시스템 응답 속도 측정을 통해 현장 사고 발생 시 신속한 초기대응 및 상황전파가 가능한 것을 확인하였다. 결론: 본 연구의 시스템은 다양한 제조기업 이 제조공정 시스템 도입 시 의사결정을 지원하는 기초자료로 활용될 수 있을 것으로 기대된다.

Purpose: This study aims to enhance productivity in manufacturing companies while prioritizing worker safety and developing a digital twin-based AI vision to improve quality, creating manufacturing execution systems, and implementing an integrated monitoring system for intelligent manufacturing processes that leverages location positioning technologies. Method: Considering the characteristics of the manufacturing process, particularly for tasks involving injection molding machines, real-time video-based AI detection technology was applied to tailor its design and application to the production line structure, enhancing the specialized detection of defective products. Result: Using the proposed method, the system was developed and implemented, and a comparison of production PPM before and after its adoption demonstrated increased productivity due to reduced defect rates. Additionally, measurements of system response speed confirmed its ability to provide a swift initial response and rapidly disseminate information in the event of an on-site accident. Conclusion: The system developed in this study is expected to serve as a foundation for supporting decision-making in the implementation of manufacturing process systems across various companies.

연구목적: 본 논문은 건설 시공현장에서 발생하는 사고 및 잠재적 위험분석을 위한 IoT 및 CCTV 기반 안전모니터링을 실시하고 추락, 충돌 등 위험 또는 이상현상을 탐지하여 무전기 등을 이용한 예･경보 및 챗봇서비스를 구축하는 방법을 제시하는데 목적이 있다. 연구방법: 건설현장 스마트 건설기술 사례 및 문헌분석을 통하여 안전관리 모델을 제시한다. 연구결과: ‘건설사고 통계’에 따르면 2021년 건설업 사고재해자는 26,888명으로 전체 사고재해의 26.3%가 건설업에서 발생하였고, 건설업 안전사고 사망 자는 417명으로 전체 산업재해 사망자의 50.5%에 달한다. 이런한 건설재해의 개선 방안으로, IoT 건전 성모니터링 기반 스마트 건설기술을 활용한 건설현장 안전관리 AI 챗봇서비스를 제시한다. 근로자 등 이해관계자가 참여하는 건설현장은 비계공정 및 개구부, 위험기계기구류 접근 등 사업장 내부 주요 위 험구역을 선정하여 인공지능 챗봇시스템을 구현하여 실증하였다. 결론: 건설현장 인공지능 챗봇서비 스 실증결과에 대한 참여근로자의 만족도 조사에서 90점 이상을 받아 상업화 가능성을 확인하였다.

Purpose: This paper conducts IoT and CCTV-based safety monitoring to analyze accidents and potential risks occurring at construction sites, and detect and analyze risks such as falls and collisions or abnormalities and to establish a system for early warning using devices like a walkie-talkie and chatbot service. Method: A safety management service model is presented through smart construction technology case studies at the construction site and review a relevant literature analysis. Result: According to ‘Construction Accident Statistics,’ in 2021, there were 26,888 casualties in the construction industry, accounting for 26.3% of all reported accidents. Fatalities in construction-related accidents amounted to 417 individuals, representing 50.5% of all industrial accident-related deaths. This study suggests implementing AI chatbot services for construction site safety management utilizing IoTbased health monitoring technologies in smart construction practices. Construction sites where stakeholders such as workers participate were demonstrated by implementing an artificial intelligence chatbot system by selecting major risk areas within the workplace, such as scaffolding processes, openings, and access to hazardous machinery. Conclusion: The possibility of commercialization was confirmed by receiving more than 90 points in the satisfaction survey of participating workers regarding the empirical results of the artificial intelligence chatbot service at construction sites.

본 연구는 한국 건설산업의 고질적인 병폐인 불법하도급 문제를 해결하기 위해 기존의 수기 및 서류 중심 단속 체계가 가진 한계점을 분석하고, 국가 행정정보와 인공지능(AI) 기술을 결합한 지능형 감시체계의 설계 방향을 제안하였다. 현재 우리나라의 불법하도급 단속은 연간 17만 건이 넘는 방대한 공사 물량에 비해 전담 인력은 10여 명에 불과하여 물리적인 감시 사각지대가 발생할 수밖에 없는 심각한 구조적 제약에 직면해 있다. 이러한 자원의 부족은 위장계약이나 다단계 재하도급과 같이 정교화된 위법 행위를 실시간으로 포착하는 것을 어렵게 만든다. 이러한 문제들을 해결하기 위해 본 연구는 국세청의 전자세금계산서, 고용노동부의 고용 및 4대 보험 자료, 그리고 건설기계 등록정보 등 여러 기관에 산재한 행정정보를 통합하는 AI 기반 감시 모형을 제시하였다. 다차원적인 데이터 세트에 대한 실시간 교차검증을 통해 AI는 계약 서류와 실제 현장 투입 주체(근로자 및 장비) 간의 불일 치나 비정상적인 자금 흐름과 같은 이상 징후를 자동으로 식별할 수 있다. 본 연구는 ‘위장계약’, ‘이면계약’, ‘다단계 재하도급’에 대한 구체적인 탐지 시나리오를 상세화함으로써 기술적 구현을 위한 실질적인 윤곽을 수립하였다. 나아가 본 연구는 AI 감시시스템의 조속한 구축, 범정부 차원의 정보공유 플랫폼 마련, 민간 공사 하도급 계약등록 의무화, 그리고 특별사법경찰권 도입을 통한 단속 권한 강화 등을 포함하는 종합적인 정책 패키지를 제안하였다. 본 연구는 기술적 혁신, 제도적 개선, 행정적 효율성을 통합함으로써 단속의 정확도를 높이고 건설 관리의 공정성과 안전성을 확보하기 위한 전략적 대안을 제공한다는 점에서 학술적·정책적으로 중요한 의의를 지닌다.

This study analyzes the limitations of existing manual and document-based enforcement systems to address the chronic issue of illegal subcontracting in the Korean construction industry. It proposes an intelligent monitoring framework that integrates national administrative data with Artificial Intelligence (AI) technology. Currently, enforcement faces severe structural constraints; with over 170,000 annual construction projects but only about 10 dedicated personnel, vast monitoring blind spots are inevitable. This lack of resources makes it difficult to detect sophisticated illegalities such as disguised contracts or multi-level subcontracts in real time. To solve these problems, this research proposes an AI-driven monitoring model that integrates fragmented administrative data—including tax invoices from the National Tax Service, employment and four major insurance data from the Ministry of Employment and Labor, and machinery registration information. By performing real-time cross-verification of these multi- dimensional datasets, the AI can automatically identify anomalies, such as discrepancies between contract documents and actual on-site entities (laborers and equipment) or irregular financial patterns. The study establishes a concrete technical outline by detailing specific detection scenarios for "sham contracts," "back-door contracts," and "multi-level subcontracting". Furthermore, the study recommends a comprehensive policy package: the rapid deployment of an AI monitoring system, the legalization of a government-wide data-sharing platform, mandatory subcontract registration for private projects, and the strengthening of enforcement through special judicial police powers. This research holds significant academic and policy value as it provides a strategic framework to enhance enforcement accuracy, promote fairness, and ensure safety in construction management by integrating technology, institutional reform, and administrative efficiency.

This paper deals with research on innovative systems using Python-based artificial intelligence technology in the field of plant growth monitoring. The importance of monitoring and analyzing the health status and growth environment of plants in real time contributes to improving the efficiency and quality of crop production. This paper proposes a method of processing and analyzing plant image data using computer vision and deep learning technologies. The system was implemented using Python language and the main deep learning framework, TensorFlow, PyTorch. A camera system that monitors plants in real time acquires image data and provides it as input to a deep neural network model. This model was used to determine the growth state of plants, the presence of pests, and nutritional status. The proposed system provides users with information on plant state changes in real time by providing monitoring results in the form of visual or notification. In addition, it is also used to predict future growth conditions or anomalies by building data analysis and prediction models based on the collected data. This paper is about the design and implementation of Python-based plant growth monitoring systems, data processing and analysis methods, and is expected to contribute to important research areas for improving plant production efficiency and reducing resource consumption.

This study proposes an AI-based wireless network temperature measurement system designed to capture and monitor fine-grained temperature variations within a grid-partitioned indoor environment. Unlike conventional thermostats that provide only a few point measurements, the proposed system employs a dense array of wireless sensor nodes distributed across a spatial grid to achieve high-resolution spatial coverage. Sensor readings are transmitted in real time to a centralized data aggregation server, which performs data cleaning, anomaly filtering, and integration with historical records. The core of the approach is an LSTMbased prediction model capable of short-horizon forecasting of temperature dynamics in each grid cell by leveraging temporal dependencies across multi-sensor data streams. By combining dense sensing, robust preprocessing, and predictive analytics, the system enhances spatial resolution and improves both the accuracy and responsiveness of indoor climate control. Early detection of anomalies such as localized hotspots or cooling inefficiencies allows proactive HVAC adjustments that reduce energy consumption, mitigate thermal discomfort, and prevent operational faults. Field experiments demonstrate that the proposed framework delivers real-time forecasting with low latency and high detection sensitivity, supporting adaptive decision-making in dynamic indoor environments. This integrated sensing-and-prediction paradigm represents a scalable and energy-aware solution for next-generation smart buildings, contributing to improved occupant comfort, sustainable energy management, and overall operational efficiency.

This is a research thesis on the development of a monitoring and diagnosis device that prevents the risk of an accident through monitoring and diagnosis of a modular Main Circuit Breaker (MCB) using Vacuum Interrupter (VI) for Korean high-speed rolling stock. In this paper, a comprehensive MCB monitoring and diagnosis was performed by converging vacuum level diagnosis of interrupter, operating coil monitoring of MCB and environmental temperature/humidity monitoring of modular box. In addition, to develop an algorithm that is expected to have a similar data processing before the actual field test of the MCB monitoring and diagnosis device in 2023, the cluster analysis and factor analysis were performed using the WEKA data mining technique on the big data of Korean railroad transformer, which was previously researched by Tae Hee Evolution with KORAIL.

In this paper, for the purpose of designing an real-time unmanned monitoring system, the YOLOv5s (small) object detection model was applied on the NVIDIA TX2TM AI (Artificial Intelligence) edge computing platform in order to design the fundamental function of an unmanned monitoring system that can detect objects in real time. YOLOv5s was applied to the our real-time unmanned monitoring system based on the performance evaluation of object detection algorithms (for example, R-CNN, SSD, RetinaNet, and YOLOv5). In addition, the performance of the four YOLOv5 models (small, medium, large, and xlarge) was compared and evaluated. Furthermore, based on these results, the YOLOv5s model suitable for the design purpose of this paper was ported to the NVIDIA TX2TM AI edge computing system and it was confirmed that it operates normally. The real-time unmanned monitoring system designed as a result of the research can be applied to various application fields such as an security or monitoring system. Future research is to apply NMS (Non-Maximum Suppression) modification, model reconstruction, and parallel processing programming techniques using CUDA (Compute Unified Device Architecture) for the improvement of object detection speed and performance.

Objectives: Diabetic foot ulcer (DFU) is a critical complication of diabetes that can lead to severe outcomes such as infection,amputation, and increased mortality if left untreated. Early detection and continuous monitoring are essential but remainchallenging, especially in resource-limited settings such as India. This study developed and validated a deep learning algorithmto classify diabetic foot images into severity grades based on the International Working Group on the Diabetic Footclassification: grade 0 (healthy), grade 1 (mild), grade 2 (moderate), and grade 3 (severe). Methods: A dataset of 407 clinicalimages was collected from open-source platforms and clinics in South India and expanded to 612 images through dataaugmentation. The dataset was divided into training (70%), validation (15%), and testing (15%) subsets. Multiple machinelearning models were tested, including MobileNet_V2, EfficientNet-b0, DenseNet121, ResNet_50, VGG16, and ViT_b_16. Results: Among the evaluated models, MobileNet_V2 demonstrated the highest validation accuracy (82%) and achieved anF1-score of 79% on the test set. Although the model showed strong training accuracy, minor overfitting was observed, particularlyin distinguishing adjacent severity grades. To address this, dropout, batch normalization, and early stopping wereemployed. Overall, the model generalized well, showing high accuracy in detecting healthy cases and acceptable performanceacross ulcer severity grades. Conclusions: This study underscores the potential of machine learning-based tools to supportfrontline healthcare workers and facilitate patient self-monitoring in low-resource environments. Future work will focus onrefining the model and integrating it into user-friendly applications.

A reactor protection system (RPS) is a core safety system for the stable operation of nuclear power plants (NPPs). Modern RPSs adopt digital platforms based on programmable logic controllers, offering enhanced reliability and maintainability compared with analog systems. Despite these advancements, current practices such as self-diagnosis functions, periodic operational tests, and scheduled maintenance remain limited in assessing component conditions during normal operation. To address this limitation, this study proposes an artificial intelligence (AI)-based condition-monitoring framework to improve the maintenance strategy for the internal electronic components of digital RPSs. The framework combines a rule-based model to ensure data integrity with an AI-based model to monitor component condition. The methodology was specifically applied to the photocoupler, a critical electronic component whose failure can significantly affect the digital RPS. To train and validate the AI-based model, a scenario dataset was generated using accelerated aging data for photocouplers to simulate the actual operating environment. In addition, a conceptual monitoring interface was designed to evaluate the practical applicability of this approach. The framework early detected the faults of photocouplers in RPS and accurately diagnosed their conditions. It is expected that the proposed framework can enable condition-based maintenance, reduce unnecessary inspections, and improve system availability.