Earticle

초록

영어

The rapid expansion of electric vehicles and energy storage systems has led to an unprecedented accumulation of end-of-life lithium-ion batteries (LIBs), creating both environmental challenges and safety hazards. This study proposes an AI-based risk estimation framework that integrates real-time sensor data and machine learning analytics to enhance the safety and efficiency of LIB recycling processes. The framework comprises four key layers—data acquisition, preprocessing, AI risk estimation, and decision support—and applies a hybrid Deep Neural Network (DNN) and Gradient Boosted Tree (GBT) model to predict potential hazards across mechanical dismantling, thermal pretreatment, and hydrometallurgical recovery stages. A pilot-scale experimental setup was constructed, collecting over 10,000 time-series samples of temperature, gas emission rate, pH, pressure, and current under varying recycling conditions. The proposed model achieved 98% accuracy, 0.97 AUC, and a 45% reduction in mean time-to-alarm (MTTA) compared with rule-based monitoring systems, providing proactive early warnings 2–3 minutes before hazardous events. Explainability analysis using SHAP values identified temperature deviation (ΔT) and gas emission rate (G) as the dominant contributors to risk prediction, jointly explaining over 60% of model variance. The system maintained stable performance under noisy and incomplete data, confirming its applicability for real-time industrial deployment. Overall, this work demonstrates that AIdriven predictive intelligence can significantly improve safety, sustainability, and operational reliability in LIB recycling, supporting the transition toward smart, circular, and Industry 5.0–aligned recycling systems.

목차

Abstract
1. Introduction
2. Overview of Battery Recycling Processes
2.1 Mechanical Dismantling
2.2 Thermal Pretreatment
2.3 Hydrometallurgical Recovery
2.4 Process Data and Risk Factors
3. AI-Based Risk Estimation Framework
3.1 System Architecture
3.2 State Prediction Platform (SPP)
3.3 Model Formulation
3.4 Training and Optimization
3.5 Explainability and Safety Feedback
4. Experimental Setup and Validation
4.1 Experimental Configuration
4.2 Dataset Description
4.3 Training and Validation Methodology
4.4 Evaluation Metrics
4.5 Baseline Comparison
4.6 Validation Results and Observations
5. Results and Discussion
5.1 Model Performance Evaluation
5.2 Early Hazard Detection Performance
5.3 Feature Importance and SHAP Interpretability
5.4 Robustness and Industrial Applicability
6. Conclusions
Acknowledgement
References

키워드

저자

• Seongsoo Cho [ Professor, Department of Applied Artificial Intelligence, Hansung University, Seoul 02876, Republic of Korea ]
• Hiedo kim [ Director of R&D Department Chief Technical Officer, SUNGSAM Co., Ltd., Gyeonggi-do, Republic of Korea ] Corresponding Author

참고문헌

발행기관

간행물

표지보기 관심저널 등록

• 간행물명

  The International Journal of Advanced Smart Convergence

• 간기

  계간

• pISSN

  2288-2847

• eISSN

  2288-2855

• 수록기간

  2012~2025

• 십진분류

  KDC 326 DDC 380