스마트 제조의 초자율화: 딥러닝 기반 고장 예측 및 최적 후보군 탐색
본문
- Conference
- 대한기계학회 CAE 및 응용역학부문 2025년 춘계학술대회
- Date
- 2025-04-17
- Presentation Type
- 구두
Abstract
The casting process is a manufacturing method where molten metal is poured into a mold and solidified, making it
ideal for mass production with precise shape formation. However, traditional casting relies heavily on operator experience,
leading to inconsistent quality. While deep learning-based defect prediction has gained attention, the black box nature of
AI models limits interpretability, reducing operator trust. To address this, we propose a deep learning-based Multi-Role
AI Model that integrates three key functions: Pass/Fail prediction, defect cause analysis using Explainable AI
(XAI), and optimal process variable exploration. Experimental results show 98.7% accuracy in defect prediction,
effective identification of defect causes, and optimized process parameters that minimize defects and reduce costs. This
study enhances fault prediction reliability and process efficiency in casting through AI-driven optimization.
- 이전글노이즈에 강건한 베어링 고장진단 모델 개발을 위한 규제 기법 25.04.22
- 다음글데이터 드리프트 관리를 위한 객체 탐지 기반 베어링 상태 추정 프레 임워크 25.04.22