http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
딥러닝 기반의 최대응력과 위치 예측 기법: 로드 휠 충격 테스트 예시
진아현(Ah-hyeon Jin),이성희(Sunghee Lee),유소영(Soyoung Yoo),신승연(Seungyeon Shin),김창곤(ChangGon Kim),허성필(Sungpil Heo),강남우(Namwoo Kang) 대한기계학회 2021 대한기계학회 춘추학술대회 Vol.2021 No.8
The impact test of road wheels is one of the important tests to ensure the safety of the wheels. Automakers use CAE simulation to analyze the location and magnitude of maximum stress, reducing prototype testing costs and time. However, the time required for modeling and analysis is still large, so it is difficult to quickly evaluate a large amount of conceptual design. In addition, it is difficult for general engineering designers to utilize because it requires high expertise in CAE. This study develops an AI-based wheel performance evaluation process that uses deep learning with CAE data to learn the location and magnitude of maximum stress. This deep learning model can predict the strength performance of road wheels in real time, allowing rapid evaluation at the conceptual design stage without domain knowledge.
좌민영(Minyoung Jwa),진아현(Ah-hyeon Jin),신동주(Dongju Shin),신승연(Seungyeon Shin),강남우(Namwoo Kang) 대한기계학회 2021 대한기계학회 춘추학술대회 Vol.2021 No.8
Deep learning-based inverse design is an design optimization methodology that solves the design performance prediction problem inversely using deep learning. While traditional optimization methods find the optimum in an iterative way, deep learning-based inverse design can generate the optimum immediately using a given target. However, studies that analyze the pros and cons of both methodologies in depth are limited. In particular, there is not enough research analyzing the types of optimization problems in which inverse design outperforms conventional optimization. This research compares the optimization performance of conventional optimization methods with deep learning-based inverse design using a benchmark function. The numerical results can provide guidance and insight for using deep learning-based inverse design in real design problems.