화학공학소재연구정보센터
학회 한국재료학회
학술대회 2019년 가을 (10/30 ~ 11/01, 삼척 쏠비치 호텔&리조트)
권호 25권 2호
발표분야 특별심포지엄2. 재료공학에 적용 가능한 인공지능 기술 심포지엄-오거나이저:이승철(포항공대)
제목 AI 기반 분말 압축 공정: 구성방정식과 인공 신경망 알고리즘의 결합 Analysis of cold compaction based on constitutive relation and artificial neural networks
초록 Powder metallurgy (P/M) is one of the key manufacturing processes to produce near-net shaped parts. Generally, P/M process consists of two steps, cold compaction, and sintering. In P/M process, not only sintering but also cold compaction are the important steps, and defects-free compact should be required to avoid the subsequent sintered part defects such as cracking, distortion, and low sintered density. Especially, the compaction behavior is strongly dependent on the powder design parameters such as particle size characteristics and additive contents. They influence the overall densification mechanism in the compaction process, and consequentially yielded different compaction properties. As a way of understanding the relationship between the powder design parameters and final compaction properties, the integrated system was developed by combining the constitutive relation, artificial neural network (ANN), and compaction simulation. The effect of five major powder design parameters, particle size, graphite, lubricant, particle size distribution, and copper, were investigated in Fe based compaction with the concept of materials informatics. Firstly, the constitutive relation of compaction was studied with the powder design parameters using Shima-Oyane constitutive model. Calculated material coefficients in Shima-Oyane model were defined as the material related properties (ρTap, γ, a, b, n), which were the interim parameters between the powder design parameters and final properties. Secondly, a new ANN model was developed to predict the connection pattern between the powder design parameters and material related properties. Finally, the final properties (ρgreen, Δρ, σeffect, σhydro, εeffect, εvol) were calculated with compaction simulation by employing the material related properties as the input condition. The sensitivity analysis was also carried out among three constituents, powder design parameters, material related properties, and final properties.
저자 신다슬, 이치헌, 김석현, 박성진, 이승철
소속 포항공과대
키워드 Powder metallurgy; Cold compaction; Fe-C; Fe-C-Cu; Artificial neural networks
E-Mail