³í¹®¸í |
±â°èÇнÀ ±â¹Ý ¼³°è ±â¹ýÀ» È°¿ëÇÑ °ÝÀÚ ±¸Á¶ ¿Àüµµµµ ÃÖÀû¼³°è / Optimal Lattice Structure Thermal Conductivity Design using Machine Learning-based Design Optimization |
¼ö·Ï»çÇ× |
Çѱ¹Àü»ê±¸Á¶°øÇÐȸ³í¹®Áý, Vol.37 No.5 (2024-10) |
ÆäÀÌÁö |
½ÃÀÛÆäÀÌÁö(353) ÃÑÆäÀÌÁö(7) |
ÁÖÁ¦¾î |
°ÝÀÚ ±¸Á¶; ÃÖÀû ¼³°è; ¿Àüµµµµ; ±â°èÇнÀ; À¯Àü ¾Ë°í¸®Áò ; lattice structure; optimal design; thermal conductivity; machine learning; genetic algorithm; neural network |
¿ä¾à1 |
º» ³í¹®¿¡¼´Â 15Â÷ bezier °î¼±À» »ç¿ëÇÏ¿© ±âÁ¸ÀÇ ¿¬±¸º¸´Ù ´õ À¯¿¬ÇÑ ºö Çü»óÀ» ¼³°èÇÏ°í, ´õ ³ÐÀº ¼³°è °ø°£¿¡¼ ÃÖÀû ¼³°è¸¦ ¼öÇàÇÏ¿© ÃÖÀûÀÇ ¿Àüµµµµ¸¦ °®´Â ºö Çü»óÀ» ¼³°èÇÑ´Ù. ¼³°è °ø°£ÀÌ ³Ð¾îÁö¸é ±×¸¸Å °è»ê¾çÀÌ Áõ°¡ÇÏ°Ô µÇ´Âµ¥, °íÂ÷¿ø º¯¼ö °ø°£¿¡¼ È¿À²ÀûÀ¸·Î ÀÛµ¿ÇÏ´Â Àΰø½Å°æ¸ÁÀ» »ç¿ëÇÏ¿© ÃÖÀû ¼³°è¸¦ °¡¼ÓÈÇÏ¿© °è»ê ÇѰ踦 ±Øº¹ÇÏ¿´´Ù. ´õ ³ª¾Æ°¡ ÃÖÀûÀÇ Åº¼º°è¼ö¸¦ °®´Â ºöÀÇ Çü»ó°ú ºñ±³ÇÏ¿´À¸¸ç ¿Àüµµ¿Í ź¼ºÇÐ »çÀÌÀÇ ¼öÇÐÀû À¯»ç¼ºÀ» ÀÌ¿ëÇÏ¿© ºö Çü»óÀ» ¼³¸íÇÑ´Ù. º» ¿¬±¸¿¡¼´Â ÀΰøÁö´ÉÀ» È°¿ëÇÑ Çü»ó ÃÖÀû¼³°è¸¦ ÅëÇØ ±âÁ¸ÀÇ ÇѰ踦 ¶Ù¾î³Ñ´Â °ÝÀÚ±¸Á¶ÀÇ ºö Çü»óÀ» Á¦¾ÈÇÑ´Ù. ¸ÕÀú, SC(Simple Cubic), BC(Body Centered Cubic) °ÝÀÚ ±¸Á¶ ºö Çü»óÀ» bezier °î¼±À¸·Î ¸ðµ¨¸µÇÏ°í bezier °î¼±ÀÇ Á¦¾îÁ¡ ÁÂÇ¥¸¦ ¹«ÀÛÀ§·Î ¼³Á¤ÇÏ¿© ÇнÀµ¥ÀÌÅ͸¦ È®º¸ÇÏ¿´´Ù. NN(Neural Network) ¹× GA(Genetic Algorithm)¸¦ ÅëÇØ ¿ì¼öÇÑ À¯È¿ ¿Àüµµµµ¸¦ °¡Áø ºö Çü»óÀ» »ý¼ºÇÏ¿© ÃÖÀûÀÇ ºö Çü»óÀ» ¼³°èÇÏ¿´´Ù. º» ¿¬±¸¸¦ ÅëÇØ ÃßÈÄ ´Ù¾çÇÑ ¿ Á¶°Ç¿¡¼ °ÝÀÚ±¸Á¶ÀÇ ÀûÀýÇÑ ±¸Á¶Àû ÇØ´äÀ» Á¦½ÃÇÒ ¼ö ÀÖÀ» °ÍÀ¸·Î ±â´ëµÈ´Ù. |
¿ä¾à2 |
Lattice structures exhibit good thermal performance due to the high surface-to-volume ratio. Previous studies have investigated the thermal conductivity to improve the performance of lattice structures. However, the conventional approach simplifies the geometry of lattice structures using limited design parameters due to the high computational or experimental costs. This study introduces a lattice structure with optimal thermal conductivity. We propose a lattice beam shape that overcomes the existing design limitations through shape optimization using artificial intelligence. First, the beam shape of the body-centered (BC) lattice structure is modeled as a smooth Bezier curve. Second, the coordinates of the control points of the Bezier curve are randomly set to obtain training data. Finally, the optimal beam shape is designed by generating a beam shape with excellent effective thermal conductivity through a neural network combined with a genetic algorithm. A mechanism of optimized thermal conductivity is suggested and the optimal beam shape is compared with a lattice structure with optimal elastic stiffness. The results of this study are expected to provide an appropriate structural solution for lattice structures under various thermal conditions in the future. |