°ÇÃ൵½Ã°ø°£¿¬±¸¼Ò

Architecture & Urban Research Institute

pdf¿ø¹®º¸±â ¿¡·¯ ÇØ°á¹æ¹ý ¹Ù·Î°¡±â



¹®ÇåȨ > ¿¬±¸³í¹® > »ó¼¼

[¿ø¹®º¸±â½Ã ¼ÒºñµÇ´Â Æ÷ÀÎÆ® : 100 Æ÷ÀÎÆ®] ¹Ì¸®º¸±â Àοë

Çѱ¹°Ç¼³°ü¸®ÇÐȸ|³í¹®Áý 2024³â 11¿ù

³í¹®¸í BIM ¹× OpenAI ±â¹Ý ±³·® ¹× ÅͳΠÃʱ⠰ø»ç°èȹ ¼ö¸³ Áö¿ø ¸ðµ¨ / BIM and OpenAI-based Model for Supporting Initial Construction Planning of Bridges and Tunnels
ÀúÀÚ¸í ȲÁرâ(Hwang, Jungi) ; ¼Û½ÂÈ£(Song, Seung Ho) ; ÀÌâ¼ö(Lee, Changsu) ; ¾ÈÈñÀç(Ahn, Heejae) ; Á¶ÈÆÈñ(Cho, Hunhee) ; °­°æÀÎ(Kang, Kyung-In)
¹ßÇà»ç Çѱ¹°Ç¼³°ü¸®ÇÐȸ
¼ö·Ï»çÇ× Çѱ¹°Ç¼³°ü¸®ÇÐȸ ³í¹®Áý, Vol.25 No.6 (2024-11)
ÆäÀÌÁö ½ÃÀÛÆäÀÌÁö(24) ÃÑÆäÀÌÁö(10)
ISSN 2005-6095
ÁÖÁ¦ºÐ·ù ½Ã°ø(Àû»ê)
ÁÖÁ¦¾î Ãʱ⠰ø»ç°èȹ; BIM; OpenAI; GPT-4.0 ; Construction Planning; BIM; OpenAI; GPT-4.0
¿ä¾à1 ÃÖ±Ù µðÁöÅÐ ±â¼úÀÇ ¹ßÀü¼Óµµ°¡ °æÁ¦ »ý»ê¼º¿¡ ¹ÌÄ¡´Â ¿µÇâÀÌ ±Þ¼Óµµ·Î Áõ°¡ÇÔ¿¡ µû¶ó °Ç¼³»ê¾÷ÀÇ »ý»ê¼º Çâ»óÀ» À§ÇØ µðÁöÅÐ Àüȯ ¹æ¾ÈÀÌ ¿¬±¸µÇ°í ÀÖ´Ù. ƯÈ÷, ÅëÇÕ °Ç¼³ Á¤º¸°ü¸® ¹× ÀÌÇØ°ü°èÀÚ °£ µðÁöÅÐ Çù¾÷ÀÇ °­È­¸¦ À§ÇØ BIM ±â¼úÀÌ ÁÖ¸ñ¹Þ°í ÀÖ´Ù. º» ¿¬±¸¿¡¼­´Â OpenAI(GPT-4.0) ¾ð¾î¸ðµ¨À» È°¿ëÇÏ¿© 3´Ü°è·Î ºÐ·ùµÈ Á¤º¸Á¦°ø ȯ°æ¿¡¼­ »êÃâÇÑ Ãʱ⠰Ǽ³Á¤º¸¸¦ ±â¼öÇàµÈ °Ç¼³ÇÁ·ÎÁ§Æ®ÀÇ °ø»çÁ¤º¸¿Í ºñ±³¡¤ºÐ¼®ÇÏ¿© Á¤º¸ÀÔ·Â ¼öÁØ¿¡ µû¶ó »êÃâµÇ´Â µ¥ÀÌÅÍÀÇ Á¤È®µµ¸¦ ºÐ¼®ÇÏ¿´´Ù. ÃÖ¼ÒÇÑÀÇ Á¤º¸¸¸ ÀÔ·ÂÇÏ´Â 1´Ü°è Á¤º¸Á¦°ø ȯ°æ¿¡¼­´Â Á¤·®ÀûÀÎ µ¥ÀÌÅ͸¦ »êÃâÇÒ ¼ö ¾ø¾ú´Ù. Ãʱ⠰ø»ç±âȹ Á¤º¸¸¦ ÀÔ·ÂÇÏ´Â 2´Ü°è Á¤º¸Á¦°ø ȯ°æ¿¡¼­´Â Á¤·®ÀûÀÎ µ¥ÀÌÅ͸¦ »êÃâÇÒ ¼ö ÀÖ¾úÀ¸³ª ±â¼öÇàµÈ ÇÁ·ÎÁ§Æ®¿Í 54.45%ÀÇ ¿ÀÂ÷À²ÀÌ È®ÀεǾú´Ù. ¸¶Áö¸·À¸·Î, ±¸Á¶¹°ÀÇ Çü»óÁ¤º¸¿Í Âø°ø ¹× ÁØ°øÀÏÀÚ¸¦ Ãß°¡·Î ÀÔ·ÂÇÏ´Â 3´Ü°è Á¤º¸Á¦°ø ȯ°æ¿¡¼­ »êÃâÇÑ µ¥ÀÌÅÍ´Â ±â¼öÇàµÈ ÇÁ·ÎÁ§Æ®¿Í Æò±Õ 19.70% ¿ÀÂ÷À²ÀÌ È®ÀεǾî GPT-4.0 ¾ð¾î¸ðµ¨·Î »êÃâÇÑ Ãʱ⠰ø»çÁ¤º¸ÀÇ È°¿ë °¡´É¼ºÀ» °ËÁõÇÏ¿´°í »êÃâÇÑ °ø»çÁ¤º¸¸¦ ±â¹ÝÀ¸·Î Á¤º¸ ºÐ·ù ½Ã½ºÅÛÀ» ÅëÇØ Ã³¸®ÇÏ¿© BIM¼³°è¸ðµ¨°ú ¿øÈ°ÇÏ°Ô ¿¬°áµÉ ¼ö ÀÖ´Â Á¤º¸°ü¸® ÇÁ·Î¼¼½º¸¦ ±¸ÃàÇÏ¿´´Ù. º» ¿¬±¸¿¡¼­ ±¸ÃàÇÑ Á¤º¸°ü¸® ÇÁ·Î¼¼½º´Â °øÁ¤°èȹ ÀÛ¼º ¹× ¼ö·®»êÃâ ¾÷¹«ÀÇ È¿À²È­¸¦ ÅëÇØ ±âº»¼³°è ´Ü°èÀÇ Á¦ÇÑµÈ ½Ã°£ ³» ÃÖÀû ¼³°è¾ÈÀ» µµÃâÇÏ¿© Àü¸é BIM Àû¿ë¿¡ ±â¿©ÇÑ´Ù.
¿ä¾à2 BIM is a prominent technology for enhancing integrated construction information management and extensive stakeholder coordination. However, the difficulty of delivering practical construction planning in the initial design phase and the lack of qualified BIM experts within the industry are the challenges of BIM-based construction digitization. This research utilizes GPT-4.0, a large language model developed by OpenAI that has gained significant traction across academia and industry for its ease of use and effectiveness. The construction information was categorized into three distinct detail levels to derive construction schedules for NATM and the Extradosed bridge. Derived schedules were compared and analyzed against the actual schedules for accuracy. Only qualitative data were generated with the lowest detail level. With the highest detail level, which included structural information and commencement and completion dates, the quantitative data extracted exhibited an average error rate of 19.70% compared to the actual schedules, thus validating the use of GPT-4.0 for extracting relevant construction schedules. The quantitative data were subsequently processed via a classification system, establishing an information management process to be integrated with BIM models. These results contribute to achieving a better construction plan under the limited time and resources in the early design phase and facilitate broader BIM implementation throughout all phases of construction projects.
¼ÒÀåó Çѱ¹°Ç¼³°ü¸®ÇÐȸ
¾ð¾î Çѱ¹¾î
DOI https://dx.doi.org/10.6106/KJCEM.2024.25.6.024