미래형 직업교육을 위한 메타버스기반 환경 연구

이효섭(Hyo-Sub Lee),심호석(Ho-Seok Shim) 한국산업기술융합학회(구. 산업기술교육훈련학회) 2022 산업기술연구논문지 (JITR) Vol.27 No.4

코로나19 팬데믹의 장기화는 오프라인 중심의 교육을 온라인 방식으로 바꿔놓았다. 초창기 메타버스는 교육보다일회성 행사 중심의 이벤트로 운영되었지만, 현재는 현실과 상호작용하는 새로운 가상의 학습공간으로 주목받고 있다. 본 연구에서는 한국폴리텍대학에서 구축한 메타버스 플랫폼(Poly Meta)과 콘텐츠 활용 사례들을 분석하였다. 이를 기반으로 일반 교육적 의미의 메타버스 개념을 뛰어넘어, 직업교육에 필요한 메타버스의 환경요건과 메타버스 콘텐츠의 활용방안을 다루고자 한다. 향후 후속연구에서는 구축된 메타버스 플랫폼의 사용성과 접근성, 직업교육 훈련적용 효과에 대하여 정량적·정성적 분석을 수행할 것이다. The prolonged COVID-19 pandemic has changed offline-oriented education by increasingly employing online methods. Although in the early days, the metaverse was operated to be used as a one-time event rather than for education, it is now gaining attention as a new virtual learning space that interacts with reality. In this study, the metaverse platform Poly Meta and content use cases built by Korea Polytechnics were analyzed. Based on this, we intend to go beyond the usage of the metaverse in general education and further develop metaverse environmental requirements and content in a direction necessary for vocational education. In future research, quantitative and qualitative analysis will be performed on the metaverse platform’s usability and accessibility and on the effect of applying it to vocational education and training

환경친화적 토양개량제의 농경지 중 잔류농약 경감효과

이효섭(Hyo Sub Lee),홍수명(Su Myeong Hong),김택겸(Taek Kyum Kim),권혜영(Hye Young Kwon),김단비(Dan Bi Kim),문병철(Byeong Chul Moon),문준관(Joon Kwan Moon) 한국농약과학회 2016 농약과학회지 Vol.20 No.4

This study investigated the reduction of residual pesticide in soil by using eight kinds of soil amendments such as shell, quick lime, dolomite, silicic acid, leaf mold, oil cake, bed soil and husk. A total of ten pesticides most frequently detected in agricultural soil were selected (azoxystrobin etc.) and soil amendments were treated in soils at a 2% application rate and incubated for 7 days. Pesticides were analyzed by HPLC-MS/MS with QuEChERS-based sample preparation procedure and pH was measured on 1, 3 and 7 days. This study showed that, soil amemdments were effective in reducing pesticides after 7days of incubation. Over 90% of azoxystrobin, cadusafos, chlorpyrifos, fluquinconazole, imidacloprid, isoprothiolane and procymidone were decomposed from soils amended with 2% quick lime, whereas the concentration of boscalid, dimethomorph and triycyclazole were not decreased. The soil pH increased to 12.8-12.9 in soil amended with quick lime, but other soil conditioners did not change the soil pH. Quick lime was particularly effective in reducing residual pesticides.

챗GPT기반 메타버스플랫폼 모델 설계에 관한 연구

이효섭(Hyo-Sub Lee),심호석(Ho-Seok Shim) 한국산업기술융합학회(구. 산업기술교육훈련학회) 2023 산업기술연구논문지 (JITR) Vol.28 No.2

메타버스에서 상호작용은 무엇보다 중요하다. 아바타와의 상호작용과 공간에서의 공통된 경험을 통해 이루어지는 상호작용은 메타버스 사용자 모두에게 독특한 경험을 제공한다. 대화형 인공지능서비스인 챗GPT는 기존의 정적인 메타버스 내 상호작용을 획기적으로 개선할 수 있는 방법일 수 있다. 인공지능을 통해 아바타와 가상기술이 연결되어 인간의 언어를 이해하고 응답할 수 있게 되면서, 보다 몰입감있고 현실감이 느껴지는 메타버스 경험을 할 수 있기 때문이다. 따라서 본 연구에서는 챗GPT를 활용해 한국폴리텍대학의 메타버스플랫폼 가상캠퍼스에서 상호작용을 증강시킬 수 있는 모델을 설계하고자 한다. 향후연구에서는 한국폴리텍대학 고유의 도메인 데이터, NCS 및 산업용 데이터, 개인 데이터 등으로 LLM을 구축하고 충분히 학습한 파인튜닝 상세설계와 프레임웍을 추가하여 시스템으로 구현할 예정이다. 이를 폴리텍대학 메타버스 가상캠퍼스에 적용하면서 테스트와 모니터링을 진행하고, 기술적 측면에서의 전문가 자문과 객관적 평가 등을 적용한 정량적·정성적 분석을 수행하려고 한다. Interactivity is the most critical factor in the success of metaverse platforms, which can provide a unique user experience through interaction with digital avatars and engagement in a shared space. As a conversational artificial intelligence (AI) service, ChatGPT is considered a promising approach to dramatically improve the quality of interactions with computational agents on existing metaverse platforms. The ability of large language models (LLMs) to understand and respond to human communication can be expected to enable more immersive and realistic experiences. Therefore, in this study, we developed a computational model to enhance interaction with AI agents on a virtual campus on the Korea Polytechnics University metaverse platform by using ChatGPT. In future research, we plan to create an LLM with unique domain data related to Korea Polytechnic University and the National Competency Standards (NCS) along with industrial and personal data and implement the model by fine‑tuning a specially designed architecture with a pretrained framework. We plan to apply this system to the virtual Polytech Metaverse Campus for testing and monitoring. We will also perform quantitative and qualitative analyses based on advice from technical experts and the results of objective evaluations.

머신러닝기반 메타버스플랫폼 고도화 모델 설계에 관한 연구

이효섭(Hyo-Sub Lee),심호석(Ho-Seok Shim) 한국산업기술융합학회(구. 산업기술교육훈련학회) 2023 산업기술연구논문지 (JITR) Vol.28 No.4

메타버스는 실세계와 유사한 소통과 상호작용이 가능하며 다양한 콘텐츠와 서비스 제공을 통해 새로운 경험을 추구할 수 있는 공간이다. 근래 인공지능을 메타버스에 적용하려는 움직임으로 인해 메타버스 세계가 확장될 것이라는 기대가 커지고 있다. 메타버스와 인공지능의 융합은 커뮤니케이션의 한계를 극복하고 공간 생성의 어려움해소와 대내외 데이터의 활용 극대화에 핵심적인 역할 수행이 가능해지기 때문이다. 따라서 본 연구에서는 한국폴리텍대학 메타버스플랫폼 가상캠퍼스내에 구현한 기존 연구 현황을 공유하고, 이를 기반으로 최신 GPT기술과 LLM 등 머신러닝을 활용해 상호작용을 증강시키는 고도화 모델을 설계하였다. LLM기반의 대화형 내부정보 검색서비스 제공으로 대학 내 보유하고 있는 데이터를 효율적으로 활용하는 동시에 모든 의사소통을 자연어로 진행함 으로써 사용자 환경을 개선할 수 있도록 제안하였다. 향후연구에서는 범용서비스가 가능한 도메인 데이터 세부설계와 본 연구 모델을 구현하며 외부 LLM 모델과의 성능평가를 수행하려고 한다. Metaverse is a space where communication and interaction similar to those in the real world are possible and new experiences can be pursued through the provision of various contents and services. Owing to the recent advancements in the application of artificial intelligence (AI), Metaverse is expected to expand. The convergence of metaverse and AI overcomes the limitations of communication, resolves difficulties in creating space, and maximizes the use of internal and external data. In this study, we present the status of existing studies implemented within the Metaverse Platform virtual campus of Korea Polytechnic and the design of an advanced model that enhances interaction using machine learning such as the latest GPT technology and LLM. We propose to provide an LLM-based interactive internal information search service to improve the user environment by efficiently utilizing internal data and communicating in natural language. In future studies, we will focus on the detailed design of domain data capable of general-purpose services and implementation of the advanced model designed in this study. Thereafter, we intend to design and construct our own reliable LLM model and perform performance evaluation through with external LLM models.

프로그램 표절 검출 방법에 대한 조사

이효섭(Hyo-Sub Lee),도경구(Kyung-Goo Doh) 한국정보과학회 2005 한국정보과학회 학술발표논문집 Vol.32 No.1

표절이란 원본 자료의 출처를 밝히지 않고 사용하는 행위인데, 프로그램 표절은 표절로 의심되는 유사 소스코드를 말한다. 프로그램 표절을 검출하는 방법은 소스코드의 토큰 요약 후 비교하거나 단어와 키워드의 개수를 측정해 통계적으로 비교, 계산하거나 또는 프로그램의 구문구조를 비교하는 등 여러 접근 방법이 있다. 본 조사 보고서에서는 프로그램 표절 검출 방법 및 응용 도구에 대한 최근 연구 동향을 조사하여 정리하고 향후 발전 방향을 토론한다.

펜스형 태양광 구조시스템의 모듈배치별 CFD(Computational Fluid Dynamics) 유동해석

이효섭(Hyo-Sub Lee),유원조(Won-Jo Yoo) 대한전기학회 2022 대한전기학회 학술대회 논문집 Vol.2022 No.11

경엽 및 관주 처리에 따른 딸기 중 침투성 농약 잔류패턴 비교

이효섭 ( Hyo-sub Lee ),홍수명 ( Su-myeong Hong ),권혜영 ( Hye-young Kwon ),김단비 ( Dan-bi Kim ),문병철 ( Byeong-chul Moon ) 한국응용생명화학회 2017 Journal of Applied Biological Chemistry (J. Appl. Vol.60 No.4

Strawberries are one of the main commodities in Korea and have been exported over 30 million dollar. And the safety has recently become an important issue in the agro-trade, pesticide residue is most important matter of safe agricultural products. As strawberries can be cultivated using elevated production system, so it is possible to use the drenching. When spraying the pesticides, the use of drenching can be lower amount of residual than using foliar treatment. In this study, time sequential residual amount of pesticides were compared between drenching and foliar application in strawberries. Pesticides were treated with the standard and double dosage and samples were collected at 0, 3, 6, 9, 13, 16 and 21 day and analyzed by LC and LC-MS/MS after liquid-liquid extraction and QuEChERS. Two pesticides were detected less than 1 mg/kg in all samples for drenching treatment. Carbendazim residues in drenching treatment were 5 to 25 times lower than foliar application. And azoxystrobin was not detected in strawberries of drenching treatment.

태양광 구조물 지지용 나선형 파일 기초 시스템

이윤 ( Lee Yun ),이효섭 ( Lee Hyo Sub ),이상섭 ( Lee Sang Sub ) 한국구조물진단유지관리공학회 2014 한국구조물진단유지관리공학회 학술발표대회 논문집 Vol.18 No.2

녹색꽃양배추, 양상추, 상추 수확 후 토양 중 농약의 잔류량 조사

손경애(Kyeong-Ae Son),김찬섭(Chan Sub Kim),이효섭(Hyo Sub Lee),이은영(Eun Young Lee),이희동(Hee Dong Lee),박성은(Sung-eun Park),이지원(Ji-won Lee),홍수명(Su-Myeong Hong),조범행(Buemhaeng Cho),김윤한(Yun-han Kim) 한국농약과학회 2020 농약과학회지 Vol.24 No.4

This study was carried out to confirm the necessity of a residual test after PLS because some pesticide components remain in the soil and are absorbed into the succeeding crop, which may exceed 0.01 mg/kg on a default MRL. Broccoli, head lettuce and lettuce were cultivated after applying pesticide granules according to the standards for safe use of pesticides, and pesticide residues in the soil and crops at harvest period were investigated. On the 31st day after granulation treatment, lettuce was planted. The soil was drip irrigation during the test period. The residual amounts of cadusafos in broccoli and pencycuron in head lettuce were less than 0.01 mg/kg of the quantitative limit, and fluopyram in broccoli and flutolanil in head lettuce exceeded 0.01 mg/kg in the reference amount and doubling treatment. The residual amount in the soil was 0.2-3.58 mg/kg, which was more than 40% of the theoretical treatment concentration. When the lettuce was planted on the 2nd day after the hexaconazole granule treatment, the residual amount was <0.01 mg/kg in harvest. But when the plot was drip-irrigated for 1 month after treatment and lettuce was planted on the 31st day, hexaconazole was 0.02 mg/kg in harvest. Through this test, it was confirmed that in the case of granules that were released slowly after treatment in the soil, a significant amount of pesticide could remain in the soil even if the pesticide was not detected in the crop. And if sufficient water is supplied to the soil, it is likely to be transferred to the succeeding crop. Therefore, it is considered that the plant back intervals and the MRL for the rotational crops on pesticides with a high possibility of remaining in rotational crops should be established through the pesticide residue test for succeeding crop.

농산물 중 신규 등록 농약 pydiflumetofen의 잔류 분석을 위한 QuEChERS 방법 적용

노현호(Hyun Ho Noh),권혜영(Hyeyoung Kwon),김단비(Danbi Kim),김창조(Chang Jo Kim),이효섭(Hyo Sub Lee),경기성(Kee Sung Kyung) 한국농약과학회 2019 농약과학회지 Vol.23 No.2

This study was carried out to determine analytical method for the residual pesticide of new fungicide pydiflumetofen in agricultural products using QuEChERS method with LC-MS/MS. In order to optimize extraction method, recoveries of pydiflumetofen in agricultural products were calculated after extraction using acetate-buffering (AOAC Official Method 2007.01) and citrate-buffering (EN 15662) methods. Those were found to be 70-120%, efficiently, indicating that two methods were suitable to extract pydiflumetofen in agricultural products. Purification of samples was used by different sorbents compositions (25 mg PSA, 25 mg PSA+25 mg C<SUB>18</SUB>, 25 mg PSA+2.5 mg GCB, 50 mg PSA, 50 mg PSA+50 mg C<SUB>18</SUB> and 50 mg PSA+50 mg GCB). As results, recoveries ranged from 70 to 120% were found to be great under all sorbents. However, recovery of purified pydiflumetofen using d-SPE tube containing GCB of 50 mg was found to be low. That is because test pesticide, pydiflumetofen, was adsorbed to GCB removing pigment like a chlorophyll and non-polar hydrocarbon compound. The matrix effect has been calculated to survey the effect during ionization of the analyte in the MS detector. The matrix effects of the instrumental analysis process were mostly low, but showed a moderate matrix effect in some samples of brown rice and soybean. Finally, in order to analyze residual pesticide of pydiflumetofen in agricultural products, modified QuEChERS method with LC-MS/MS can be applied. However, it is suitable to quantify amount of pesticide using matrix matched standard calibration considering results of matrix effects.