營建業為社會經濟的主要推動力，帶來巨大的經濟利益，但也必須警惕職安意外可能導致的龐大經濟損失，以及這些事件對營建業所帶來的負面觀感之深遠影響。本研究通過文獻回顧和工程案例訪查，探討了施工職安風險評估方法的現狀。對這些文獻、案例資料進行整理和澄清後，進行了施工職安風險等級的評估。研究將進一步探討施工職安風險分析，運用建築資訊模型(BIM)來創建更直觀和易於理解的施工職安風險模型，幫助專案利害關係人基於研究中整理的施工職安風險資訊做出正確的決策。最後，本研究運用生成式預訓練轉換器(GPT)針對風險因素與相對應的風險緩解對策，進行微調讓專案利害關係人可以運用問答的方式，來獲取風險緩解策略的初步建議，以幫助專案更有效地因應潛在施工職安風險。

The construction industry serves as a major driving force in socio-economic
development, bringing significant economic benefits. However, it is essential to be vigilant about the substantial financial losses that occupational safety accidents can cause, as well as the profound negative perception these incidents can engender for the construction sector. This study, through a review of the literature and investigation of engineering case studies, examines the current state of occupational safety risk assessment methods in construction, after organizing and clarifying these documents and case study data, construction's occupational safety risk levels were evaluated. The research further explores construction occupational safety risk analysis, employing Building Information Modeling (BIM) to create more intuitive and understandable models of construction occupational safety risks. This assists Project stakeholders in making informed decisions based on the organized occupational safety risk information from the study. Finally, the study utilizes Generative Pre-trained Transformer (GPT) to fine-tune responses related to risk factors and corresponding risk mitigation strategies. This allows project
stakeholders to use a question-and-answer format to obtain preliminary
recommendations for risk mitigation strategies, aiding in more effectively
addressing potential construction occupational safety risks.

摘 要 i
ABSTRACT ii
表 目 錄 vi
圖 目 錄 vii
壹、緒論 1
一、研究背景 1
二、研究目的 3
三、研究範圍與限制 3
四、研究方法與流程 4
貳、文獻回顧 7
一、工程風險管理 7
(一) 國內工程風險管理相關文獻 7
(二) 國外工程風險管理相關文獻 12
二、風險可視化 15
三、生成式預訓練轉換器(GPT) 20
四、小結 23
參、研究方法 24
一、研究架構 24
二、案例風險評估 26
(一) 透過文獻彙整後確立風險評估方向 26
(二) 案例風險評級 26
三、風險可視化模型 31
(一) 風險模型建置流程 31
(二) Revit模型建立 31
(三) 建立Power BI的風險模型 33
(五) 風險等級的顏色分配 37
(六) 風險與風險顏色關聯 39
(七) 風險模型展現 40
四、運用生成式預訓練轉換器(GPT)產出風險緩解策略建議 43
(一) 風險緩解策略微調流程 43
(二) 資料集預處理 43
(三) 建置環境介紹以及引入OpenAI套件 45
(四) Open AI API微調Role-User-Content資料格式定義 46
(五) 分割資料集 47
(六) 微調設定 49
(七) 微調模型連接至Power BI 50
五、小結 57
肆、研究結果與分析 58
一、生成式預訓練轉換器(GPT)微調之成果 58
(一) 匯出微調成果 58
(二) 模型訓練損失、訓練準確率、驗證損失以及驗證準確率分析 59
(三) 不同指標之間的相關性熱力圖 60
(四) 標準化指標的箱形圖 62
二、風險數位儀表展示 64
三、小結 66
伍、結論與建議 67
一、結論 67
二、未來研究建議 68
參考文獻 69
附錄一、案例資料 74
附錄二、風險緩解對策資料 86

[1]勞動部職業安全衛生署 - 職業安全衛生署全球資訊網
https://www.osha.gov.tw/
[2]行政院研究發展考核委員會 - 行政院及所屬各機關風險管理及危機處理作業手冊 ,(2020)
https://www.ndc.gov.tw/Content_List.aspx?n=47CBA512BC0478E9
[3]El-Sayegh, S. M., and M. H. Mansour. (2015). Risk assessment allocation in highway construction projects in the UAE. J. Manage. Eng. 31 (6): 04015004.
https://doi.org/10.1061/(ASCE)ME.1943-5479.0000365.
[4]PMI (Project Management Institute). (2008). A guide to the project management body of knowledge, 4th Ed., Project Management Institute, Newton Square, PA.
[5]Jiahao Zeng, Min An, Nigel John Smith.(2007). Application of a fuzzy based decision making methodology to construction project risk assessment. International Journal of Project Management,25(6), p. 589-600.
https://doi.org/10.1016/j.ijproman.2007.02.006.
[6]Health and Safety Executive (HSE). Good practice and pitfalls in risk assessment. HSE Books: Research Report 151. Sudbury, Suffolk; 2003.
[7]Zeng J, An M, Chan AHC. (2015). Fuzzy reasoning decision making approach based multi-expert judgement for construction project risk analysis. In: Khosrowshahi F, editor. Proceedings of the twenty-first annual conference. Association of Researchers in Construction Management (ARCOM). London, UK. p. 841–52.
[8]A. Qazi, I. Dikmen. (2019). From risk matrices to risk networks in construction projects. IEEE Transactions on Engineering Management, p. 1-12.
https://doi.org/10.1108/EEMCS-03-2019-0059
https://doi.org/10.1205/095758203762851949
[9]勞動部職業安全衛生署, (2021), 營造工程風險評估技術指引. ROC.
[10]N.J. Duijm. (2015). Recommendations on the use and design of risk matrices. Safety Science, 76, p. 21-31.
https://doi.org/10.1016/j.ssci.2015.02.014
[11]X. Ruan, Z. Yin, D.M. Frangopol. (2015). Risk matrix integrating risk attitudes based on utility theory. Risk Analysis, 35 (8), p. 1437-1447.
https://doi.org/10.1111/risa.12400
[12]N.J. Duijm. (2015). Recommendations on the use and design of risk
matrices. Safety Science, 76, p. 21-31.
https://doi.org/10.1016/j.ssci.2015.02.014
[13]Abroon Qazi, Abdulrahim Shamayleh, Sameh El-Sayegh, Steven Formaneck. (2021). Prioritizing risks in sustainable construction projects using a risk matrix-based Monte Carlo Simulation approach. Sustainable Cities and Society, Volume 65.
https://doi.org/10.1016/j.scs.2020.102576.
[14]Roberta Selleck, Marcus Cattani, Maureen Hassall. (2022). Proposal for and validation of novel risk-based process to reduce the risk of construction site fatalities (Major Accident Prevention (MAP) program). Safety Science, Volume 158.
https://doi.org/10.1016/j.ssci.2022.105986
[15]Andrew Hopkins, (2010). Risk-management and rule-compliance: Decision-making in hazardous industries. Safety Science, Volume 49, Issue 2.
https://doi.org/10.1016/j.ssci.2010.07.014
[16]X. Xu, L. Ma, L. Ding. (2014). A Framework for BIM-Enabled Life-Cycle Information Management of Construction Project, Int. J. Adv. Rob. Syst. 11 126.
https://doi.org/10.5772/58445
[17]Soowon Chang, Ph.D., A.M.ASCE; Heung Jin Oh; JeeHee Lee, Ph.D.,
Aff.M.ASCEand Joe Perkins.(2023). Proof-of-Concept Study for Model-Based Construction Safety Diagnosis and Management Driven by Prevention through Design. American Society of Civil Engineers. https://doi.org/10.1061/JMENEA.MEENG-5474
[18]Ying Lu, Peizhen Gong, Yuchun Tang, Shuqi Sun, Qiming Li. (2021). BIM-integrated construction safety risk assessment at the design stage of building projects. Automation in Construction, Volume 124.
https://doi.org/10.1016/j.autcon.2021.103553
[19]Jingfeng Yuana, Xuewei Lib, Xiaer Xiahoub, Nicholas Tymviosc, Zhipeng Zhoud, Qiming Lib. (2019). Accident prevention through design (PtD): Integration of building information modeling and PtD knowledge base. Automation in Construction, Volume 102.
https://doi.org/10.1016/j.autcon.2019.02.015
[20]P. Smith. (2014). Bim implementation – global strategies. Procedia Engineering, vol. 85, p. 482 – 492.
https://doi.org/10.1016/j.proeng.2014.10.575
[21]J. Kunz, M. Fischer. (2020). Virtual design and construction, Constr. Manage. Econ. 38 p.355–363.
https://doi.org/10.1080/01446193.2020.1714068
[22]Y. Zou, A. Kiviniemi, S.W. Jones. (2017). A review of risk management through BIM and BIM-related technologies, Saf. Sci. 97 p.88-98.
https://doi.org/10.1016/j.ssci.2015.12.027
[23]S. Zhang, J. Teizer, J.-K. Lee, C.M. Eastman, M. Venugopal. (2013). Building information modeling (BIM) and safety: automatic safety checking of construction models and schedules. Autom. Constr., 29 , p. 183-195.
https://doi.org/10.1016/j.autcon.2012.05.006
[24]J.R. Lin, Z.Z. Hu, J.P. Zhang, F.Q. Yu. (2016). A natural-language-based approach to intelligent data retrieval and representation for cloud BIM. Computer-Aided Civil And.
https://doi.org/10.1111/mice.12151
[25]S. Durdyev, M. Ashour, S. Connelly, A. Mahdiyar. (2022). Barriers to the implementation of Building Information Modelling (BIM) for facility management, Journal of Building Engineering. 46 103736.
https://doi.org/10.1016/j.jobe.2021.103736
[26]D. Kelly, B. Ilozor. (2019). A quantitative study of the relationship between project performance and BIM use on commercial construction projects in the USA. Int. J. Constr. Educ. Res., 15 (1) , p. 3-18.
https://doi.org/10.1080/15578771.2016.1202355
[27]A. Tomek, P. Matějka. (2014). The impact of BIM on risk management as an argument for its implementation in a construction company. Procedia Eng., 85 , p. 501-509
https://doi.org/10.1016/j.proeng.2014.10.577
[28]El Kaed, C., Leida, B., & Gray, T. (2016, December). Building management insights driven by a multi-system semantic representation approach. In 2016 IEEE 3rd World Forum on Internet of Things, WF-IoT, p. 520-525. IEEE.
[29]Harode A, Ensafi M, Thabet W. (2022). Linking BIM to Power BI and HoloLens 2 to Support Facility Management: A Case Study Approach. Buildings. 12(6):852.
https://doi.org/10.3390/buildings12060852
[30]L. Floridi, M. Chiriatti. (2020). GPT-3: its nature, scope, limits, and consequences. Mind. Mach.
https://doi.org/10.1007/s11023-020-09548-1
[31]Spyros Kamnis. (2023). Generative pre-trained
transformers (GPT) for surface engineering. Surface and Coatings Technology,Volume 466.
https://doi.org/10.1016/j.surfcoat.2023.129680.
[32]Christopher D. Manning. (2022). Human Language Understanding & Reasoning. Daedalus, Volume 151, Issue 2, Spring 2022.
https://doi.org/10.1162/daed_a_01905
[33]Chaobo Zhang, Jie Lu, Yang Zhao. (2023). Generative
pre-trained transformers (GPT)-based automated data mining for building energy management: Advantages, limitations and the future,Energy and Built Environment,Volume 5, Issue 1, p.143-169.
https://doi.org/10.1016/j.enbenv.2023.06.005.
[34]A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L.
Jones, A.N. Gomez, Ł. Kaiser, I. Polosukhin. (2017). Attention is all you need. Adv. Neural Inf. Process.
https://doi.org/10.48550/arXiv.1706.03762
[35]van Eck, N. J., & Waltman, L. (2010). Software survey: VOSviewer, a computer
program for bibliometric mapping. Scientometrics, 84(2), p.523–538.
https://doi.org/10.1007/s11192-009-0146-3
[36]Blogic. Available online: https://www.bimservices.it/ (accessed on 20
September 2023)