在淨零轉型與基礎設施減碳需求下，若能於設計早期掌握橋梁與道路鋪面之蘊含碳（embodied carbon），即可於方案仍具調整彈性時進行低碳比較。然而現有方法存在四項限制：傳統碳足跡盤查須待工程資訊完整後才能執行；參數化資料高度相關，隨機切分易高估模型泛化能力；非線性模型缺乏可追溯性；橋梁方案未經工程可行性篩選即依碳排排序，可能誤選不合格方案。為解決上述問題，本研究建立設計早期適用之預力混凝土（PC）I 型梁橋上部結構與市區道路鋪面蘊含碳代理模型框架，整合工程可行性篩選、群組隔離驗證與可解釋性人工智慧（XAI）分析。橋梁資料源自林祖瑀（2025），36,864 組理論方案經規則式結構檢核後取 4,611 組可行方案；鋪面資料源自王怡涵（2025），計 521 筆。橋梁建立 5、9、20 參數模型，鋪面建立 6、8 參數模型，並比較單一深度神經網路（DNN）與 5、10 個 ANN 集成模型；資料以工程群組切分訓練、驗證、測試集，集成模型透過群組感知折外預測（OOF）建立 Ridge 後設模型，並以 SHAP 分析特徵貢獻。結果顯示，橋梁 5、9 參數模型測試集 R²／RMSE 分別為 0.9651／7.934 與 0.9858／5.055 tCO₂e，20 參數模型（獨立切分）達 0.9962／2.767 tCO₂e；鋪面 6、8 參數模型均以單一 DNN 表現最佳，R² 為 0.9859 與 0.9999，RMSE 為 23.877 與 2.401 tCO₂e（5,000 次配對群組 bootstrap 之 RMSE 差值 95% CI：5.933–32.284 tCO₂e）。增加具工程意義之設計資訊可降低預測誤差，集成效益則依資料特性而異。5 參數模型適用於概念設計快速篩選，9 參數模型適用於基本設計比較，20 參數模型可作細部設計參考，方案仍以規則式結構檢核為最終可行性依據。

Embodied-carbon assessment often requires detailed design information, limiting its application during early design. Moreover, record-wise random splitting of parameterized engineering data may overestimate model generalization. This study develops A1–A3 embodied-carbon surrogate models for prestressed concrete I-girder bridge superstructures and urban pavements by integrating engineering feasibility screening, group-isolated validation, and explainable artificial intelligence (XAI).
The bridge dataset from Lin (2025) contained 36,864 theoretical alternatives, with 4,611 feasible alternatives retained after structural checks, while the pavement dataset from Wang (2025) included 521 modelling records. Bridge models were developed using 5, 9, and 20 parameters, and pavement models using 6 and 8 parameters. A single deep artificial neural network (DNN) was compared with Ridge stacking ensembles of five and ten neural networks under group-isolated data splitting. SHAP analysis was applied to interpret feature contributions.
The selected bridge models achieved test R²/RMSE values of 0.9651/7.934 tCO₂e, 0.9858/5.055 tCO₂e, and 0.9962/2.767 tCO₂e for the 5-, 9-, and 20-parameter models, respectively. Pavement models achieved R²/RMSE values of 0.9859/23.877 tCO₂e and 0.9999/2.401 tCO₂e for the 6- and 8-parameter models, respectively. Bootstrap analysis further confirmed the benefit of additional engineering information. Results indicate that richer design information improves prediction accuracy, whereas larger ensembles do not always enhance generalization. The proposed framework provides a traceable and interpretable approach for early-stage embodied-carbon screening while maintaining engineering feasibility.

AASHTO. (2008). Mechanistic-empirical pavement design guide: A manual of practice (Interim ed.). American Association of State Highway and Transportation Officials.
Adeli, H. (2001). Neural networks in civil engineering: 1989–2000. Computer-Aided Civil and Infrastructure Engineering, 16(2), 126–142. doi: 10.1111/0885-9507.00219
AzariJafari, H., Yahia, A., & Ben Amor, M. (2016). Life cycle assessment of pavements: Re-viewing research challenges and opportunities. Journal of Cleaner Production, 112, 2187–2197. doi: 10.1016/j.jclepro.2015.09.080
British Standards Institution. (2011). PAS 2050:2011: Specification for the assessment of the life cycle greenhouse gas emissions of goods and services. doi: 10.3403/30227173
British Standards Institution. (2023). PAS 2080:2023: Carbon management in buildings and infrastructure. doi: 10.3403/30438881
California Department of Transportation. (2022). Bridge design practice: Chapter 5.3, precast pretensioned concrete I-girders. California Department of Transportation.
Cawley, G. C., & Talbot, N. L. C. (2010). On over-fitting in model selection and subsequent selection bias in performance evaluation. Journal of Machine Learning Research, 11, 2079–2107.
Ceylan, H., Bayrak, M. B., & Gopalakrishnan, K. (2014). Neural networks applications in pave-ment engineering: A recent survey. International Journal of Pavement Research and Tech-nology, 7(6), 434–444. doi: 10.6135/ijprt.org.tw/2014.7(6).434
Chen, T., & Guestrin, C. (2016). XGBoost: A scalable tree boosting system. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Min-ing (pp. 785–794). Association for Computing Machinery. doi: 10.1145/2939672.2939785
Cheng, H.-T., Koc, L., Harmsen, J., Shaked, T., Chandra, T., Aradhye, H., Anderson, G., Cor-rado, G., Chai, W., Ispir, M., Anil, R., Haque, Z., Hong, L., Jain, V., Liu, X., & Shah, H. (2016). Wide & deep learning for recommender systems. In Proceedings of the 1st Work-shop on Deep Learning for Recommender Systems (pp. 7–10). Association for Computing Machinery. doi: 10.1145/2988450.2988454
Chou, J.-S., & Pham, A.-D. (2013). Enhanced artificial intelligence for ensemble approach to predicting high performance concrete compressive strength. Construction and Building Ma-terials, 49, 554–563. doi: 10.1016/j.conbuildmat.2013.08.078
Chou, J.-S., & Pham, A.-D. (2015). Smart artificial firefly colony algorithm-based support vector regression for enhanced forecasting in civil engineering. Computer-Aided Civil and Infra-structure Engineering, 30(9), 715–732. doi: 10.1111/mice.12121
Chou, J.-S., Ngo, N.-T., & Pham, A.-D. (2016). Shear strength prediction in reinforced concrete deep beams using nature-inspired metaheuristic support vector regression. Journal of Com-puting in Civil Engineering, 30(1), 04015002. doi: 10.1061/(ASCE)CP.1943-5487.0000466
Dietterich, T. G. (2000). Ensemble methods in machine learning. In J. Kittler & F. Roli (Eds.), Multiple classifier systems: First international workshop, MCS 2000 (Lecture Notes in Computer Science, Vol. 1857, pp. 1–15). Springer. doi: 10.1007/3-540-45014-9_1
Du, G., & Karoumi, R. (2013). Life cycle assessment of a railway bridge: Comparison of two superstructure designs. Structure and Infrastructure Engineering, 9(11), 1149–1160. doi: 10.1080/15732479.2012.670250
Du, G., & Karoumi, R. (2014). Life cycle assessment framework for railway bridges: Literature survey and critical issues. Structure and Infrastructure Engineering, 10(3), 277–294. doi: 10.1080/15732479.2012.749289
Du, G., Safi, M., Pettersson, L., & Karoumi, R. (2014). Life cycle assessment as a decision sup-port tool for bridge procurement: Environmental impact comparison among five bridge de-signs. The International Journal of Life Cycle Assessment, 19(12), 1948–1964. doi: 10.1007/s11367-014-0797-z
European Committee for Standardization. (2011). EN 15978:2011: Sustainability of construction works—Assessment of environmental performance of buildings—Calculation method.
European Committee for Standardization. (2019). EN 15804:2012+A2:2019: Sustainability of construction works—Environmental product declarations—Core rules for the product cate-gory of construction products.
European Committee for Standardization. (2022). EN 17472:2022: Sustainability of construction works—Sustainability assessment of civil engineering works—Calculation methods.
Feng, D.-C., Wang, W.-J., Mangalathu, S., & Taciroglu, E. (2021). Interpretable XGBoost-SHAP machine-learning model for shear strength prediction of squat RC walls. Journal of Structural Engineering, 147(11), 04021173. doi: 10.1061/(ASCE)ST.1943-541X.0003115
Finnveden, G., Hauschild, M. Z., Ekvall, T., Guinée, J., Heijungs, R., Hellweg, S., Koehler, A., Pennington, D., & Suh, S. (2009). Recent developments in life cycle assessment. Journal of Environmental Management, 91(1), 1–21. doi: 10.1016/j.jenvman.2009.06.018
Forrester, A. I. J., Sóbester, A., & Keane, A. J. (2008). Engineering design via surrogate model-ling: A practical guide. John Wiley & Sons. doi: 10.1002/9780470770801
García-Segura, T., & Yepes, V. (2016). Multiobjective optimization of post-tensioned concrete box-girder road bridges considering cost, CO2 emissions, and safety. Engineering Structures, 125, 325–336. doi: 10.1016/j.engstruct.2016.07.012
García-Segura, T., Yepes, V., & Frangopol, D. M. (2017). Multi-objective design of post-tensioned concrete road bridges using artificial neural networks. Structural and Multidiscipli-nary Optimization, 56(1), 139–150. doi: 10.1007/s00158-017-1653-0
García-Segura, T., Yepes, V., Frangopol, D. M., & Yang, D. Y. (2017). Lifetime reliability-based optimization of post-tensioned box-girder bridges. Engineering Structures, 145, 381–391. doi: 10.1016/j.engstruct.2017.05.013
Gong, H., Sun, Y., Shu, X., & Huang, B. (2018). Use of random forests regression for predicting IRI of asphalt pavements. Construction and Building Materials, 189, 890–897. doi: 10.1016/j.conbuildmat.2018.09.017
Hammervold, J., Reenaas, M., & Brattebø, H. (2013). Environmental life cycle assessment of bridges. Journal of Bridge Engineering, 18(2), 153–161. doi: 10.1061/(ASCE)BE.1943-5592.0000328
Hansen, L. K., & Salamon, P. (1990). Neural network ensembles. IEEE Transactions on Pattern Analysis and Machine Intelligence, 12(10), 993–1001. doi: 10.1109/34.58871
Harvey, J., Meijer, J., Ozer, H., Al-Qadi, I. L., Saboori, A., & Kendall, A. (2016). Pavement life-cycle assessment framework (Report No. FHWA-HIF-16-014). Federal Highway Admin-istration.
Hornik, K., Stinchcombe, M., & White, H. (1989). Multilayer feedforward networks are universal approximators. Neural Networks, 2(5), 359–366. doi: 10.1016/0893-6080(89)90020-8
Idowu, O. S., & Lam, K. C. (2020). Conceptual quantities estimation using bootstrapped support vector regression models. Journal of Construction Engineering and Management, 146(4), 04020018. doi: 10.1061/(ASCE)CO.1943-7862.0001780
Institution of Civil Engineers. (2023). Guidance document for PAS 2080: Practical actions and examples to accelerate the decarbonisation of buildings and infrastructure.
International Organization for Standardization. (2006a). ISO 14040:2006: Environmental man-agement—Life cycle assessment—Principles and framework.
International Organization for Standardization. (2006b). ISO 14044:2006: Environmental man-agement—Life cycle assessment—Requirements and guidelines.
International Organization for Standardization. (2018). ISO 14067:2018: Greenhouse gases—Carbon footprint of products—Requirements and guidelines for quantification.
Kaufman, S., Rosset, S., Perlich, C., & Stitelman, O. (2012). Leakage in data mining: Formula-tion, detection, and avoidance. ACM Transactions on Knowledge Discovery from Data, 6(4), Article 15. doi: 10.1145/2382577.2382579
Khademi, F., Akbari, M., Jamal, S. M., & Nikoo, M. (2017). Multiple linear regression, artificial neural network, and fuzzy logic prediction of 28 days compressive strength of concrete. Frontiers of Structural and Civil Engineering, 11(1), 90–99. doi: 10.1007/s11709-016-0363-9
Lundberg, S. M., & Lee, S.-I. (2017). A unified approach to interpreting model predictions. In I. Guyon et al. (Eds.), Advances in neural information processing systems (Vol. 30, pp. 4765–4774). Curran Associates, Inc.
Luo, L., Chen, Y., Feng, W., Wu, J., & Cai, W. (2025). Progressive prediction of embodied car-bon emissions across stages of schematic design with machine learning. Journal of Cleaner Production, 517, 145799. doi: 10.1016/j.jclepro.2025.145799
Mangalathu, S., Hwang, S.-H., & Jeon, J.-S. (2020). Failure mode and effects analysis of RC members based on machine-learning-based SHapley Additive exPlanations (SHAP) ap-proach. Engineering Structures, 219, 110927. doi: 10.1016/j.engstruct.2020.110927
Pham, A.-D., Hoang, N.-D., & Nguyen, Q.-T. (2016). Predicting compressive strength of high-performance concrete using metaheuristic-optimized least squares support vector regression. Journal of Computing in Civil Engineering, 30(3), 06015002. doi: 10.1061/(ASCE)CP.1943-5487.0000506
Rabinowicz, A., & Rosset, S. (2022). Cross-validation for correlated data. Journal of the Ameri-can Statistical Association, 117(538), 718–731. doi: 10.1080/01621459.2020.1801451
Santero, N. J., Masanet, E., & Horvath, A. (2011). Life-cycle assessment of pavements. Part I: Critical review. Resources, Conservation and Recycling, 55(9–10), 801–809. doi: 10.1016/j.resconrec.2011.03.010
United Nations Environment Programme. (2025). Global status report for buildings and con-struction 2024/2025: Not just another brick in the wall—The solutions exist. Scaling them will build on progress and cut emissions fast. United Nations Environment Programme.
United Nations Office for Project Services, United Nations Environment Programme, & Universi-ty of Oxford. (2021). Infrastructure for climate action. United Nations Office for Project Services.
Varma, S., & Simon, R. (2006). Bias in error estimation when using cross-validation for model selection. BMC Bioinformatics, 7, 91. doi: 10.1186/1471-2105-7-91
Wang, T., Lee, I. S., Harvey, J., Kendall, A., Lee, E. B., & Kim, C. (2012). UCPRC life cycle assessment methodology and initial case studies on energy consumption and GHG emis-sions for pavement preservation treatments with different rolling resistance (UCPRC-RR-2012-02). University of California Pavement Research Center.
Wolpert, D. H. (1992). Stacked generalization. Neural Networks, 5(2), 241–259. doi: 10.1016/S0893-6080(05)80023-1
Zhou, Z.-H. (2012). Ensemble methods: Foundations and algorithms. Chapman & Hall/CRC. doi: 10.1201/b12207
王怡涵（2025）。台灣市區道路低碳鋪面設計策略：基於力學經驗之鋪面設計與減碳策略分析［碩士論文，國立成功大學］。
交通部（2009）。交通技術標準規範公路類公路工程部—公路橋梁設計規範。
李牧軒（2022）。應用深度學習輔助RC建築之柱斷面設計［碩士論文，國立臺灣大學］。
李育帆（2016）。利用倒傳遞類神經網路預測透水混凝土抗壓強度及透水係數之研究［碩士論文，逢甲大學］。
李毅家（2021）。以圖像辨識深度學習進行即時淹水預測［碩士論文，國立臺灣大學］。
沈錦鴻（2013）。應用類神經網路配合ACI規範輔助卜作嵐混凝土配比設計［碩士論文，國立交通大學］。
林祖瑀（2025）。邁向淨零之橋梁工程：預力混凝土I型梁橋上部結構碳足跡模型與低碳設計探討［碩士論文，國立成功大學］。
黃廷堅（2010）。以類神經網路預估建築工程造價之研究［碩士論文，國立臺灣大學］。
楊敬緯（2022）。基於複合生命週期產品碳足跡分析建立營建產品碳足跡係數資料庫：以道路工程為例［碩士論文，國立成功大學］。
劉庭瑋（2011）。預力I型梁橋下部結構不同時之二氧化碳排放量評估［碩士論文，逢甲大學］。
環境部（2023）。氣候變遷因應法。全國法規資料庫。
魏維男、李志隆（2012）。公路總局預力混凝土簡支梁橋橋梁工程標準圖適用性之探討。臺灣公路工程，38（4–5），2–34。
羅紹文（2010）。預力I型梁橋之二氧化碳排放量評估［碩士論文，逢甲大學］。