臺灣博碩士論文加值系統

混凝土為當今土木工程中不可或缺的材料，各式工程中，混凝土品質優劣 對於結構物的穩定有著舉足輕重的影響。於混凝土抗壓強度之品質評估上時常 使用非破壞檢測方式進行測量，以避免鑽心試驗對結構本體產生應力集中的情 況發生。經過去文獻中得知，透過非破壞檢測測量混凝土抗壓強度時有 15%~20%之誤差範圍，故本研究採用卷積神經網路模型進行混凝土抗壓強度之 預測，搜集某大樓之非破壞檢測(反彈錘試驗及超音波波速試驗)數據與鑽心 抗壓試驗數據，施測點位共 100 組，將非破壞數值轉換為圖片表示，但由於初 始之資料量不足以訓練預測模型，因此利用資料增強技術增加訓練樣本。根據 不同誤差範圍建立三種預測模型區間，透過圖像辨識方法辨識圖中特徵值，進 而推估其抗壓強度範圍，研究結果如下所示:
1. 模型經訓練完成後，利用測試樣本評估其模型準確度得知，三種模型之準 確率皆高於 80%以上，證明模型具有其效益。
2. 有別以往預測混凝土抗壓強度之方法，將預測結果以區間表示而並非數值， 透過模型之結果應證此方法可行。
3. 利用資料增強技術改善訓練樣本不足的情況，可降低訓練過程中過度擬合 的情況，有效提升模型訓練成效。
由上述可知，可利用卷積神經網路建立混凝土抗壓強度預測模型，以模型所預測之區間評估混凝土品質，以確保結構物之安全及穩定。

Concrete is a crucial material in the civil engineering industry. In various projects, the quality of concrete is a key factor influencing the structure stabilities. When assessing compressive strength of concrete, Nondestructive testing (NDT) is often used to avoid structure damage caused by the coring process. According to the literatures, using NDT method to estimate concrete compressive strength has an average error of 15% to 20%. Therefore, this research proposes a Convolutional Neural Networks (CNNs) model to estimate the compressive strength of concrete. A total of 100 sets of NDT (Rebound Hammer Test and Ultrasonic Plus Velocity Test) measurements were collected from the basement of a large residential complex. Because of the limited number of samples collected, this research incorporates Data Augmentation to increase the dataset. Three prediction model with different error margins were created. CNNs are used to develop the compressive strength prediction models, which classify the images according to the eigenvectors.The research results are summarized as follows:
1. The compressive strength estimation accuracies for all three models are higher than 80%, which proves that the models are effective.
2. The research results show that the error range prediction method is feasible.
3. Data Augmentation can be used to increase the dataset when there are limited samples available.

摘要
Abstract
誌謝
目錄
表目錄
圖目錄
第 1 章 緒論
1.1研究背景與動機
1.2研究目的
1.3研究範圍與限制
1.4研究步驟及流程
1.5章節架構
第 2 章 文獻回顧
2.1 非破壞檢測
2.1.1 反彈錘試驗(Rebound Hammer Test)
2.1.2 超音波波速試驗（Ultrasonic Pulse Velocity Test)
2.1.3 結合超音波與反彈錘的試驗法（SonReb）
2.2人工智慧 (Artificial Intelligence)
2.3 資料增強技術(Data Augmentation)
2.4小結
第 3 章 研究方法
3.1研究數據
3.1.1 反彈錘試驗
3.1.2超音波波速試驗
3.2預測模型開發工具
3.2.1 Python
3.2.2 Matplotlib
3.2.3 Keras
3.2.4 TensorFlow
3.2.5 Jupyter Notebook
3.3 研究模型建立
3.3.1預測模型之輸出項設計
3.3.2資料準備
3.3.3 建立卷積神究網路模型架構
3.3.4 預測結果之輸出值
3.4 小結
第 4 章 研究結果與分析
4.1預測模型相關資訊
4.1.2輸入項資訊
4.1.3研究工具資訊
4.1.4訓練次數
4.1.5 卷積神經網路架構
4.2研究結果
4.2.1 模型一結果
4.2.2 模型二結果
4.2.3 模型三結果
4.2.4 各區間之結果評估
4.2.5 資料增強前後結果比較
4.3 模型預測
4.3.1 預測結果
4.3.2 混淆矩陣
4.4 小結
第 5 章 結論與建議
5.1結論
5.2建議
參考文獻

一、英文文獻
[1]ASTM C597-02.,2002, Standard Test Method for Pulse Velocity Through Concrete.
[2]ASTM C805.,2008, Standard Test Method for Rebound Number of Hardened Concrete, Annual Book of ASTM Standards, Vol.04.02.
[3]Albawi, S., Mohammed, T. A., & Al-Zawi, S.,2017, “Understanding of a convolutional neural network”. In 2017 International Conference on Engineering and Technology , ICET, pp. 1-6.
[4]Almutairi, M., Nikitas, N., Abdeljaber, O., Avci, O., & Bocian, M. ,2021, “A methodological approach towards evaluating structural damage severity using 1D CNNs”. In Structures ,Vol. 34, pp. 4435-4446, Elsevier.
[5]Alwash, M., Breysse, D., Sbartaï, Z. M., Szilágyi, K., & Borosnyói, A.,2017,” Factors affecting the reliability of assessing the concrete strength by rebound hammer and cores”, Construction and Building Materials, 140, 354-363.
[6]Bogas, J. A., Gomes, M. G., & Gomes, A.,2013, “Compressive strength evaluation of structural lightweight concrete by non-destructive ultrasonic pulse velocity method”, Ultrasonics, 53(5), 962-972.
[7]Breccolotti, M., Bonfigli, M. F., & Materazzi, A. L.,2013, “Influence of carbonation depth on concrete strength evaluation carried out using the SonReb method. NDT & E International”, 59, 96-104.
[8]Carino, N. J., & Malhotra, V. M. (Eds.).,2004, “Handbook on nondestructive testing of concrete”. CRC press.
[9]Choi, Y., Yoon, G., & Kim, J.,2021, “Unsupervised learning algorithm for signal validation in emergency situations at nuclear power plants”, Nuclear Engineering and Technology.
[10]Cristofaro, M. T., Viti, S., & Tanganelli, M.,2020, “New predictive models to evaluate concrete compressive strength using the SonReb method”. Journal of Building Engineering, 27, 100962.
[11]Cui, F., Ning, M., Shen, J., & Shu, X. ,2022, “Automatic recognition and tracking of highway layer-interface using Faster R-CNN”. Journal of Applied Geophysics, 196, 104477.
[12]Deng, L., & Yu, D.,2014, “Deep learning: methods and applications.Foundations and trends in signal processing”, 7(3–4), 197-387.
[13]Du, B., Liu, Y., & Tian, G.,2020, “Improving Motor Imagery EEG Classification by CNN with Data Augmentation”. In 2020 IEEE 19th International Conference on Cognitive Informatics & Cognitive Computing (ICCI* CC) (pp. 111-118).
[14]Fawzi, A., Samulowitz, H., Turaga, D., & Frossard, P. ,2016, “Adaptive data augmentation for image classification”, In 2016 IEEE international conference on image processing (ICIP) (pp. 3688-3692).
[15]French, R. M.,2000, “The Turing Test: the first 50 years. Trends in cognitive sciences”, 4(3), 115-122.
[16]Ghosh, R., Sagar, S. P., Kumar, A., Gupta, S. K., & Kumar, S. ,2018, “Estimation of geopolymer concrete strength from ultrasonic pulse velocity (UPV) using high power pulser”. Journal of building engineering, 16, 39-44.
[17]H Kim, H Kim, YW Hong, H Byun,2018,” Detecting Construction Equipment Using a Region-Based Fully Convolutional Network and Transfer Learning”, Journal of Computing in Civil Engineering, Volume 32.Issue 2.
[18]Hoła, J., & Schabowicz, K.,2005, “New technique of nondestructive assessment of concrete strength using artificial intelligence”. NDT & E International, 38(4), 251-259.
[19]Jo, H., Na, Y. H., & Song, J. B.,2017, “ Data augmentation using synthesized images for object detection”, In 2017 17th International Conference on Control, Automation and Systems (ICCAS),pp. 1035-1038.
[20]Kazemi, M., Khodaiyan, F., & Hosseini, S. S.,2019, “Eggplant peel as a high potential source of high methylated pectin: Ultrasonic extraction optimization and characterization”, Lwt, 105, 182-189.
[21]Kewalramani, M. A., & Gupta, R.,2006, “Concrete compressive strength prediction using ultrasonic pulse velocity through artificial neural networks”, Automation in Construction, 15(3), 374-379.
[22]Kim, S., Lee, K., Lee, M., Lee, J., Ahn, T., & Lim, J. T.,2021, “Evaluation of saturation changes during gas hydrate dissociation core experiment using deep learning with data augmentation”, Journal of Petroleum Science and Engineering, 109820.
[23]Kumar, P., Sharma, A., & Kota, S. R.,2021, “Automatic Multiclass Instance Segmentation of Concrete Damage Using Deep Learning Model”. IEEE Access, 9, 90330-90345.
[24]Kusrini, K., Suputa, S., Setyanto, A., Agastya, I. M. A., Priantoro, H., Chandramouli, K., & Izquierdo, E.,2020, “Data augmentation for automated pest classification in Mango farms”, Computers and Electronics in Agriculture, 179, 105842.
[25]Learned-Miller, E. G.,2014, Introduction to supervised learning. I: Department of Computer Science, University of Massachusetts.
[26]Ligthart, A., Catal, C., & Tekinerdogan, B.,2021,” Analyzing the effectiveness of semi-supervised learning approaches for opinion spam classification”, Applied Soft Computing, 101, 107023.
[27]Mohammed, B. S., Adamu, M., & Liew, M. S.,2018,” Evaluating the effect of crumb rubber and nano silica on the properties of high volume fly ash roller compacted concrete pavement using non-destructive techniques”, Case studies in construction materials, 8, 380-391.
[28]Naik T. R. and Malhotra V. M.,1991, “The Ultrasonic Pulse Velocity Method in CRC Handbook on Nondestructive Testing of Concrete”, pp. 169-188, CRC Press, Boca Raton, Florida.
[29]Naskar, S., & Chakraborty, A. K. ,2016, “Effect of nano materials in geopolymer concrete”, Perspectives in science, 8, 273-275.
[30]Nobile, L.,2015, “Prediction of concrete compressive strength by combined non-destructive methods”. Meccanica,Springer, 50(2), 411-417.
[31]Nobile, L., & Bonagura, M.,2013,”Accuracy of non-destructive evaluation of concrete compression strength”, In The 12th International Conference of the Slovenian Society for Non-Destructive Testing, Portorož, Slovenia.
[32]Panedpojaman, P., & Tonnayopas, D.,2018, “Rebound hammer test to estimate compressive strength of heat exposed concrete”,Construction and Building Materials, 172, 387-395.
[33]Provost, F., & Kohavi, R.,1998,”Glossary of terms”, Journal of Machine Learning, 30(2-3), 271-274.
[34]S. Hochreiter, Y. Bengio, P. Frasconi, and J. Schmidhuber.,2001, “Gradient flow in recurrent nets: the difficulty of learning long-term dependencies”, In S. C. Kremer and J. F. Kolen, editors, A Field Guide to Dynamical Recurrent Neural Networks. IEEE Press.
[35]Sastrawan, I. K., Bayupati, I. P. A., & Arsa, D. M. S. ,2021, “Detection of fake news using deep learning CNN–RNN based methods”, ICT Express.
[36]Shariq, M., Prasad, J., & Masood, A.,2013, “Studies in ultrasonic pulse velocity of concrete containing GGBFS”, Construction and Building Materials, 40, 944-950.
[37]Silver, D., Huang, A., Maddison, C. J., Guez, A., Sifre, L., Van Den Driessche, G., ... & Hassabis, D,2016, “Mastering the game of Go with deep neural networks and tree search” ,Nature, 529(7587), 484-489.
[38]Trtnik, G., Kavčič, F., & Turk, G.,2009, “Prediction of concrete strength using ultrasonic pulse velocity and artificial neural networks”, Ultrasonics ,49(1), 53-60.
[39]Tsioulou, O., Lampropoulos, A., & Paschalis, S.,2017, “Combined non-destructive testing (NDT) method for the evaluation of the mechanical characteristics of ultra high performance fibre reinforced concrete (UHPFRC)”, Construction and Building Materials, 131, 66-77.
[40]Wang, F. Y., Zhang, J. J., Zheng, X., Wang, X., Yuan, Y., Dai, X., ... & Yang, L. ,2016, “Where does AlphaGo go: From church-turing thesis to AlphaGo thesis and beyond”, IEEE/CAA Journal of Automatica Sinica, 3(2), 113-120.
[41]Xiao, B., Zhang, Y., Chen, Y., & Yin, X. (2021). A semi-supervised learning detection method for vision-based monitoring of construction sites by integrating teacher-student networks and data augmentation. Advanced Engineering Informatics, 50, 101372.
[42]Yakimovich, A., Beaugnon, A., Huang, Y., & Ozkirimli, E.,2021, “Labels in a haystack: Approaches beyond supervised learning in biomedical applications”, Patterns, 2(12), 100383.
[43]Zhu, X., & Goldberg, A. B.,2009,”Introduction to semi-supervised learning”, Synthesis lectures on artificial intelligence and machine learning, 3(1), 1-130.


二、中文文獻
[1]王俊揚，1998，非破壞性混凝土品質檢測系統之研發與應用， 國立台灣大學，碩士論文。
[2]王勝榮，2019，應用卷積神經網路於預測台灣鋼筋價格漲幅之研究，國立成功大學，碩士論文。
[3]王嘉聖，2001，反彈錘有效性之研究，逢甲大學，碩士論文。
[4]沈永年、林彥余，2008，“混凝土版抗壓強度與反彈錘數之關係研究”，高雄應用科技大學學報，第37卷，第5期。
[5]林國榮，1996，超音波縱波波速對不同性質混凝土強度關係之研究，國立中興大學，碩士論文。
[6]林錫國，2006，反彈錘法在卜作嵐混凝土強度檢測評估之應用，國立中興大學，碩士論文。
[7]莊凌雲，2020，使用無人機飛行載具及基於卷積神經網路辨識建築物外牆裂縫檢測之研究，國立高雄科技大學，碩士論文。
[8]陳穎君，2021，應用機器學習於混凝土抗壓強度預測及RC建築梁柱設計，國立台灣大學，碩士論文。
[9]程理明，1999，老舊RC 建築物強度非破壞檢測方法比較研究，中華大學，碩士論文。
[10]魏士翔，2008，“應用電子式反彈錘及適應性類神經模糊推論系統預測混凝土抗壓強度”，技術學刊，第28卷，第2期。