| 摘要: | 這項研究提出了一種新的數據驅動非破壞性檢測技術(NDT),專門用於評估與水接觸的結構健康監測(SHM)系統。該方法將射頻辨識(RFID)感測器數據與深度學習模型相結合。實驗驗證包含在混凝土和鋼筋混凝土梁(RCB)試體上測試負載-開裂-水滲入機制,以及模擬鋼箱梁的模擬滲漏。我們的研究還探討了奈米材料的整合,以增強智慧感測標籤的靈敏度。結果表明, 智慧感測標籤可以通過sensor code 的總損壞指數(TDI)檢測圓柱形混凝土試樣中的滲漏和裂縫。此外,對鋼筋混凝土梁(RCBs)的測試證實,帶有3D 列印保護外殼的智慧感測標籤可用 嵌入式應用中,通過sensor code 指數檢測滲漏和裂縫。通過改性等離子體輔助電化學剝離石墨(MPGE)奈米塗層改善智慧感測標籤,在奈米片合成過程中不使用額外的架橋劑(CA)時,感應靈敏度提高了約53%,使用額外CA 時提高了約121%。這些結果突出了智慧感測標籤在檢測水的有效性,並確認了試體中存在裂縫。結果還表明,帶有3D 打印保護外殼的智慧標籤可以通過sensor code 和頻率的關係檢測鋼箱梁內的滲漏。此外,我們利用電磁波影像數據集,並使用傳統和預訓練的卷積神經網絡(CNNs)進行分析。研究結果顯示準確率達96%,F1 分數達92%。這項研究通過為水接觸結構的結構評估提供一個強大的框架,為SHM 的發展做出了貢獻。;This research presents a novel data-driven non-destructive technique (NDT) designed specifically for assessing structures in contact with water, aimed at enhancing structural health monitoring (SHM) systems. The methodology integrates empirical data of radio-frequency identification (RFID)-based smart sensor system with deep learning models. The experimental validation involved testing the loading-cracking-immersing mechanism on concrete and reinforced concrete beam (RCB) specimens, as well as simulating constant water flow through a steel box girder to mimic leakage. Our study also explores the integration of nanosheet materials to augment smart tag sensitivity.
Results indicate that smart sensor tags can detect seepage and cracks in cylindrical concrete specimens through the total damage index (TDI) of the sensor code. Additionally, tests on reinforced concrete beams (RCBs) verified that smart tags with 3D-printed protective cases can be used to detect seepage and cracks through the sensor code index in embedding applications. Modifications to the smart tags with a modified plasma-assisted electrochemical exfoliated graphite (MPGE) nanosheet coating layer improved sensing sensitivity by approximately 53% without an additional crosslinking agent (CA) in the nanosheet synthesis process, and by approximately 121% with the inclusion of an additional CA. These results highlight the efficacy of smart tags in detecting seepage as water penetrates through cracks due to the loading-cracking-immersing mechanism, and confirming the presence of cracks in the specimens. The results also show that smart tags with 3D-printed protective cases can detect leakage within the steel box girder through the relation of sensor code and frequency. Moreover,we utilized the electromagnetic-wave (EM-wave) images dataset and analyzed them using classical and pre-trained convolutional neural networks (CNNs). The findings demonstrate promising results, with an accuracy of 96% and an F1-score of 92% in predicting the condition of RCBs under water-contact scenarios. This study contributes to the advancement of SHM by providing a robust framework for the structural assessment of water-contacted structures. |
