Intelligent monitoring and assessment on early-age hydration and setting of cement mortar through an EMI-integrated neural network. (15th November 2022)

• Record Type: Journal Article
• Authors: Yan, Qixiang; Liao, Xiaolong; Zhang, Chuan; Zhang, Yifeng; Luo, Shuangsheng; Zhang, Dongmei
• Is Part Of: Measurement. Volume 203(2022)
• Publication Date: 2022-11-15
• Subjects: Early-age hydration -- Cement mortar -- PZT sensor -- Electromechanical impedance (EMI) -- Deep learning; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2022.111984
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: The conductance signatures in response to d31 and d33 vibration modes were monitored. The localized and overall changes of EMI spectra well correspond to the mortar's settings. An integrated neural network was built to evaluate the mortar's penetration impedance change. The proposed EMI-IntNet presents superior accuracy and reliability over conventional methods. Abstract: In this study, the authors presented a smart sensing method for monitoring and evaluating the early-age hydration and hardening process of cement mortar through the integration of the EMI technique with deep learning (DL) models. The main research works and contributions of this research are given as follows: The development of the penetration resistance was acquired by measuring the Vicat needle penetration depth. Meanwhile, the EMI spectra within two scanning frequency ranges, 1000 kHz and 3000 kHz, were continuously captured. The key innovation of this research is that a DL-based framework named EMI-integrated neural network (EMI-IntNet) was proposed, which could facilitate the achievement of intelligent monitoring and assessment. The EMI-IntNet, which combines a one-dimensional convolutional neural network (1D-CNN), a long short-term memory neural network (LSTM) and an attention mechanism, can accurately predict the dynamic penetration impedance from raw EMI data without any pre-processing or subjective judgment. Two datasets collected in two separate vibration modes ( d31 and d33 ) were used to validate the effectiveness of the proposed method. The results show that the EMI-IntNet can quantitatively evaluate the dynamic penetration impedance of cement mortar with an error of less than 2 % and a coefficient of determination ( R 2 ) value over 0.99 for both datasets. The EMI-IntNet's forecasting accuracy, noise immunity and adaptability were thoroughly investigated, which outperform that of the machine learning-based approaches, 1D-CNN and 1D-CNN-LSTM. In conclusion, the proposed method enables automatic data acquisition, feature extraction and identification of EMI signals, contributing to the intelligent monitoring and accurate assessment of the mechanical impedance change throughout the early-age hydration and setting of cement mortar.