Investigation Of A Non Destructive System To Evaluation Critical Properties Of Asphalt

Author : Cheng Zhang
Publisher :
ISBN 13 :
Total Pages : 0 pages
Book Rating : 4.:/5 (144 download)

DOWNLOAD NOW!

Download or read book In-situ Evaluation of Asphalt Pavement Modulus with Embedded Wireless Sensors written by Cheng Zhang and published by . This book was released on 2024 with total page 0 pages. Available in PDF, EPUB and Kindle. Book excerpt: The in-situ dynamic modulus properties of asphalt mixture play a significant role in assessing pavement mechanical responses under traffic loading, determining the pavement performance and condition, and making optimized maintenance decisions. Several methods, such as the falling weight deflectometer (FWD), have been utilized as a non-destructive test to back-calculate the in-situ pavement modulus and conditions; however, the FWD test can only be performed periodically and has the disadvantage of disturbing traffic due to lane-closure needs. With the recent advancement in data science and sensing technologies, the application of micro-electromechanical system (MEMS) sensors and machine learning techniques in pavement nondestructive tests has attracted more research attention. This research aims to develop an in-situ evaluation system that can automatically collect, process, and interpret data to determine the in-situ dynamic modulus of the asphalt mixture under traffic loads using embedded wireless sensors and machine learning techniques. The proposed system is a self-adaptive process and can predict in-situ dynamic modulus based only on mechanical responses and environmental conditions. Ultimately, the well-trained predictive model can be integrated into the pavement management system for the automated and cost-effective assessment of pavement conditions, facilitating informed decision-making. The research program encompasses three types of dynamic modulus experiments: laboratory uniaxial dynamic modulus tests, the one-third scale model mobile load simulator (MMLS3) tests, and in-situ dynamic modulus tests. Particle-size wireless sensors, SmartKli sensors, were implemented in the laboratory specimens and the pavements to collect data from sine wave loads and moving loads. Finite element models (FEM) were also developed and calibrated to generate pavement mechanical response data for more pavement types. The collected data and the FEM simulations were integrated into a database for a proposed adaptive data processing procedure. In addition, because the data collected by embedded sensors in infrastructure health monitoring are inevitably contaminated with noise, and the data features have a distinct discrepancy in different types of tests, a secondary objective of this research is to propose a data processing method capable of removing noises, recognizing data feature discrepancies, and extracting hidden features. An adaptive data processing procedure was developed by combining an empirical mode decomposition (EMD) method and an intrinsic mode function (IMF) selection processing to enhance the reliability of the pavement dynamic modulus prediction. Different EMD techniques were applied to decompose signals from wireless sensors embedded in the pavements. The maximum normalized cross-correlation (MNCC) and signal noise ratio (SNR) were selected as indices in the K-means classification to select the effective IMFs. The results indicated that ensemble EMD (EEMD) and multivariant EMD (MEMD) methods can extract more information from the mechanical responses and extend data dimensions. The EEMD method gives the lowest mean relative error (MRE). Therefore, the EEMD method was recommended for infrastructure data processing. The K-means method can adaptively select the effective IMFs based on the MNCC and SNR. Finally, three dynamic modulus predictive models were developed for different situations. An artificial neural network (ANN) model was developed based on the laboratory test data. This model verified that the ANN model can predict in-situ dynamic modulus. The second dynamic modulus predictive model was developed using the ensemble ANN model to improve the stability of the ANN model, which was trained and tested by the data collected from the MMLS3 test. The third model was developed to predict the dynamic modulus of various asphalt mixtures by fusing a transfer learning approach and Transformer architecture. Besides, the training database was extended with the FEM simulations. The results indicated that the ensemble ANN model is feasible and robust in predicting the dynamic modulus of the asphalt mixture in the MMLS3 test. The transfer learning model is reasonable and robust in predicting the in-situ dynamic modulus of the asphalt pavement.