TY - JOUR
T1 - Artificial neural network-based intelligent compaction analyzer for real-time estimation of subgrade quality
AU - Imran, Syed Asif
AU - Barman, Manik
AU - Commuri, Sesh
AU - Zaman, Musharraf
AU - Nazari, Moeen
N1 - Publisher Copyright:
© 2018 American Society of Civil Engineers.
PY - 2018/6/1
Y1 - 2018/6/1
N2 - The quality and long-term performance of asphalt pavement depends significantly on the stiffness of the underlying subgrade. The modification of virgin soil, with additives such as cement kiln dust (CKD) and lime and its subsequent compaction with a pad foot roller, can ensure proper support for pavement. The quality of the compacted subgrade is usually verified in spot tests at discrete locations on the base. However, such spot tests do not accurately reflect the quality of support and could potentially leave soft spots undetected, thereby contributing to early deterioration of the pavement. Thus, there is a need to develop a system to estimate the stiffness of an entire subgrade during compaction. Complete coverage of the subgrade will enable the identification of regions of inadequate compaction during construction that can then be rectified before any overlays are placed. An artificial neural network (ANN)-based intelligent compaction (IC) system for estimating the stiffness of subgrade during construction was proposed in this study. The IC systemwas mounted on a vibratory smooth steel drum compactor, used to proof-roll the compacted subgrade. This system was based on the hypothesis that the drum and the underlying subgrade form a coupled system during vibratory compaction. Therefore, changes in the stiffness of the subgrade alter the vibratory response of the drum. The ANNbased system analyzed the vibration of the drum, extracted the patterns, and classified them into vibration levels, translating this information into stiffness, represented as a resilient modulus values for an assumed stress state of the soil. A method was proposed to train the ANN during field compaction and a calibration procedure was developed to map the ANN output to corresponding modulus values. The utility of the system in estimating the stiffness of the subgrade was investigated during the construction of different projects in Oklahoma. Field evaluations indicate that the system was capable of providing real time estimate of subgrade stiffness with an accuracy sufficient for quality-control operations. The IC system presented in this paper was a first step in bringing mechanistic-empirical design to construction of pavement subgrades.
AB - The quality and long-term performance of asphalt pavement depends significantly on the stiffness of the underlying subgrade. The modification of virgin soil, with additives such as cement kiln dust (CKD) and lime and its subsequent compaction with a pad foot roller, can ensure proper support for pavement. The quality of the compacted subgrade is usually verified in spot tests at discrete locations on the base. However, such spot tests do not accurately reflect the quality of support and could potentially leave soft spots undetected, thereby contributing to early deterioration of the pavement. Thus, there is a need to develop a system to estimate the stiffness of an entire subgrade during compaction. Complete coverage of the subgrade will enable the identification of regions of inadequate compaction during construction that can then be rectified before any overlays are placed. An artificial neural network (ANN)-based intelligent compaction (IC) system for estimating the stiffness of subgrade during construction was proposed in this study. The IC systemwas mounted on a vibratory smooth steel drum compactor, used to proof-roll the compacted subgrade. This system was based on the hypothesis that the drum and the underlying subgrade form a coupled system during vibratory compaction. Therefore, changes in the stiffness of the subgrade alter the vibratory response of the drum. The ANNbased system analyzed the vibration of the drum, extracted the patterns, and classified them into vibration levels, translating this information into stiffness, represented as a resilient modulus values for an assumed stress state of the soil. A method was proposed to train the ANN during field compaction and a calibration procedure was developed to map the ANN output to corresponding modulus values. The utility of the system in estimating the stiffness of the subgrade was investigated during the construction of different projects in Oklahoma. Field evaluations indicate that the system was capable of providing real time estimate of subgrade stiffness with an accuracy sufficient for quality-control operations. The IC system presented in this paper was a first step in bringing mechanistic-empirical design to construction of pavement subgrades.
KW - Artificial neural network
KW - Intelligent compaction
KW - Pavement construction
KW - Soil subgrade
KW - Vibration
UR - http://www.scopus.com/inward/record.url?scp=85044842107&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85044842107&partnerID=8YFLogxK
U2 - 10.1061/(ASCE)GM.1943-5622.0001089
DO - 10.1061/(ASCE)GM.1943-5622.0001089
M3 - Article
AN - SCOPUS:85044842107
SN - 1532-3641
VL - 18
JO - International Journal of Geomechanics
JF - International Journal of Geomechanics
IS - 6
M1 - 04018048
ER -