Prediction of Undisturbed Clay Rebound Index Based on Soil Microstructure Parameters and PSO-SVM Model

• Published in: KSCE journal of civil engineering Vol. 26; no. 5; pp. 2097 - 2111
• Main Authors: Dong, Jiaqi, Wang, Boxin, Yan, Xuexin, Xu, Xinchuan, Xiao, Guangping, Yu, Qingbo, Yao, Meng, Wang, Qing
• Format: Journal Article
• Language: English
• Published: Seoul Korean Society of Civil Engineers 01-05-2022; Springer Nature B.V; 대한토목학회
• Subjects: Accuracy; Algorithms; Civil Engineering; Clay; Correlation; Design parameters; Electron microscopy; Engineering; Geotechnical Engineering; Geotechnical Engineering & Applied Earth Sciences; Industrial Pollution Prevention; Liquid limits; Mathematical models; Microstructure; Moisture content; Parameters; Physical properties; Plasticity index; Population number; Predictions; Scanning electron microscopy; Sensitivity analysis; Shoals; Soil; Soils; Support vector machines; Void ratio; Water content; 토목공학; PSO-SVM; Microstructure parameters; Sensitivity analysis; Microscopic mechanism; Rebound index
• ISSN: 1226-7988; 1976-3808
• DOI: 10.1007/s12205-022-1031-3

The rebound index
C
r
is an important design parameter in engineering construction, and its determination is cumbersome and susceptible to errors. Explaining the macroscopic rebound characteristic parameter
C
r
from the perspective of microscopic mechanism is an important research area that is addressed in this study. In this paper, the different soil parameters, including the
C
r
parameter and the physical parameters (void ratio
e
r
liquid limit water content
W
L
, and plasticity index
I
p
), have been determined through experiments for the undisturbed clay of Chongming East Shoal (CES), Shanghai. Further, scanning electron microscopy (SEM) imaging was used to analyze the microstructural features. Through SEM, the grey correlation degree, the average abundance of structural units
AC
p
and the average equivalent diameter of pores
AD
v
were determined as the soil microstructure parameters with the most significant correlation with
C
r
. The predictive analysis model of
C
r
was then carried out combined with the PSO-SVM algorithm. In order to evaluate the influence of microscopic parameters of soil on the prediction accuracy, four sets of input parameter combinations were used. The results indicate the high prediction accuracy of the developed model. Sensitivity analysis was also carried out, which showed that the sensitivity of
C
r
to
AC
p
and
AD
v
was significantly lower than
e
; however, the difference from
w
L
and
I
p
was small, indicating that it is imperative to consider microscopic parameters while predicting
C
r
. This study, thus, provides a basis and method for predicting the rebound index of soil from the microstructure perspective.