Assessing frost heave susceptibility of gravelly soils based on multivariate adaptive regression splines model
Introduction
Significant frost heave is observed in embankments of high-speed rails (HSR) constructed on seasonally frozen ground, posing a severe threat to railway trains. Field investigations of Harbin-Dalian HSR suggested that the frost heave of subgrade surface contributed substantially to the total deformation (Luo et al., 2019). Since coarse fills (e.g., well-graded gravels) used for constructing roadbed (i.e., subgrade surface) were universally recognized as insensitive to frost action, attempts were made to elucidate the potential frost heave mechanism in HSR embankments. To date, the scientific community primarily attributes the frost-related issues to the moisture accumulation, excessive fines in coarse fills, or the combination of both (Zhang et al., 2015). Field observations confirmed unexpected moisture accumulation phenomena in some HSR foundations, and thus Niu et al. (2020) proved that the frost heave performance of track foundation would be significantly enhanced. Sheng et al. (2015) indicated that the cyclic train loads would cause the accumulation of excess pore water pressure in the subgrade, and hence facilitating moisture migration toward the freezing front and feeding ice formation. In recent years, researchers have developed a new concept that the moisture transfer could occur through coupled liquid and vapor migration (Bai et al., 2018; Gao et al., 2018; Zhang et al., 2016). To this end, Zhang et al. (2019a) and Wang et al. (2019b) applied a fluorescein tracer in one-dimensional freezing tests to investigate the liquid and vapor migration in gravelly soils under open-system conditions. Results indicated that the contribution of vapor migration is non-negligible regarding HSR frost heave issues. In the case of Lanzhou-Xinjiang HSR embankments that were constructed in arid regions, delamination of frost heave still takes place in railway roadbed (Lin et al., 2018; Wu et al., 2018). To sum up, the hydraulic condition and moisture redistribution should be carefully accounted for concerning gravelly soil in earthwork construction.
Meanwhile, a strong correlation was found between the frost heave susceptibility of fill material and its fines content (grain size threshold may vary between cases, e.g., 0.25 mm and 0.075 mm). She et al. (2019) demonstrated the clustering effect of fines in saturated coarse fills upon freezing with X-ray computed tomography technique and proposed an indicator known as clustering ratio for a quantitative assessment of fines distribution. To reveal the effect of fines and other influential factors on the frost heave susceptibility of coarse fills, laboratory unidirectional freezing tests were extensively performed on compacted gravel columns for obtaining the frost heave ratios (FHRs). Wang et al. (2016) and Luo et al. (2019) examined four factors, namely fines content, moisture content, compaction degree, and freezing temperature, showing that fines exerted a major influence on frost heave while the temperature gradient is weakly related. Wang et al. (2014) and Long et al. (2018) further considered the impact of overburden pressure, finding that the applied vertical loads could considerably mitigate the ultimate frost heave with and without water intake. Other studies (Konrad and Lemieux, 2005; Konrad, 2008; Li et al., 2017; Liu et al., 2020; Vinson et al., 1986; Zhang et al., 2019a, Zhang et al., 2019b, Zhang et al., 2019c) investigated the effects of repeated freeze-thaw cycles, compaction energy, segregation potential and constant head (i.e., open/closed system) on frost heave susceptibility of fill material used in HSR embankments. In brief, frost heave of gravelly soil is profoundly affected by the presence of excessive fines; laboratory freezing test has been proven a competent measurement method, which could provide useful information on frost heave of graded gravels under the impact of not only fines content, but also some other fundamental factors. Nevertheless, an appropriate characterization of frost heave and developing a practical guideline in field require an in-depth insight into influential factors interaction and establishing feasible prediction models of soil frost susceptibility from a quantitative point of view.
Multivariate Adaptive Regression Splines (MARS) is a nonlinear and nonparametric method for flexible regression of high dimensional data (Friedman, 1991). It can reflect the underlying functional relationships between predictor and response variables without knowing any assumptions before model training. The model shares its ability to capture optimal variable transformations and high order interactions. The method has been widely reported in the geotechnical literature, such as estimating the liquefaction-induced settlement of shallow foundations (Zheng et al., 2020b), the earthquake-induced uplift displacement of tunnels (Adoko et al., 2013; Zheng et al., 2020a), the undrained shear strength of clay (Samui, 2008; Samui and Kurup, 2012), the shaft resistance of piles (Lashkari, 2013; Zhang et al., 2019a, Zhang et al., 2019b, Zhang et al., 2019c), and the slope stability analysis (Wang et al., 2019a, Wang et al., 2019b; Wang et al., 2020). Therefore, the MARS algorithm is adopted in this study to make a connection between laboratory test data and predictive models and to facilitate field operation.
The paper is organized as follows. Typical unidirectional freezing tests of gravelly soils focusing on high levels of compaction are first presented. The methodology of multivariate adaptive regression splines (MARS) is introduced, and two MARS models for FHRs prediction are then developed based on laboratory freezing dataset from this study and previous literature. The proposed models are validated by a randomly chosen subset that is not used and assessed in terms of statistical performance, relative error, and cumulative probability. The general and coupling effects of four influential factors (predictors) on the FHR (response variable) are finally discussed in conjunction with a relative importance analysis.
Section snippets
Laboratory freezing test
Some methods in terms of segregation potential, plasticity testing, principally gradation and standard testing (e.g., ASTM D5918 - 13e1) have been developed for an indirect analysis of the frost susceptibility of typical geomaterials (Bilodeau et al., 2008; Carter and Bentley, 2016; Lund et al., 2018; Vinson et al., 1986). However, they do not apply to aggregates that contain much larger grains. Modified one-dimensional freezing test (Long et al., 2018; Wang et al., 2014, Wang et al., 2016;
MARS model for predicting frost heave ratio
The normalized heave of the HSR subgrade surface is an efficient and practical indicator of frost heave susceptibility of the subgrade surface aggregates (Konrad, 2008). As a consequence, the frost heave ratio (FHR, ηf) is defined for the unidirectional freezing tests considered in this study, given bywhere ΔH is the measured heave of a specimen, and tf is the initial thickness of soil to be frozen (i.e., the thickness of frozen zone minus frost heave amount). This section focuses on the
Concluding Remarks
The normalized heave of geomaterial, termed frost heave ratio, is proposed as a useful indicator of frost heave susceptibility of gravelly soils. This paper primarily presents two MARS models to predict frost heave ratio of subgrade aggregates under different hydraulic conditions with preconditions provided. The statistical performance of proposed models is assessed in terms of the coefficient of determination (R2), root mean squared error (RMSE), and mean absolute error (MAE). The reliability
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
This work was supported by the National Natural Science Foundation of China [grant numbers 41901073 and 51878560], China Postdoctoral Science Foundation [grant numbers 2019M663556 and 2019M663557], and Fundamental Research Funds for the Central Universities [grant numbers 2682020CX66 and 2682020CX65].
References (41)
- et al.
Predicting tunnel convergence using multivariate adaptive regression spline and artificial neural network
Tunn. Undergr. Sp. Technol.
(2013) - et al.
Water-vapor-heat behavior in a freezing unsaturated coarse-grained soil with a closed top
Cold Reg. Sci. Technol.
(2018) - et al.
Laboratory investigation of the freezing point of saline soil
Cold Reg. Sci. Technol.
(2011) - et al.
Undrained lateral load capacity of piles in clay using artificial neural network
Comput. Geotech.
(2006) - et al.
Experimental study on the water-heat-vapor behavior in a freezing coarse-grained soil
Appl. Therm. Eng.
(2018) - et al.
Laboratory investigation of the freezing point of oil-polluted soils
Cold Reg. Sci. Technol.
(2001) - et al.
Experimental measurement and numerical simulation of frost heave in saturated coarse-grained soil
Cold Reg. Sci. Technol.
(2017) - et al.
Characteristics and controlling factors of frost heave in high-speed railway subgrade, Northwest China
Cold Reg. Sci. Technol.
(2018) - et al.
Evaluating frost heave susceptibility of well-graded gravel for HSR subgrade based on orthogonal array testing
Transp. Geotech.
(2019) Support vector machine applied to settlement of shallow foundations on cohesionless soils
Comput. Geotech.
(2008)
New insights into the frost heave behavior of coarse grained soils for highspeed railway roadbed: Clustering effect of fines
Cold Reg. Sci. Technol.
An experimental study on the frost heave properties of coarse grained soils
Transp. Geotech.
The experiment study of frost heave characteristics and gray correlation analysis of graded crushed rock
Cold Reg. Sci. Technol.
Investigation of liquid and vapor migration in coarse-grained soil during open-system step-freezing test
Cold Reg. Sci. Technol.
Delamination frost heave in embankment of high speed railway in high altitude and seasonal frozen region
Cold Reg. Sci. Technol.
Importance of vapor flow in unsaturated freezing soil: a numerical study
Cold Reg. Sci. Technol.
Characteristics of the liquid and vapor migration of coarse-grained soil in an open-system under constant-temperature freezing
Cold Reg. Sci. Technol.
Multivariate adaptive regression splines model for prediction of the liquefaction-induced settlement of shallow foundations
Soil Dyn. Earthq. Eng.
Gradation influence on frost susceptibility of base granular materials
Int. J. Pave Eng.
Frost Susceptibility
Cited by (24)
Permeation grouting of low-permeability silty sands with colloidal silica
2023, Case Studies in Construction MaterialsExperimental study on the influence of initial water saturation on segregation frost-heaving behavior in silty clay columns
2023, Applied Thermal EngineeringPavement preventive maintenance decision-making for high antiwear and optimized skid resistance performance
2023, Construction and Building MaterialsDiscrete element modelling of the effect of aspect ratio on compaction and shear behaviour of aggregates
2023, Computers and GeotechnicsSurrogate-assisted uncertainty modeling of embankment settlement
2023, Computers and Geotechnics