Dynamic characteristics and constitutive model of improved coarse-grained roadbed filler under traffic cyclic loading Authors: Weihong Wu, Bo Liang, Shiliang Shan, Tao Ouyang Published: 2025 ABSTRACT Based on the orthogonal experimental design method, the dry density of each specimen of coal gangue roadbed filling was obtained by the compaction experiment. The maximum dry density of the sample was obtained as follows, fly ash 3%, lime 9%, water 12%, soil 5%. The larger hysteresis loop area indicates the bigger energy consumption of the specimen under a traffic load. The degree of damage to the specimen deepens as the dynamic strain increases, the hysteresis loop area gradually increases, and the damping ratio grows. The smaller hysteresis loop area occurs with the larger the confining pressure. The inclination degree of the long axis of the specimen first rapidly decreases and then tends to flatten with increasing dynamic strain. As the confining pressure of the specimen increases, the inclination degree of the long axis of the hysteresis loop gradually increases. As the dynamic strain increases, the non-closure degree of the hysteresis loop increases approximately linearly. The larger the amplitude of dynamic stress is, the greater the residual deformation generated by the roadbed filling material will be. With increasing confining pressure, the non-closure degree of the hysteresis loop decreases continuously. As dynamic strain increases, dynamic stress also increases. Under increasing confining pressure, the material’s failure occurs at a higher level of dynamic stress. An improved Hardin-Drnevich model was introduced to express the relationship between dynamic stress and dynamic strain for coal gangue roadbed filling material under traffic cyclic loading conditions. Additionally, it is observed that the damping ratio diminishes with increasing dynamic strain and confining pressure. Introduction Coal gangue has many crushed stone particles and fine grains, with properties very similar to those of soil, making it an excellent coarse-grained material1–3. Coal gangue is the rock that dry base greater than 50% produced in the process of coal mine construction, development and excavation, coal mining and coal washing4. It is a kind of black and gray rock with low carbon content and harder than coal associated with coal seam in the process of coal formation, and is the waste generated in the process of coal mining5–7. The long-term accumulation of coal gangue will have a serious impact on the surrounding ecological environment, and its use in roadbed filling will produce good benefits, and the application prospect is very broad8–10. Hence, investigating improvement methods and dynamic behaviors of coal gangue roadbed fill under traffic loading holds significant engineering and scientific value. There are many comprehensive utilization methods for coal gangue. In terms of roadbed material, Qiu et al.11 explained the feasibility of coal gangue as roadbed filler by summarizing the research results and engineering examples of domestic coal gangue as roadbed material. He et al.9 studied the engineering road performance and filling technology of soil-doped coal gangue by using screening, compaction, consolidation, penetration, load-bearing ratio and direct shear tests. They also studied the mechanical properties of fine grain through small-scale triaxial tests under cyclic load, and analyzed the feasibility of soil-doped coal gangue as filling material for highways8. Zhang et al. studied the dynamic characteristics of coal gangue roadbed filling3 and the evolution of particle breakage under cyclic load through large-scale triaxial tests and artificially prepared samples2,12–14. Yang et al.15 studied the parameter indexes of residual strength of coal gangue fillers with different soil content through large-scale direct shear tests. Yang et al.16 investigated the displacement of coal gangue filling material in road construction and microstructural behavior of treated coal gangue by considering the environmental factors of aqueous solutions with different acidity. Overall, some research results have been achieved on coal gangue as a conventional roadbed fill and effective soil blending modification techniques. Yet, few studies focus on the activation and improvement of roadbed fill with consideration of coal gangue’s coarse-grained characteristics, especially its intrinsic properties. Moreover, there are still insufficient studies on its dynamic characteristics under traffic dynamic loads. Hence, the purpose of this paper is to (1) add some admixtures to improve the strength of coal gangue roadbed filling, (2) conduct the compaction test and large-scale triaxial test to study its dynamic behaviors under traffic load, and (3) study the hydration effect and the hysteresis curve and the dynamic constitutive for coal gangue roadbed filling. Test material profile Test material sampling In this paper, coal gangue, a solid waste selected from coal mining, is selected as the research object for coarse grain roadbed filling. The test material is taken from an abandoned Tangjiashan mine in Xiangtan City of Hunan Province. The site of coal gangue accumulation is shown in Fig. 1. Fig. 1Coal gangue accumulation. Coal gangue accumulation. Coal gangue accumulation will lead to a series of environmental problems17. Major environmental concerns involve: (i) land use competition, (ii) particulate matter pollution, (iii) aqueous heavy metal contamination, and (iv) slope stability issues. Therefore, it is urgent to carry out comprehensive utilization of accumulated coal gangue. Coal gangue is widely utilized as roadbed filling and has a broad prospect15. The use of clay cladding can well avoid secondary pollution caused by coal gangue road use. Therefore, the use of coal gangue waste in roadbed filler in this mining area can turn waste into treasure and save costs, which has good engineering and practical significance. Basic properties of test materials Test materials were transported to the laboratory and were dried in an oven at 110℃ holding 24 h to eliminate the effect of initial moisture content. The liquid limit, plastic limit and California bearing ratio (CBR) are 11%, 29% and 0.23 respectively. The specific gravity is 2.08, the natural moisture content is 2.06%. The gradation of sample is displayed in Fig. 2. Gradation curve can also be given by gradation equations12,14,18, and then the effect of gradation on the roadbed filling could be studied19. Considering that the material used in this experiment is crushed roadbed filling, the crushed particle size distribution can be directly adopted in Fig. 2. The main mineral composition of coal gangue was obtained, that is, quartz 41%, calcite 27%, illite 15%, kaolinite 11%, chlorite 4%, feldspar 1%. Fig. 2Particle size distribution test materials. Particle size distribution test materials. Improvement of coal gangue roadbed filling Compaction test of coarse-grained roadbed filling Considering the inherent behaviors of coal gangue coarse particle roadbed filling, the bearing characteristics of coal gangue will be enhanced by adding some other materials. Because the coal gangue includes a large number of coarse grains and the percentage of fine particles is lower, in order to make the gradation of roadbed filling more uniform, the soil is joined for improvement, and the mass percentage of soil is 0%, 5%, 10%, 15%. Considering that coal gangue includes a lot of oxides of Si, Al, Ca and the other, by adding coal ash and lime, it can activate with coal gangue in the condition of water to enhance the bearing properties. Coal ash mass percentage should be 0%, 3%, 6%, 9%, lime mass percentage should be 0%, 3%, 6%, 9%, water mass percentage should be 0%, 6%, 12%, 18%. Note that, the different materials added to improve the coal gangue was replaced by mass percentage. According to the orthogonal testing design20, a four-factor and four-level orthogonal table was utilized to carry out the compaction test. This test was conducted according to the Chinese scientific standards for Test Methods of Soils for Highway Engineering (JTG 3430 − 2020). Firstly, the coal gangue and each admixture are generated following the compaction test design, the coal gangue is fully stirred, and then the coal gangue is sealed and placed for more than 6 h to make full contact21,22. After that, the dry density of the specimen was captured by heavy compaction experiment. Hydration analysis of mixture materials The maximum dry density of the coal gangue coarse particle roadbed filling specimen can be captured by the above compaction experiment23. After comparison, the dry density of the eighth experiment is the biggest i.e., 2.29 g/cm3. Hence, the proportion of admixture corresponding to the eighth group of samples is adopted, that is, 3% fly ash, 9% lime, 12% water, 5% soil. The reason for the above phenomenon can be attributed to the influence of hydration effect. The above-mentioned study of coarse particle roadbed filling indicates that coal gangue in this test includes many compounds of Si, Al, Ca and other elements, then the materials including these substances will undergo hydration reaction under the condition of water after adding coal ash and lime, and then generating calcium hydroxide, calcium silicate hydrate, calcium aluminate hydrate and carbonate crystals. The structure of the specimen is continuously improved with the process of hydration reaction. Therefore, the strength and water stability of the specimen can be continuously enhanced. Hence, the compaction influence and bearing behavior of the specimen are improved. Dynamic characteristics of coal gangue coarse particle roadbed filling under traffic load Large-scale triaxial test design of coarse-grained roadbed filling Dry density is an important indicator for measuring the compactness of roadbed filling, and the larger dry density value indicates the denser roadbed filling. Hence, the above experimental results can be evaluated using dry density, and the experimental group corresponding to the maximum dry density in the above experiment is selected. Then, the percentage of mixed materials corresponding to the maximum dry density of experiments are used to carry out the large-scale triaxial testing. The test method refers to the Chinese scientific standards for Test Methods of Soils for Highway Engineering (JTG 3430 − 2020)9, the compaction degree is 93%, and the compaction of large-scale triaxial test specimens must meet dry density requirements, and the sample size is φ300mm×600 mm. The test equipment adopts a large-scale triaxial test equipment, as illustrated in Fig. 3. This instrument uses sine-wave to carry out traffic cyclic loading to conduct dynamic behavior test, and the traffic cyclic loading frequency is set at 1 Hz. The mass needed for large-scale triaxial experiment of coarse-grained roadbed filling is nearly 80 kg for each specimen, which takes a long time, expensive cost in specimen preparation. Hence, the large-scale dynamic experiment is conducted on the basis of the maximum dry density group of admixture proportion. Fig. 3The large triaxial apparatus. The large triaxial apparatus. The loading methods of large-scale triaxial test is as follows. The cyclic loading amplitude for axial stress of one specimen is taken as: 50 kPa, 70 kPa, 90 kPa… The confining pressures for various specimens were selected as 50 kPa, 100 kPa, 150 kPa and 200 kPa respectively until the loading failure, and the experiments were parallel conducted three times with different test condition. Each dynamic amplitude of axial load was loaded at 10 cycles under traffic load. Hysteresis loop analysis The hysteretic loop reflects the changing process of dynamic stress and strain of coarse-grained roadbed filling samples under traffic load. Figure 4 shows the hysteretic curve of the middle section under loading when the confining pressure is 100 kPa and the axial dynamic stress amplitude is 180 kPa. As shown in Fig. 4 that with increasing dynamic strain, the opening at the bottom of the hysteretic loop of the samples gradually decreases and becomes narrow. The plastic strain increases gradually and the rate of increase decreases gradually during traffic load. As a fundamental reflection of the dynamic behaviors of coarse-grained roadbed filling under traffic load, the hysteresis loop is the core to establishing its dynamic constitutive relationship24. Its shape can characterize the macroscopic mechanical behaviors of the filling, making the study of its shape characteristics of great theoretical and practical significance. Therefore, the study of the shape characteristics of the hysteresis loop has important theoretical and practical significance. As shown in Fig. 4 that the hysteresis loop of the sample under traffic load are approximately elliptical, and the hysteresis loops are not very smooth, which was also confirmed in25. With increasing dynamic strain, a series of changes occur in the distance between the loading and unloading sections of the hysteresis loops, the overall inclination angle, the enclosed area, and the degree of flatness26. Hence, to accurately study the variation law of the hysteresis loop under traffic load, indicators such as the degree of non-closure, the inclination degree of the long axis, the area of the hysteresis loop, and the plumpness of the hysteresis loop are introduced to quantitatively analyze the changes in the shape of the sample hysteresis curve. Fig. 4Hysteretic loop. Hysteretic loop. Shape of hysteresis curve Analysis of the hysteresis loop area Due to the non-closure of the hysteresis loop, the starting and ending points of the loading and unloading sections were connected when calculating the hysteresis loop, and then the area of the closed hysteresis loop was solved. Firstly, the hysteresis loop is artificially divided into two segments, with a positive hysteresis loop slope and a negative hysteresis loop slope. After that, the area of the hysteresis curve can be obtained by subtracting the area enclosed by the positive segment of the hysteresis loop slope and the x-axis from the area enclosed by the negative segment of the hysteresis curve slope and the x-axis. The hysteresis loop area indicates the energy consumption of the specimen due to damping. The bigger hysteresis loop area indicates the greater energy consumption of the specimen under a cyclic load. Figure 5 shows the variation between hysteresis curve area (S) and the dynamic strain (ε). As shown in Fig. 5, as the dynamic strain increases, the damage degree to the specimen deepens, the hysteresis loop area gradually increases, and the damping ratio increases, the smaller hysteresis loop area occurs with the larger the confining pressure. Fig. 5The variation between hysteresis loop area and strain. The variation between hysteresis loop area and strain. Analysis of the inclination degree of the long axis According to Fig. 4, the hysteresis loop of the specimen under traffic load is not completely symmetrical between the loading and unloading sections, and the hysteresis loop is not closed. The slope of the line between the turning point of the loading and unloading section and the midpoint of the starting and ending points of the loading and unloading section within a cycle is utilized to reflect the inclination degree of the long axis for hysteresis loop. The inclination degree of the long axis of hysteresis loop reflects the stiffness and elastic behaviors of the sample. The bigger the inclination degree of the long axis is, the better the elastic performance and dynamic modulus will be. On the contrary, it indicates that it is prone to softening and has a lower dynamic modulus. The relationship between inclination degree of the long axis and dynamic strain is shown in Fig. 6. As shown in Fig. 6, (1) as the dynamic strain increases, the inclination degree of the long axis of the specimen first rapidly decreases and then tends to flatten. This indicates that the sample demonstrates a strong ability to resist deformation in the early stage for cyclic loading, and as the dynamic stress increases, the specimen’s ability to resist deformation gradually weakens. (2) As the confining pressure of the specimen increases, the inclination degree of the long axis of the hysteresis curve gradually increases. With increasing confining pressure, the initial stress on the sample increases, the porosity decreases, the cohesion and friction of the specimen increase, and relative slip becomes difficult, thereby improving the specimen’s ability to resist deformation. Fig. 6The relationship between inclination degree of the long axis and strain. The relationship between inclination degree of the long axis and strain. Analysis of the non-closure degree According to the hysteresis curve of the sample presented in Fig. 4, it is observed that the hysteresis loop of the sample under traffic cyclic load is non-closed. The hysteresis loop initiates at the start of the traffic cyclic load in the experiment and terminates at the end of cyclic load unloading. The degree of non-closure of the hysteresis loop is characterized by the strain difference between the load application start point and the unloading end point within one cycle. The non-closure degree of hysteresis loop reflects the residual strain after one cycle of traffic cyclic loading and unloading. As shown in Fig. 4, under traffic load, the hysteresis loop of the specimen is not closed and asymmetric. This phenomenon occurs because traffic loading induces irreversible plastic deformation in the specimen. Consequently, the center of the hysteresis loop gradually shifts toward the direction of increasing dynamic strain, exhibiting the characteristic of strain accumulation. A relationship curve was plotted with dynamic strain as the horizontal axis and non-closure degree as the vertical axis, as illustrated in Fig. 7. As shown in Fig. 7 that (1) as the dynamic strain increases, the non-closure degree of the hysteresis loop increases approximately linearly. The bigger the amplitude of dynamic stress is, the greater the residual deformation generated by the roadbed filling material will be. (2) With increasing confining pressure, the non-closure degree of the hysteresis loop reduces continuously. This indicates that the compaction influence of confining pressure on samples enhances the interlocking effect between particles, increases cohesion and friction, and makes it difficult for residual deformation to develop. The confining pressure has a significant inhibitory effect on the development of non-closure degree. Fig. 7The relationship between non-closure degree and dynamic strain. The relationship between non-closure degree and dynamic strain. Analysis of the hysteresis loop plumpness As shown in Fig. 4, the hysteresis loop plumpness varies under different experimental conditions. The fullness reflects the degree of viscosity of the sample. The larger plumpness has the greater viscosity, and the smaller plumpness occurs with the smaller viscosity. For the hysteresis loop of sample displayed in Fig. 8, the smaller fullness has the flatter hysteresis loop, the greater fullness happens with the fuller hysteresis loop. Figure 8 shows the variation of hysteresis loop plumpness and dynamic strain. As shown in Fig. 8 that with increasing dynamic strain, the plumpness of the hysteresis loop increases approximately linearly. The effect of confining pressure on the plumpness is obvious, i.e., the lower plumpness appears with the bigger confining pressure. Further analysis shows that with the increase of dynamic strain, grains of specimen are gradually compacted, slipped, and even broken, and the structure of the sample is destroyed, gradually producing irreversible plastic deformation. Under cyclic dynamic stress conditions, the viscosity of the specimen increases, and the hysteresis loop becomes fuller. The higher the confining pressure is, the tighter the contact between the particles inside the sample, the denser the sample, and the lower viscosity will be under traffic load. This is because as the internal density of the sample increases, the spatial blocking effect between particles inside the sample is enhanced, which in turn leads to a decrease in its plastic viscosity27, resulting in a flatter hysteresis loop. Fig. 8The relationship between hysteresis loop plumpness and strain. The relationship between hysteresis loop plumpness and strain. Dynamic constitutive of coarse-grained roadbed filling under traffic loading The dynamic stress and dynamic strain data of coal gangue roadbed filling samples were obtained through the large-scale dynamic triaxial test mentioned above. Draw a curve with dynamic stress as the x-axis and dynamic strain as the y-axis, as shown in Fig. 9. As shown in Fig. 9, as the dynamic strain increases, the dynamic stress first increases approximately in a straight line, then the increasing rate becomes faster, and finally the growth slows down and gradually stabilizes. This is mainly because, under long-term cyclic loading conditions, the volumetric deformation of the sample in the early stage is primarily caused by the rearrangement and redistribution of particles28. Meanwhile, with increasing confining pressure, the dynamic stress grows when the sample is failed. Fig. 9The relationship between dynamic stress and dynamic strain. The relationship between dynamic stress and dynamic strain. To further study the relationship between dynamic stress and dynamic strain of coal gangue roadbed filling specimen under traffic cyclic loading, an improved Hardin-Drnevich model was adopted to express the relationship between dynamic stress and dynamic strain, and its formula is1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\sigma _d}=\frac{{{\varepsilon _d}{E_{d\hbox{max} }}\left( {1+a{e^{ - b\frac{{{\varepsilon _d}{E_{d\hbox{max} }}}}{{{\sigma _{d\hbox{max} }}}}}}} \right)}}{{1+\left| {\frac{{{\varepsilon _d}{E_{d\hbox{max} }}\left( {1+a{e^{ - b\frac{{{\varepsilon _d}{E_{d\hbox{max} }}}}{{{\sigma _{d\hbox{max} }}}}}}} \right)}}{{{\sigma _{d\hbox{max} }}}}} \right|}}$$\end{document} where, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\sigma\:}_{dmax}$$\end{document} is the maximum dynamic stress (kPa), \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{E}_{dmax}$$\end{document} is the maximum dynamic modulus (MPa), \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\sigma\:}_{d}$$\end{document} is dynamic stress, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\varepsilon _d}$$\end{document} is the dynamic strain, a and b are correction coefficients, a is 1/Edmax, b is 1/σdmax. The fitting curve of the dynamic constitutive model (i.e., Eq. (1)) was compared with the dynamic stress-strain test data, as shown in Fig. 10, and the model parameters are shown in Table 1. It can be found that the model curve is in good agreement with the test results, and the two curves are extremely close. This can demonstrate the correctness of the above model. Fig. 10Comparison between the improved Hardin–Drnevich model and test results. Comparison between the improved Hardin–Drnevich model and test results. Table 1The improved Hardin-Drnevich model parameters.σ350100150200 E dmax 10294.2310133.8812304.5315281.80 a −1.21−1.61−1.71−1.93 b 1.301.481.311.17R20.970.980.980.96 Table 1The improved Hardin-Drnevich model parameters.σ350100150200 E dmax 10294.2310133.8812304.5315281.80 a −1.21−1.61−1.71−1.93 b 1.301.481.311.17R20.970.980.980.96 The improved Hardin-Drnevich model parameters. The energy loss can be expressed by using damping ratio under traffic cyclic loading, and the formula is2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\lambda =\frac{1}{{4\pi }}\frac{{\Delta W}}{W}$$\end{document} where, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\lambda\:$$\end{document} is the damping ratio; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\Delta\:}W$$\end{document} is the energy loss under a single cycle load, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:W$$\end{document} is the total energy. The values of damping ratio can be obtained by the above Eq. (2). Figure 11 shows the relationship between damping ratio and dynamic strain. As shown in Fig. 11, with the increase of dynamic strain, the damping ratio of the specimen gradually decreases, and the larger confining pressure of the sample has the smaller damping ratio. The reasons can be explained that as the dynamic strain increases, the stiffness increases and the ability to assimilate dissipation energy decreases. This is because the loose sample resisted stiffness degradation during most of the applied loading cycles and developed cumulative axial extensional strain, which resulted in a decrease in the damping ratio29, which was also confirmed in the research of Hu et al.30 and Zhang et al.31. From a micromechanical perspective, the small-strain stiffness of granular soils is associated with both the stiffness of individual particle contacts and the total number of such contacts. As dictated by the Hertz-Mindlin contact model, the stiffness of a single particle contact is determined by the particle properties32,33. Consequently, the overall stiffness of the granular soil is governed by the number of particle contacts34. Fig. 11The relationship between damping ratio and dynamic strain. The relationship between damping ratio and dynamic strain. Conclusions Based on the orthogonal experimental design method, the dry density of the specimen of each group of coal gangue roadbed fillings was obtained through compaction test. When the dry density of the samples reached the maximum, the proportion of each additive was as follows: fly ash 3%, lime 9%, water 12% and soil 5% respectively.As the dynamic strain increases, the opening at the bottom of the hysteresis curve of the sample gradually decreased and narrowed, reflecting that the plastic strain of the specimen gradually increased and the increase rate gradually decreased during the traffic load process. The failure mode of large dynamic triaxial sample of coal gangue roadbed fill under traffic cyclic loading is mainly intermediate swelling.The larger hysteresis loop area indicates the bigger energy consumption of the specimen under a traffic loading. As the dynamic strain increases, the damage degree to the specimen deepens, the hysteresis loop area gradually increases, and the damping ratio increases. The smaller hysteresis loop area occurs with the larger the confining pressure.The inclination degree of the long axis of the specimen first rapidly decreases and then tends to flatten with the increasing dynamic strain. As the confining pressure of the specimen increases, the inclination degree of the long axis of the hysteresis loop gradually increases.As the dynamic strain increases, the non-closure degree of the hysteresis loop increases approximately linearly. The bigger the amplitude of dynamic stress is, the greater the residual deformation generated by the roadbed filling material will be. The non-closure degree of the hysteresis loop decreases continuously with increasing confining pressure.The dynamic stress of the coal gangue roadbed filling specimen increases as the dynamic strain increases. The larger confining pressure indicates bigger dynamic stress at the time of specimen failure. An improved Hardin-Drnevich model was introduced to express the relationship between dynamic stress and dynamic strain for coal gangue roadbed filling material under traffic cyclic loading conditions. Additionally, it is observed that the damping ratio diminishes with increasing dynamic strain and confining pressure. Based on the orthogonal experimental design method, the dry density of the specimen of each group of coal gangue roadbed fillings was obtained through compaction test. When the dry density of the samples reached the maximum, the proportion of each additive was as follows: fly ash 3%, lime 9%, water 12% and soil 5% respectively. Based on the orthogonal experimental design method, the dry density of the specimen of each group of coal gangue roadbed fillings was obtained through compaction test. When the dry density of the samples reached the maximum, the proportion of each additive was as follows: fly ash 3%, lime 9%, water 12% and soil 5% respectively. As the dynamic strain increases, the opening at the bottom of the hysteresis curve of the sample gradually decreased and narrowed, reflecting that the plastic strain of the specimen gradually increased and the increase rate gradually decreased during the traffic load process. The failure mode of large dynamic triaxial sample of coal gangue roadbed fill under traffic cyclic loading is mainly intermediate swelling. As the dynamic strain increases, the opening at the bottom of the hysteresis curve of the sample gradually decreased and narrowed, reflecting that the plastic strain of the specimen gradually increased and the increase rate gradually decreased during the traffic load process. The failure mode of large dynamic triaxial sample of coal gangue roadbed fill under traffic cyclic loading is mainly intermediate swelling. The larger hysteresis loop area indicates the bigger energy consumption of the specimen under a traffic loading. As the dynamic strain increases, the damage degree to the specimen deepens, the hysteresis loop area gradually increases, and the damping ratio increases. The smaller hysteresis loop area occurs with the larger the confining pressure. The larger hysteresis loop area indicates the bigger energy consumption of the specimen under a traffic loading. As the dynamic strain increases, the damage degree to the specimen deepens, the hysteresis loop area gradually increases, and the damping ratio increases. The smaller hysteresis loop area occurs with the larger the confining pressure. The inclination degree of the long axis of the specimen first rapidly decreases and then tends to flatten with the increasing dynamic strain. As the confining pressure of the specimen increases, the inclination degree of the long axis of the hysteresis loop gradually increases. The inclination degree of the long axis of the specimen first rapidly decreases and then tends to flatten with the increasing dynamic strain. As the confining pressure of the specimen increases, the inclination degree of the long axis of the hysteresis loop gradually increases. As the dynamic strain increases, the non-closure degree of the hysteresis loop increases approximately linearly. The bigger the amplitude of dynamic stress is, the greater the residual deformation generated by the roadbed filling material will be. The non-closure degree of the hysteresis loop decreases continuously with increasing confining pressure. As the dynamic strain increases, the non-closure degree of the hysteresis loop increases approximately linearly. The bigger the amplitude of dynamic stress is, the greater the residual deformation generated by the roadbed filling material will be. The non-closure degree of the hysteresis loop decreases continuously with increasing confining pressure. The dynamic stress of the coal gangue roadbed filling specimen increases as the dynamic strain increases. The larger confining pressure indicates bigger dynamic stress at the time of specimen failure. An improved Hardin-Drnevich model was introduced to express the relationship between dynamic stress and dynamic strain for coal gangue roadbed filling material under traffic cyclic loading conditions. Additionally, it is observed that the damping ratio diminishes with increasing dynamic strain and confining pressure. The dynamic stress of the coal gangue roadbed filling specimen increases as the dynamic strain increases. The larger confining pressure indicates bigger dynamic stress at the time of specimen failure. An improved Hardin-Drnevich model was introduced to express the relationship between dynamic stress and dynamic strain for coal gangue roadbed filling material under traffic cyclic loading conditions. Additionally, it is observed that the damping ratio diminishes with increasing dynamic strain and confining pressure.