A portable roadbed in-situ dynamic rebound modulus testing device and method

Abstract

The present invention discloses a portable in-situ roadbed dynamic rebound modulus testing device and method. The testing device includes a static load application device, a dynamic load application device, and a data acquisition device. The static load application device includes an annular outer ring loading plate, the top of which is connected to a static load cylinder, and the outer ring counterweight device is placed on a vehicle frame. The dynamic load application device includes a circular loading plate, which is installed in the middle of the outer ring loading plate, and the middle of the top of the loading plate is connected to the output end of the dynamic load cylinder, which is installed at the bottom of the inner ring counterweight device. The inner ring counterweight device is placed on the top of the inner ring reaction frame. The inner ring telescopic legs can separate the inner ring reaction frame from the vehicle frame. The data acquisition device is installed at the bottom center of the loading plate and on the road surface outside the outer ring loading plate. Based on machine learning algorithms and nonlinear finite element numerical calculation methods, the present invention achieves the acquisition of roadbed modulus under full-scale road surface constraints with a relatively small instrument weight, greatly improving portability and applicability.

Description

Technical Field

[0001] The invention belongs to the technical field of road engineering and relates to a portable roadbed in-situ dynamic rebound modulus testing device and method. Background Art

[0002] The subgrade modulus exhibits significant stress dependence, with significant differences in modulus under different stress states. This characteristic predicts significant differences between the subgrade modulus during the construction phase (without pavement structure and vehicle loads) and the operational phase (with pavement structure and vehicle loads). The subgrade design modulus should essentially be the modulus during the operational phase. Currently, after subgrade construction is completed, modulus (deflection) testing is often used to evaluate subgrade construction quality. However, traditional field testing methods (bearing plates, Beckmann beams, and PFWD) treat the subgrade as a linear elastic system and often ignore the effects of pavement structure loads. Consequently, test results fail to represent the actual mechanical properties of the subgrade during operational phases.

[0003] To solve this problem, pavement structure constraints can be applied simultaneously during the subgrade modulus test. However, how to accurately simulate pavement structure constraints becomes the primary challenge. According to the existing highway pavement structure of 75cm, its static load is about 20kPa. If the loading range is 2m 2 , the required restraint load is as high as 4 tons, which makes the development of corresponding test instruments face difficulties such as heavy weight, non-portability and high difficulty in development. Summary of the Invention

[0004] In order to solve the above problems, the present invention provides a portable roadbed in-situ dynamic rebound modulus testing device, which can simulate real driving loads, reduce the weight and size of the instrument, improve portability and applicability, and solve the problems existing in the prior art.

[0005] Another object of the present invention is to provide a portable in-situ dynamic rebound modulus testing method for roadbed. Based on machine learning algorithms and nonlinear finite element numerical calculation methods, the constraint effect of the pavement structure is established, and the roadbed modulus under full-scale pavement constraints is obtained with a smaller instrument weight, so as to quickly predict the mechanical properties of the roadbed under actual operating conditions during the roadbed construction period.

[0006] The technical solution adopted by the present invention is a portable roadbed in-situ dynamic rebound modulus testing device, comprising:

[0007] A static load applying device, comprising an annular outer ring loading plate, the top of which is connected to the telescopic end of a static load cylinder, the static load cylinder being mounted on the bottom of the vehicle frame above the outer ring loading plate, and an outer ring counterweight device being placed on the vehicle frame to serve as a reaction frame for the outer ring load application;

[0008] A dynamic load applying device, comprising a circular loading plate, the loading plate being mounted in the middle of the outer ring loading plate, the outer ring loading plate being flush with the bottom of the loading plate, the top middle of the loading plate being connected to the output end of the dynamic load cylinder, and the dynamic load cylinder being mounted at the bottom of the inner ring counterweight device; the inner ring counterweight device being placed on top of the inner ring reaction frame, the inner ring reaction frame being placed above the vehicle frame, the inner ring telescopic legs being symmetrically mounted at the bottom of the inner ring reaction frame to provide gravity support for the inner ring reaction frame; the inner ring telescopic legs being capable of separating the inner ring reaction frame from the vehicle frame to avoid mutual interference between the inner ring dynamic load and the outer ring static load;

[0009] The data acquisition device is installed at the bottom center of the loading plate and the road surface outside the outer ring loading plate, and all sensors are arranged in a straight line.

[0010] Another technical solution adopted by the present invention is a portable roadbed in-situ dynamic rebound modulus testing method, which uses the above-mentioned portable roadbed in-situ dynamic rebound modulus testing device and includes the following steps:

[0011] S1: Determine the on-site test point for the roadbed. The roadbed surface near the point must be flat. Place the in-situ dynamic rebound modulus test device above the test point. Apply loads to the loading plate and outer ring loading plate according to the set loading sequence to obtain the measured rebound deflection under different loading sequences or stress states.

[0012] S2, based on the measured rebound deflection under each loading sequence or stress state collected in S1, combines machine learning algorithms to invert a series of key parameters reflecting the nonlinear characteristics of the subgrade soil;

[0013] S3, based on the subgrade nonlinear parameters obtained in S2, constructs a finite element model that takes into account the subgrade nonlinearity and includes the actual pavement structure, and calculates the pavement surface displacement response; according to the deflection equivalence principle, the equivalent modulus of the actual designed pavement structure and the vehicle load is obtained, that is, the dynamic rebound modulus correction is realized.

[0014] The beneficial effects of the present invention are:

[0015] 1. This embodiment of the present invention fully considers the stress-dependent nature of the subgrade soil modulus. Using a testing method based on an in-situ loading sequence, it enables on-site acquisition of subgrade nonlinear model parameters. This fully considers the stress-dependent nature of the subgrade and the load effects of the pavement structure. The device is compact and features a fast back-calculation method. This allows for rapid prediction of the mechanical properties of the subgrade under actual operational conditions during construction, effectively resolving the current mismatch between the subgrade design and operational state.

[0016] 2. The embodiment of the present invention applies loads through a loading sequence, which can fully capture the stress-dependent characteristics of the roadbed soil rebound modulus. Compared with the traditional test method based on linear elasticity, it is a major innovation and can further improve the current roadbed rebound modulus design method. It is different from traditional means in both testing and calculation models. The roadbed modulus obtained by the method of the present invention is more representative of the actual mechanical properties of the roadbed during operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 3D schematic diagram of the test device according to an embodiment of the present invention (wheels are not drawn).

[0019] Figure 2 1 is a front view of a testing device according to an embodiment of the present invention (mobile mode).

[0020] Figure 3 1 is a front view of the test device according to an embodiment of the present invention (loading mode).

[0021] Figure 4 2 is a schematic structural diagram of a loading plate in an embodiment of the present invention.

[0022] Figure 5 It is a flow chart of the testing method according to an embodiment of the present invention.

[0023] Figure 6 This is the inner ring load mode in the embodiment of the present invention.

[0024] Figure 7 It is the finite element model of the roadbed in the embodiment of the present invention.

[0025] Figure 8a is the correlation coefficient of the density prediction result in the embodiment of the present invention.

[0026] Figure 8b is the correlation coefficient of the Poisson's ratio prediction result in the embodiment of the present invention.

[0027] Figure 8c is the correlation coefficient of the k1 prediction result in the embodiment of the present invention.

[0028] Figure 8d is the correlation coefficient of the k2 prediction result in the embodiment of the present invention.

[0029] Figure 8eis the correlation coefficient of the k3 prediction result in the embodiment of the present invention.

[0030] Figure 9a is the average error of the density prediction result in the embodiment of the present invention.

[0031] Figure 9b is the average error of the Poisson's ratio prediction results in the embodiment of the present invention.

[0032] Figure 9c is the average error of the k1 prediction results in the embodiment of the present invention.

[0033] Figure 9d is the average error of the k2 prediction results in the embodiment of the present invention.

[0034] Figure 9e is the average error of the k3 prediction results in the embodiment of the present invention.

[0035] Figure 10a This is a comparison of the prediction results of the test set confining pressure of 1.0 kPa in the embodiment of the present invention.

[0036] Figure 10b This is a comparison of the prediction results of the test set confining pressure of 5kPa in the embodiment of the present invention.

[0037] Figure 10c This is a comparison of the prediction results of the test set confining pressure of 15kPa in the embodiment of the present invention.

[0038] Figure 10d This is a comparison of the prediction results of the test set confining pressure of 25kPa in the embodiment of the present invention.

[0039] Figure 11 It is the finite element model of the pavement structure in the embodiment of the present invention.

[0040] Figure 12 It is the deflection time history curve at the load center in the embodiment of the present invention.

[0041] Figure 13 It is the determination of the roadbed modulus based on the deflection equivalence principle in the embodiment of the present invention.

[0042] In the figure, 1. wheel, 2. inner ring telescopic support leg, 3. loading plate, 4. outer ring loading plate, 5. flexible rubber pad, 6. static load cylinder, 7. dynamic load cylinder, 8. inner ring reaction frame, 9. inner ring counterweight device, 10. outer ring counterweight device, 11. frame, 12. sensor steel beam, 13. sensor housing, 14. spring, 15. high-precision displacement sensor, 16. sensor contact pad. DETAILED DESCRIPTION

[0043] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0044] The basic concept of the embodiment of the present invention is as follows:

[0045] On the basis of ensuring the portability of the device and the test accuracy, the embodiment of the present invention proposes a portable roadbed in-situ dynamic rebound modulus testing device and method based on machine learning algorithms and nonlinear finite element numerical calculation methods, which realizes the acquisition of the roadbed modulus under full-scale pavement constraints with a smaller instrument weight, greatly improving the portability and applicability.

[0046] The main features are as follows: 1) Test objectives: a series of key parameters that reflect the stress-dependent characteristics of the roadbed, rather than a single modulus indicator; 2) Test method: a loading sequence in the form of a series of static and dynamic load combinations, which can fully collect the on-site nonlinear characteristics of the roadbed soil and ensure the accuracy of the target parameter calculation; 3) Parameter inversion method: a fast calculation method for the nonlinear parameters of the roadbed is developed based on the machine learning algorithm; 4) Modulus calculation method: the rebound modulus of the roadbed under the constraints of the full-scale pavement structure is obtained by combining the nonlinear finite element numerical calculation method. The whole process has a fast calculation speed, and the roadbed rebound modulus can be measured and used immediately.

[0047] Example 1,

[0048] A portable in-situ dynamic rebound modulus test device for roadbed, such as Figures 1 to 4 As shown, it includes a static load applying device, a dynamic load applying device and a data acquisition device;

[0049] The static load applying device includes an annular outer ring loading plate 4. The top of the outer ring loading plate 4 is connected to the telescopic end of the static load cylinder 6. The static load cylinder 6 is installed at the bottom of the frame 11 above the outer ring loading plate 4. The outer ring counterweight device 10 is placed on the frame 11 and serves as a reaction frame for the outer ring load application.

[0050] The dynamic load applying device includes a circular loading plate 3, which is installed in the middle of the outer ring loading plate 4. The outer ring loading plate 4 is flush with the bottom of the loading plate 3. The middle of the top of the loading plate 3 is connected to the output end of the dynamic load cylinder 7, and the dynamic load cylinder 7 is installed at the bottom of the inner ring counterweight device 9; the inner ring counterweight device 9 is placed on the top of the inner ring reaction frame 8, and the inner ring reaction frame 8 is placed above the frame 11. The inner ring telescopic legs 2 are symmetrically installed at the bottom of the inner ring reaction frame 8 to provide gravity support for the inner ring reaction frame 8; the inner ring telescopic legs 2 can separate the inner ring reaction frame 8 and the frame 11 to avoid mutual interference between the inner ring dynamic load and the outer ring static load.

[0051] Among them, the position where the inner ring counterweight device 9 is placed on the inner ring reaction frame 8 is raised upward to provide sufficient installation space for the dynamic load cylinder 7. The telescopic end of the dynamic load cylinder 7 passes through the frame 11 to avoid mutual interference between the inner ring dynamic load and the outer ring static load.

[0052] The data acquisition device is installed at the bottom center of the loading plate 3 and on the road surface outside the outer ring loading plate 4, with all sensors arranged in a straight line. The data acquisition device includes a high-precision displacement sensor 15, which is installed at the bottom center of the loading plate 3. The dynamic load cylinder 7 is fixedly connected to the horizontally arranged sensor steel beam 12, and multiple high-precision displacement sensors 15 are evenly arranged on the sensor steel beam 12. A sensor contact pad 16 is installed on the section of the high-precision displacement sensor 15 close to the road surface. A sensor housing 13 is provided on the outside of the high-precision displacement sensor 15, and a spring 14 is installed between the sensor housing 13 and the high-precision displacement sensor 15. The high-precision displacement sensor 15, that is, an acceleration (seismic) sensor, has a specific structure known in the art and can achieve real-time measurement of displacement, which serves as the main input layer parameter of the prediction model.

[0053] The lower end of the inner telescopic leg 2 is mounted with a wheel 1. Before the inner telescopic leg 2 elevates the inner reaction frame 8, the inner reaction frame 8 acts directly on the vehicle frame. The entire test vehicle is in free-wheeling mode. The gravity support point of the vehicle frame 11 is the wheel 1. The static load reaction force of the outer ring comes from the vehicle frame itself and the outer ring counterweight 10. The dynamic load reaction force of the inner ring comes from the inner reaction frame 8 and the inner ring counterweight 9. During testing of the entire device, due to the action of the inner telescopic leg 2, the loads do not interfere with each other.

[0054] An annular outer ring loading plate 4 is provided on the outside of the loading plate 3. Flexible rubber pads 5 are installed on the bottom of the loading plate 3 and the outer ring loading plate 4 (the side close to the road surface). The high-precision displacement sensor 15 is tightly pressed against the roadbed surface by the spring 14 to avoid the influence of dynamic load vibration on the test results.

[0055] The inner ring counterweight device 9 and the outer ring counterweight device 10 are custom-made metal weights of a specific weight.

[0056] The research and development ideas of the test device of the embodiment of the present invention were formed during the process of developing the instrument device (patent with publication number CN 110749518A, hereinafter referred to as the original invention) applied by the inventor in the early stage.

[0057] Since the original invention can exert a complete pavement structure constraint, the total weight is more than 4 tons and the loading plate diameter is 60 cm, which makes the entire device heavy and not portable. The present invention combines the theoretical correction algorithm to propose a portable roadbed in-situ dynamic rebound modulus test device, method and calculation method, which realizes the simulation of full-scale pavement constraints with a smaller instrument weight, greatly improving portability and applicability. The solution of the present invention is to abandon the original linear elastic model and adopt a nonlinear model to consider the stress-dependent characteristics of the roadbed, and based on this, solve the negative impact of insufficient constraint force caused by lightweighting. Although the core purpose of the present invention and the original invention is to obtain the rebound modulus of the roadbed under pavement constraints, there are still significant differences in the test method, calculation method, loading device, size and weight.

[0058] 1) In terms of testing method, the original invention uses a single stress state to obtain the rebound deflection of the roadbed under the constraints of the pavement structure, while the present invention uses a loading sequence method that combines several dynamic loads and static loads to fully collect the nonlinear characteristics of the roadbed soil.

[0059] 2) In terms of calculation method, the original invention directly uses a linear elastic model to directly calculate the subgrade modulus under pavement constraints, while the present invention uses machine learning combined with a nonlinear finite element numerical calculation method to indirectly calculate the subgrade modulus under pavement constraints.

[0060] 3) As for the loading device, since the testing and calculation methods of the present invention are significantly different, the control system and processing system thereof are also quite different.

[0061] 4) In terms of size and weight, the large instrument's load plate 12 has a diameter of 30 cm, the loading plate 22 has an inner diameter of 31-32 cm, and an outer diameter of 60 cm. The total weight of the large instrument is 4-6 tons, and the counterweight 18 has a loading range of 0-5 kN. In this embodiment of the present invention, the loading plate 3 has a diameter of 15-25 cm, the outer diameter of the outer ring loading plate 4 is 30-50 cm, the loading range is 0-3 kN, the load adjustment level is 0.10 kN / level, the loading load is 40-60 kPa, and the total weight of the device is 0.6-0.8 tons. The frame is designed to be as lightweight as possible, and all reaction forces are achieved through removable counterweights. The reduction in the loading plate area reduces the total load by over 80%.

[0062] In summary, while the present invention and the original invention share certain similarities in their loading mechanisms, there are significant differences in their approaches and development concepts. The original invention primarily applies full-scale road surface constraints by applying sufficient load during testing, while the present invention applies full-scale road surface constraints indirectly through a theoretical correction algorithm.

[0063] Example 2,

[0064] A portable in-situ dynamic rebound modulus test method for roadbed, such as Figure 5 As shown, the following steps are included:

[0065] S1. Identify the on-site test point for the roadbed. The roadbed surface near the test point must be essentially flat. For minor potholes, fill them with fine sand (less than 2 mm thick). Next, place the loading device above the test point and apply loads to loading plate 3 and outer ring loading plate 4 according to the loading sequence shown in Table 1. Record the load at each level and the rebound deflection at 30 cm and 60 cm. The inner loading plate 3 is loaded with a half-sine pulse load with a loading time of 0.1 s and a loading frequency of 1 Hz.

[0066] The dynamic load diagram of the loading plate 3 in the embodiment is as follows: Figure 6 As shown. After theoretical verification, in order to minimize the weight of the device while ensuring test accuracy, the recommended parameter values ​​of the loading device are as follows:

[0067] 1) The radius of the loading plate 3 is 10 cm and the thickness is 10 mm;

[0068] 2) The inner radius of the outer ring loading plate 4 is 10.5 cm, the outer radius is 20 cm, and the plate thickness is 10 mm;

[0069] 3) Inner ring counterweight 300kg;

[0070] 4) Outer ring counterweight 300kg;

[0071] 5) The distance between the inner and outer ring telescopic legs and the load center exceeds 80cm;

[0072] 6) The displacement sensor test accuracy shall not exceed 2μm.

[0073] 7) The total weight of the device is within 800kg, and it is in the form of a small trailer, which can complete the test under the towing of various types of vehicles.

[0074] Table 1 Subgrade field loading sequence

[0075]

[0076] The specific operation of the loading device is as follows:

[0077] S11: Move to the upper part of the measuring point and adjust the inner ring telescopic legs 2 until the inner ring reaction frame 8 is completely separated from the vehicle frame 11. Then, the static load cylinder 6 and dynamic load cylinder 7 move the loading plate 3 and the outer ring loading plate 4 downward until they contact the roadbed surface.

[0078] S12, adjust the static load cylinder 6 and the dynamic load cylinder 7 to preload a predetermined load, and statically press for 1 to 2 minutes to ensure that the loading plate 3 and the outer ring loading plate 4 fully contact the roadbed surface and eliminate the plastic deformation of the roadbed;

[0079] S13, the controller applies the loading sequence step by step. When loading, first adjust the static load cylinder 6 to the predetermined static load size. After the load stabilizes, the dynamic load cylinder 7 is adjusted to apply a half-sine pulse load of 0.1s and 1Hz frequency ( Figure 6 When each level of load is applied, the rebound deflection is calculated according to formula (4), and the stability of the rebound deflection is determined according to formula (5). If the rebound deflection is stable, the rebound deflection under the load of this level is calculated according to formula (6).

[0080]

[0081]

[0082]

[0083] Where, is the rebound deflection of the i-th loading under the n-th loading level; represents the peak deflection in the i-th loading under the n-th loading level; represents the stable deflection peak value after the end of the i-th loading under the n-th loading level; ω (n) It is expressed as rebound deflection under the nth level of loading, They represent the rebound deflections obtained at the i-1th and i-2th times respectively.

[0084] S14, after the test is completed, first adjust the dynamic load cylinder 7 to unload the inner ring load; adjust the static load cylinder 6 and the dynamic load cylinder 7 to lift the inner ring loading plate and the outer ring loading plate; finally, adjust the inner ring telescopic support leg 2 until the inner ring reaction frame 8 is fully in contact with the frame 11, until it changes from the loading mode to the moving mode, and repeat steps S11 to S14 to the next measuring point to complete the test.

[0085] S2, combined with the recorded rebound deflection data for a total of 24 steps of the 8-stage loading sequence, uses a machine learning algorithm to invert the subgrade nonlinear model parameters k1, k2, and k3. The nonlinear model described in this embodiment of the present invention is the NCHRP1-28A model, which is currently widely used both domestically and internationally, but is not limited to this model. The specific form is shown in Equation (1). The machine learning algorithm is a randomized BP neural network swarm algorithm, and its learning samples are obtained through finite element batch calculation methods.

[0086]

[0087] Where k1, k2, k3 are nonlinear model parameters; θ is the body stress, θ = σ1 + σ2 + σ3; τ oct is the octahedral shear stress, σ1, σ2, σ3 represent the stress magnitude in the x, y, and z directions respectively. oct Calculate according to formula (2) and formula (3) respectively, pa Indicates standard atmospheric pressure, 100kPa.

[0088]

[0089]

[0090] Where ρ is the roadbed density; z is the calculation depth, that is, the depth from the top of the roadbed; g is the acceleration of gravity; μ is the Poisson's ratio; σ r is the horizontal dynamic load; σ z is the numerical dynamic load. Equations (2) and (3) are based on the body stress θ and the octahedral shear stress τ oct Define the obtained formula.

[0091] The specific random BP neural network group algorithm is:

[0092] S21. Using the finite element method, the dynamic response of a large number of subgrades with different densities ρ, Poisson's ratios μ, and model parameters k1, k2, and k3 under the loading sequence shown in Table 1 was theoretically simulated. A total of 24 deflection values ​​were recorded at positions r = 0.0, 0.3, and 0.6 m under each load level. Furthermore, a learning sample library containing the subgrade soil nonlinear parameters k1, k2, and k3 and the deflection values ​​under each load level was constructed. The deflection values ​​were all processed using the interference term in Equation (7) to account for the measurement error (±2 μm) of the displacement sensor.

[0093] w=w0+(1-2·rand)×2 (7)

[0094] Where w0 is the deflection value obtained by theoretical calculation; rand is a random number between 0 and 1; and w is the deflection obtained after interference processing.

[0095] The finite element calculation method for roadbed stress dependence used in S21 and S32 is my previous work. The implementation steps are as follows:

[0096] S211, assign initial modulus to each unit of granular material and subgrade soil structure layer i is the unit number and t0 is the initial time step.

[0097] S212, t is obtained by using the finite element numerical calculation method. k The strain magnitude of each unit under the time step, and the unit stress magnitude is calculated according to formula (8) in each time step.

[0098]

[0099] Wherein, (n) represents the nth iteration, n = 0, 1, 2, ...; Indicates that unit i is in the tth kTime step, node Lamé coefficient of the nth iteration; Indicates that unit i is in the tth k Time step, node shear modulus of the nth iteration; Lame coefficient and shear modulus are physical parameters in elastic mechanics, often used to describe the stress-strain relationship.

[0100] are the unit i in the tth k Time step, radial, vertical, hoop, and tangential stress magnitudes at the nth iteration; Indicates the element modulus size of element i at the nth iteration; are the radial, vertical, circumferential and tangential strain increments of element i at the nth iteration respectively; i represents the finite element number; t k is the calculation time step, t k =t0,t1,t2,...,T; T is the total calculation time; Respectively represent the t k-1 The radial, vertical, circumferential, and tangential strains at the end of the time step iteration. The superscript “—” indicates the stress and strain values ​​at the end of the iteration; μ represents Poisson’s ratio.

[0101] S213, update t according to formula (9) k The modulus of element i at the (n+1)th iteration of the time step and the recalculation of the strain increment Convergence is determined according to formula (10). When the condition is met, the iteration is terminated. If not, S212 to S213 are repeated.

[0102]

[0103] Where, Represents t k Time step, body stress calculated at the nth iteration of element i, Represents t k Time step, radial stress magnitude calculated at the nth iteration of element i, Represents t k Time step, vertical stress magnitude calculated at the nth iteration of element i; Indicates t k Nodal modulus of elasticity for element i at the (n+1)th iteration at time step.

[0104] Indicates t k The magnitude of the octahedral shear stress calculated at the nth iteration of time step, element i.

[0105]

[0106] In formula (10), mean represents the averaging function.

[0107] S214, output t k modulus at the end of the time step iteration Information, update the tth unit i according to formula (11) k Time step Jacobian matrix DDSDDE i (t k ), update the unit stress according to formula (12); then enter the next time step t k+1 Iterative calculation (S212-S214), the initial value of the unit modulus is determined according to formula (13), represents the next time step t of unit i k+1 The initial modulus is calculated until all response calculations are completed and the peak values ​​of the key response indicators shown in Table 1 are output.

[0108]

[0109]

[0110]

[0111] Indicates that unit i is in the tth k Time step, node Lame coefficient when the iterative calculation is completed, Indicates that unit i is in the tth k Time step, node shear modulus at the end of the iterative calculation; the last iterative calculation That is

[0112] Respectively represent the unit i in the t k Time step, radial, vertical, circumferential and tangential stress magnitudes at the completion of the iterative calculation; the last iterative calculation That is

[0113] Respectively represent the unit i in the t k Time step, radial, vertical, circumferential, and tangential strain increments at the completion of the iterative calculation; When the iteration is completed

[0114] S22: Normalize the learning sample library obtained in S21. Then, using 24 parameters, such as the theoretical rebound deflection under 8 levels of load, as input layer parameters, and five parameters, such as the subgrade density ρ, Poisson's ratio μ, and the subgrade soil nonlinear parameters k1, k2, and k3, as output layer parameters, randomly construct several groups (500 or more are recommended) of BP neural networks for sample learning. The constructed BP neural network uses the trainlm algorithm for training and the tansig algorithm for activation functions; all other parameters, such as the training upper limit, number of hidden layers, number of hidden layer neurons, and learning rate, are generated using random numbers.

[0115] S23, based on the field measured rebound deflection obtained in step S1, apply several groups of BP neural networks obtained in step S22 to perform parameter prediction, discard 10% of the maximum and minimum values ​​in the prediction results, and average the remaining parameters as the final roadbed parameter inversion result.

[0116] S3. Based on the subgrade nonlinear model parameters k1, k2, and k3 obtained in step S2, a nonlinear finite element model of the subgrade is constructed that includes the actual pavement structure, and its surface displacement response is calculated. According to the deflection equivalence principle, the equivalent modulus of the actual designed pavement structure and the vehicle load is obtained, thereby realizing the correction of the dynamic rebound modulus.

[0117] The specific method of obtaining the equivalent stiffness of the roadbed is as follows:

[0118] S31: Build a linear elastic finite element model based on the actual pavement. The pavement structure thickness and modulus are determined based on the "Highway Asphalt Pavement Design Specification JTG D50-2017." Subgrade moduli are assumed to be 20 MPa, 50 MPa, 75 MPa, 100 MPa, 150 MPa, 200 MPa, and 300 MPa. Calculate the road surface deflection for each subgrade modulus, and ultimately generate a subgrade modulus-road surface deflection curve.

[0119] In step S32, the road surface deflection is calculated using a finite element method that accounts for stress nonlinearity. The subgrade parameters are derived from step S23, and the remaining parameters remain the same as in step S31. After the road surface deflection is obtained, the road surface equivalent elastic modulus is calculated using the subgrade modulus-road surface deflection curve obtained in step S31.

[0120] This method breaks away from traditional linear elastic field testing of roadbed modulus of resilience by fully considering the stress-dependent nature of the roadbed soil modulus and using nonlinear roadbed model parameters as inversion targets. To avoid the multiple solutions resulting from excessive inversion parameters, a loading method based on field loading sequences is innovatively proposed. The inversion calculation algorithm utilizes cutting-edge machine learning algorithms, resulting in high accuracy and efficiency.

[0121] The embodiment of the present invention mainly proves the reliability of the proposed test method theoretically, and introduces how to further obtain the equivalent elastic modulus of the roadbed during the operation period based on the test results.

[0122] 1) Construct a subgrade rebound deflection database based on loading sequences;

[0123] Based on the parameter value range shown in Table 2, 2000 groups of roadbed parameters were randomly generated, and the rebound deflection under the loading sequence shown in Table 1 was calculated respectively. After the interference terms were processed, the roadbed rebound deflection database was finally constructed. The roadbed finite element model is as follows: Figure 7 As shown, some of the database results are shown in Table 3. (Recommended calculation software: Gu Fan, Fan Haishan, Zhang Junhui. Asphalt pavement finite element automation batch processing software based on ABAQUS considering material nonlinear characteristics V1.0, 2022SR0875907)

[0124] Table 2 Roadbed parameter values

[0125]

[0126] Table 3 Partial examples of roadbed rebound deflection database

[0127]

[0128]

[0129] The numbers in Table 2 and Table 3 are in units of (t / m 3 ), the other parameters are dimensionless parameters.

[0130] 2) Based on the data obtained in 1), a training set and a test set were randomly divided into two groups with a ratio of 7:3. Then, 500 BP neural networks were randomly generated and trained using the training set data. The parameters of each BP neural network were obtained using a random algorithm, with the value range shown in Table 4. The training method was trainlm, and the activation function was tansig.

[0131] Table 4 BP neural network parameter values

[0132]

[0133] 3) Based on the 500 sets of BP neural networks obtained in 2), the test set is used for testing to obtain 500 sets of BP neural network prediction results. 10% of the maximum and minimum values ​​in the prediction results are discarded, and the mean of the remaining parameters is taken as the final roadbed parameter inversion result. The correlation coefficient and average error of the prediction results are calculated respectively. The results of the correlation coefficient are as follows: Figures 8a to 8e The results of the average error are shown as Figures 9a to 9eAs shown in the figure, it can be seen that the accuracy of the random BP neural network group algorithm exceeds that of most BP neural networks and has higher reliability.

[0134] 4) To further verify the reliability of the algorithm, the roadbed rebound deflection under the loading sequence was calculated based on the obtained test set roadbed parameters and compared with the actual rebound deflection. The results are as follows: Figures 10a to 10d As shown in the figure, it can be seen that the random BP neural network group algorithm has high reliability, and its prediction error is within 5%.

[0135] 5) Assuming that the inversion parameters are the data 1 in Table 3, here we will introduce how to further obtain the equivalent elastic modulus of the roadbed during the operation period. Construct a 4-layer pavement structure with the parameter values ​​shown in Table 5. Apply a half-sine pulse load of 707 kPa with an action radius of 15 cm, an action time of 50 ms, and an action magnitude of 15 cm to the surface to obtain its peak deflection. The finite element model of the pavement structure is as follows: Figure 11 The calculated deflection curve is shown as Figure 12 As shown in the figure, the calculated road surface deflection is 154.8 μm.

[0136] Table 5 Pavement structure parameter values

[0137] Structural layer Modulus / MPa Poisson's ratio <![CDATA[Density / t·m -3 > Thickness (m) Surface 2000 0.25 2.4 0.18 grassroots 4000 0.25 2.3 0.35 Subbase 3000 0.30 2.0 0.20 soil base k1=2.14; k2=0.28; k3=-1.99 0.42 2.14 3.00

[0138] 6) In addition, a finite element model of a four-layer pavement structure was constructed. The parameters of the surface layer, base layer, and subbase layer were the same as those in Table 5. The roadbed was considered as a linear elastic structure with moduli of 20, 50, 75, 100, 150, 200, and 300 MPa. The deflection at the load center was calculated for different roadbed moduli and plotted as shown in the following figure. Figure 13 The subgrade modulus-surface deflection curve shown in Figure 5 is shown. Combined with the 154.8 μm obtained in step 5), the subgrade equivalent modulus for this pavement structure is 193 MPa. If traditional linear elasticity were used for calculation, the subgrade modulus would be 100 MPa. This difference suggests that the subgrade modulus during construction is 100 MPa, while the modulus during operation is 193 MPa, a significant difference. Therefore, it is necessary to consider the influence of the pavement structure when calculating the subgrade modulus during operation.

[0139] The above random BP neural network group algorithm can be pre-embedded in the device processing module; the loading sequence can be built into the device loading module; the preliminary preparation, load application, data collection, and data analysis of the entire test process are all controlled and completed automatically by the device control module.

[0140] The embodiment of the present invention collects the nonlinear characteristics of the roadbed and inverts the nonlinear parameters of the roadbed through an on-site loading sequence. The purpose of the loading sequence is to collect the nonlinear characteristics of the roadbed soil, which can greatly increase the inversion accuracy of the nonlinear parameters. During the test, the road surface constraint is simulated by the outer ring static load, and the load is applied in the form of a combination of the inner ring and the outer ring to maximize the simulation of the actual situation. Furthermore, in combination with the demand for on-site portability, the device of the prior art (CN 110749518 A) is improved. Due to the reduction in the size of the load plate and the load level, the road surface constraint load (outer ring static load) and the driving load (inner ring dynamic load) are not applied enough, which will result in a large test error. The embodiment of the present invention adopts a new on-site testing method based on the loading sequence, applies a machine learning algorithm, and proposes a fast calculation method in combination with the nonlinear finite element calculation method, eliminating the influence of size effect and load effect.

[0141] Whether it is PFWD or FWD, its load is achieved through free-falling impact load, and the impact load itself has an amplification effect, but the action time is only 30ms, which is far less than the 200ms under the actual wheel action. Therefore, the test equipment based on the impact load cannot simulate the actual load effect of the roadbed. The embodiment of the present invention uses a cylinder to achieve loading. The embodiment of the present invention focuses on roadbed testing, and the load level is 50kPa, but the size of the loading plate is smaller than that of the previous generation of devices, which leads to differences in test results. Therefore, the present invention focuses on the loading method and calculation method, and eliminates the test error caused by the reduction in the size of the loading plate through theoretical calculation. While ensuring accuracy, it improves the portability of the instrument and makes it more applicable.

[0142] The above description is only a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention are included in the scope of protection of the present invention.

Claims

1. A portable in-situ dynamic rebound modulus test method for roadbed, characterized in that: A portable in-situ dynamic rebound modulus test device for roadbed is used, including: A static load applying device, comprising an annular outer ring loading plate (4), the top of the outer ring loading plate (4) being connected to the telescopic end of a static load cylinder (6), the static load cylinder (6) being mounted on the bottom of a vehicle frame (11) above the outer ring loading plate (4), and an outer ring counterweight device (10) being placed on the vehicle frame (11) as a reaction frame for applying the outer ring load; A dynamic load applying device, wherein the dynamic load applying device comprises a circular loading plate (3), the loading plate (3) is installed in the middle of the outer ring loading plate (4), the outer ring loading plate (4) is flush with the bottom of the loading plate (3), the top middle of the loading plate (3) is connected to the output end of the dynamic load cylinder (7), and the dynamic load cylinder (7) is installed at the bottom of the inner ring counterweight device (9); the inner ring counterweight device (9) is placed on the top of the inner ring reaction frame (8), the inner ring reaction frame (8) is placed above the vehicle frame (11), and the inner ring telescopic legs (2) are symmetrically installed at the bottom of the inner ring reaction frame (8) to provide gravity support for the inner ring reaction frame (8); the inner ring telescopic legs (2) can separate the inner ring reaction frame (8) and the vehicle frame (11) to avoid mutual interference between the inner ring dynamic load and the outer ring static load; A data acquisition device, wherein the data acquisition device is installed at the bottom center of the loading plate (3) and the road surface outside the outer ring loading plate (4), and all sensors are arranged in a straight line; The following steps are involved: S1, determine the on-site test point of the roadbed, requiring the roadbed surface near the point to be flat; place the roadbed in-situ dynamic rebound modulus test device above the test point, apply loads to the loading plate (3) and the outer ring loading plate (4) according to the set loading sequence, and obtain the measured rebound deflection under different loading sequences; S2, based on the measured rebound deflection under each loading sequence collected in S1, combines the machine learning algorithm to invert the nonlinear key parameters of the roadbed; S3, based on the nonlinear key parameters of the roadbed obtained in S2, constructs a nonlinear finite element model of the roadbed including the actual pavement structure and calculates the pavement surface displacement response; according to the deflection equivalence principle, the equivalent elastic modulus under the actual designed pavement structure and vehicle load is obtained to achieve dynamic rebound modulus correction; Said S1 comprises the following steps: S11, placing the roadbed in-situ dynamic rebound modulus test device above the measuring point, adjusting the inner ring telescopic legs (2) until the inner ring reaction frame (8) is completely separated from the vehicle frame (11); then, the loading plate (3) and the outer ring loading plate (4) are moved downward by the static load cylinder (6) and the dynamic load cylinder (7) until they contact the roadbed surface; S12, adjusting the static load cylinder (6) and the dynamic load cylinder (7) to preload a predetermined load, and statically pressing for 1 to 2 minutes to allow the loading plate (3) and the outer ring loading plate (4) to fully contact the roadbed surface and eliminate the plastic deformation of the roadbed; S13, applying the loading sequence step by step; when loading, first adjust the static load cylinder (6) to the predetermined static load, and after the load is stable, apply the semi-sine pulse load through the dynamic load cylinder (7); when each load is applied, calculate the rebound deflection according to formula (4), and determine whether the rebound deflection is stable according to formula (5); after stabilization, calculate the rebound deflection under the load of this level according to formula (6); （4） （5） （6） Where, For the n Level loading i Rebound deflection after secondary loading; Indicates the n Level loading i Peak deflection during secondary loading; Indicates the n Level loading i The stable deflection peak after the first loading is completed; Expressed as n Rebound deflection under level loading, Respectively represent i -1 time and i -2 rebound deflection; S14, after the test is completed, first adjust the dynamic load cylinder (7) to unload the inner ring load; adjust the static load cylinder (6) and the dynamic load cylinder (7) to lift the loading plate (3) and the outer ring loading plate (4); finally, by adjusting the inner ring telescopic support legs (2), the inner ring reaction frame (8) is completely in contact with the vehicle frame (11), and the load mode is changed to the moving mode, and S11 to S14 are repeated at the next measuring point to complete the test.

2. A portable roadbed in-situ dynamic rebound modulus testing method according to claim 1, characterized in that: The S3 includes the following steps: S31, establishing a finite element model based on linear elasticity according to the actual road surface, calculating the road surface deflection under different roadbed moduli, and finally forming a roadbed modulus-road surface deflection curve; S32, calculates the road surface deflection using the finite element calculation method that considers stress nonlinearity. The subgrade parameter values ​​are derived from the subgrade parameter inversion results of S23, and the remaining parameters are the same as those in S31. After obtaining the road surface deflection, the road surface equivalent elastic modulus is obtained using the subgrade modulus-road surface deflection curve obtained in S31.

3. A portable roadbed in-situ dynamic rebound modulus testing method according to claim 1, characterized in that: The diameter of the loading plate (3) is 15-25 cm, the outer diameter of the outer ring loading plate (4) is 30-50 cm, the loading range is 0-3 kN, the loading adjustment level is 0.10 kN / level, the loading load is 40-60 kPa, and the total weight of the test device is 0.6-0.8 tons.

4. A portable roadbed in-situ dynamic rebound modulus testing method according to claim 1, characterized in that: The position where the inner ring counterweight device (9) is placed on the inner ring reaction frame (8) is raised upward to provide sufficient installation space for the dynamic carrier cylinder (7), and the telescopic end of the dynamic carrier cylinder (7) passes through the vehicle frame (11).

5. A portable roadbed in-situ dynamic rebound modulus testing method according to claim 1, characterized in that: The data acquisition device includes a high-precision displacement sensor (15), and the high-precision displacement sensor (15) is installed at the bottom center position of the loading plate (3); the dynamic load cylinder (7) is fixedly connected to the horizontally arranged sensor steel beam (12), and a plurality of high-precision displacement sensors (15) are evenly arranged on the sensor steel beam (12), so that all the high-precision displacement sensors (15) are arranged in a straight line.

6. A portable roadbed in-situ dynamic rebound modulus testing method according to claim 5, characterized in that: A flexible rubber pad (5) is installed on the side of the loading plate (3) and the outer ring loading plate (4) close to the road surface, and the high-precision displacement sensor (15) is tightly pressed against the roadbed surface by a spring (14).

7. A portable roadbed in-situ dynamic rebound modulus testing method according to claim 1, characterized in that: A wheel (1) is mounted on the lower end of the inner ring telescopic leg (2).

Images