A method for controlling the rolling quality of aeolian sand roadbed

Abstract

The present application belongs to the technical field of road engineering, and in particular to a aeolian sand roadbed rolling quality control method. The control method comprises the following steps: S1, layered filling and base treatment, S2, precise watering and standing curing, S3, vibration rolling and process control: a vibration roller driven by front and rear wheels is used to vibrate and roll the filled material after curing, the vibration roller works in a high-frequency low-amplitude mode, and the relationship data between rolling times and compaction degree increase is obtained in real time during the rolling process, and the rolling times are dynamically adjusted based on the relationship data. The aeolian sand roadbed rolling quality control method, through the standing time regulation formula based on the environmental temperature, the precise watering calculation model considering the material state and evaporation compensation, the dynamic prediction model reflecting the compaction attenuation law, and the intelligent rolling termination logic integrating multiple criteria, changes the calculation and decision of the key construction parameters from experience qualitative to model quantitative, and realizes the scientific prediction and dynamic optimization of the rolling times.

Description

Technical Field

[0001] This invention relates to the field of road engineering technology, and in particular to a method for controlling the compaction quality of aeolian sand roadbed. Background Technology

[0002] Aeolian sand, a special type of roadbed filler widely distributed in desert and Gobi regions, is characterized by its loose structure, uniform particle size, poor cohesion, and weak water retention. When using traditional roadbed compaction methods to compact it, the following technical challenges are frequently encountered: The compaction effect of aeolian sand is extremely sensitive to moisture content, with a narrow optimum moisture content range. Traditional manual watering methods are difficult to control evenly, easily leading to localized over-drying or over-wetting. When too dry, the frictional resistance between particles is high, making compaction difficult; when too wet, it easily forms springy soil, exhibiting elastic plasticity and failing to compact properly, severely affecting the bearing capacity and stability of the roadbed. Existing construction methods rely heavily on fixed compaction passes and operator experience, lacking real-time perception and dynamic control of the compaction process. This often leads to two drawbacks: insufficient compaction passes and substandard compaction degree; and blind over-compaction, resulting in energy waste and even damage to the already formed dense structure. The lack of real-time and comprehensive monitoring of the fill material condition, environmental factors, and the operating status of the compaction equipment leads to construction adjustments relying on subjective judgment, and quality control lacks data support. Therefore, the present invention solves the shortcomings of the above-mentioned technical problems. Summary of the Invention

[0003] Based on existing technical problems, this invention proposes a method for controlling the compaction quality of aeolian sand roadbed.

[0004] This invention proposes a method for controlling the compaction quality of aeolian sand roadbed, the method comprising the following steps: S1. Layered filling and foundation treatment: Aeolian sand filler is laid according to the predetermined loose thickness, and edging soil is filled on both sides of the roadbed simultaneously.

[0005] S2. Precision watering and settling: The working surface is divided into grids. Based on the initial moisture content of the aeolian sand, a fixed amount of water is calculated and applied to bring the filler to the optimal moisture content range. Then, it is left to stand for a preset time.

[0006] S3. Vibratory compaction and process control: A vibratory roller with front and rear wheel drive is used to vibrate and compact the matured fill material. The vibratory roller operates in a high-frequency, low-amplitude mode. During the compaction process, the relationship data between the number of compaction passes and the increase in compaction degree is acquired in real time, and the number of compaction passes is dynamically adjusted based on the relationship data.

[0007] S4. Final compaction and sealing: Use a rubber-tired roller to perform 1-2 passes of static compaction on the sand layer after vibratory rolling to eliminate wheel tracks.

[0008] S5. Quality Inspection and Feedback: The compaction degree is determined by the ring cutter method and quickly verified by the nuclear density meter. The deflection value of the top surface of the subgrade is detected by the falling weight deflectometer. The test data is compared with the design standard. If it is not qualified, return to step S3 for additional compaction.

[0009] Preferably, in step S2, the preset settling time... With ambient temperature Determined through the following empirical formula: .

[0010] Through the above technical solution, in order to ensure that aeolian sand filler can obtain sufficient and appropriate maturation time under different ambient temperatures, allowing for uniform moisture penetration to improve its engineering properties, thereby providing a uniform material base for subsequent compaction and improving the first-pass compaction qualification rate, an empirical formula is established to correlate the static curing time with the environment. This enables dynamic and quantitative control of the static curing cycle, overcoming the problem of poor adaptability to environmental changes in the fixed-duration method. Calculation example: When the ambient temperature... hour, Therefore, it is recommended to let it stand for at least 4.5 hours.

[0011] Preferably, in step S2, the calculation of the quantitative water volume is performed using the following formula: ,in: Water replenishment per unit area, unit: .

[0012] and These are dimensionless empirical coefficients.

[0013] This represents the optimal moisture content.

[0014] Both represent the current moisture content and are expressed as decimals.

[0015] This refers to the dry density of the packing material, in units of... .

[0016] This represents the thickness of the compacted layer, in units of... .

[0017] Also referring to ambient temperature, the unit is... .

[0018] To estimate the operation time, the unit is... .

[0019] This is the wind speed correction factor, which is dimensionless.

[0020] This is a simplified term for the latent heat of vaporization of water, in units of... .

[0021] Through the above technical solutions, in order to achieve precise control of the moisture content of aeolian sand filler and keep it stable in the optimal range during compaction, avoiding low compaction efficiency or springy soil caused by improper moisture, a comprehensive calculation model is adopted. This model not only compensates for the basic requirement of adjusting the material to the target moisture content, but also innovatively introduces the estimated compensation for moisture evaporation loss during the operation, thereby realizing the scientific and precise calculation of water replenishment. Springy soil is an engineering disease state: when soil is compacted or subjected to load when the moisture content is too high, it cannot be compacted to the density required by the design, but instead exhibits obvious elasticity, plasticity and rebound phenomena.

[0022] Preferably, in step S3, the number of rolling passes... The dynamic prediction model for the relationship between compaction degree growth and the growth rate is an exponential decay growth model, assuming that three sets of data are available. , , and target compaction degree Its single-pass compaction increment satisfy: Its cumulative compaction model is: ,in: This represents the theoretical maximum single-pass compaction increment, and . Let be the compaction efficiency attenuation coefficient, and ,in: For crushing The cumulative compaction degree after each pass. This refers to the initial compaction degree of the filler layer before the rolling operation begins. In order to complete After compaction, the compaction degree of the filler is immediately measured. In order to complete After compaction, the compaction degree of the filler is immediately measured.

[0023] In order to grasp the compaction growth pattern and predict subsequent effects in real time during the compaction process through the above technical solutions, and thus transform the compaction operation from an extensive mode that relies on fixed number of passes or experience to a precise dynamic control mode based on data feedback, an exponential decay growth model is established. The model parameters can be quickly calibrated using a small amount of detection data in the early stage of construction, so as to achieve a quantitative description and prediction of the compaction process.

[0024] Preferably, in step S3, the real-time acquisition of the relationship data between the number of compaction passes and the increase in compaction degree is achieved through an intelligent compaction detection system integrated on the vibratory roller, and the target compaction degree is predicted using a preliminarily calibrated exponential decay growth model. Total number of rolling passes required : .

[0025] Through the above technical solution, in order to scientifically predict the total number of compaction passes required to achieve the target compaction degree in the early stages of construction, so as to optimize resource allocation, improve construction efficiency and avoid blind compaction, the exponential decay growth model is transformed to derive a formula for predicting the total number of compaction passes. This predicted value can be dynamically updated with new data from the intelligent compaction system, providing a forward-looking decision-making basis for process management.

[0026] Preferably, dynamic decision-making is based on the output of the exponential decay growth model, after obtaining the first... After the first test data, i.e., the total number of tests. Compaction degree It includes at least one of the following judgment logics: Judgment condition 1: If If compaction fails, stop rolling immediately and ensure the compaction meets the standards.

[0027] Judgment condition 2: Calculate the compaction increment from the most recent measurement: ,when Stopping the rolling process indicates that the process has entered an inefficient zone; continuing to roll will yield minimal gains. It represents the minimum acceptable increment in compaction between two adjacent passes during continuous compaction.

[0028] Judgment condition 3: When Stop compaction, indicating that the predicted number of passes has been reached.

[0029] Judgment condition 4: To prevent excessive compaction, a maximum safe number of compaction passes is preset. ,when The rolling process was forcibly stopped to check for problems with the process and materials.

[0030] Through the above technical solution, in order to establish an intelligent and reliable compaction termination judgment mechanism, while ensuring that the compaction quality meets the standards, preventing energy waste and potential structural damage caused by excessive compaction, and being able to identify abnormal working conditions in a timely manner, a comprehensive decision-making system integrating target value, efficiency value, predicted value and safety threshold is constructed by setting logical judgment conditions including stopping when the standard is met, stopping when efficiency is too low, stopping when the predicted number of passes is reached, and stopping when the safety limit is reached. This achieves the coordinated optimization of quality, efficiency and safety.

[0031] Preferably, the method further includes step S0: constructing a machine learning prediction model, wherein, between steps S1 and S2, the material property parameters of the aeolian sand, the current construction parameters, and the environmental parameters are input into the machine learning prediction model to predict the expected compaction degree and final moisture content under the preset compaction scheme, and the water spraying amount and compaction parameters are adjusted based on the prediction results.

[0032] The machine learning prediction model is a gradient boosting decision tree model, and its input feature vector consists of the following three types of parameters: a. Material characteristic parameters: percentage of particles with a particle size ≤ 0.075 mm, initial moisture content, and initial dry density.

[0033] b. Construction process parameters: loose layer thickness, planned number of vibratory compaction passes, and characteristics of the roller model after numerical conversion.

[0034] c. Environmental parameters: ambient temperature, ambient relative humidity, and wind speed.

[0035] Output target vector: the predicted final compaction degree and predicted final moisture content under the corresponding working conditions. Then, the data is cleaned, outliers are removed, missing values ​​are imputed, and all input features are normalized. The data is divided into training set, validation set and test set. The model is trained using the training set data and the mean square error between the predicted value and the true value as the loss function. Hyperparameters are tuned using the validation set to prevent overfitting. Finally, the model performance is evaluated on the test set.

[0036] The decision-making logic is based on the output target vector data of the gradient boosting decision tree model, including: Step 1: Compaction degree compliance judgment: Compare the predicted compaction degree with the design specification requirements. If the predicted value is significantly lower than the requirements, the model will issue an early warning.

[0037] Step 2, Moisture Content Window Judgment: Compare the predicted final moisture content with the dynamic optimal moisture content window. If the predicted value deviates from the window, it indicates that the current watering plan and environmental loss estimation are inaccurate.

[0038] Step 3: Parameter Optimization Suggestion Generation: Based on the prediction results, the model recommends adjustment schemes, including: When the predicted compaction degree is low and the predicted moisture content is low, it is recommended to increase the amount of water sprayed and increase the number of compaction passes. all over.

[0039] If the predicted moisture content is too high, it is recommended to extend the settling time and reduce the current watering volume.

[0040] If the forecast indicates that the existing equipment parameters are not good, it is recommended to adjust the vibration frequency of the road roller.

[0041] Execution and Feedback: Based on the model's recommendations, the construction personnel adjust the water spraying amount in step S2 and the compaction parameters in step S3. After the construction is completed, the actual measured compaction degree and moisture content are used as new data samples and fed back to the database for continuous iterative optimization of the model, forming a closed loop of data collection, model prediction, decision execution, and effect feedback.

[0042] The above technical solution aims to embed intelligent prediction and optimization layers into traditional construction processes, enabling accurate pre-construction prediction of compaction effects and moisture content outcomes. Based on these predictions, specific adjustments to process parameters are provided to improve the first-pass yield and overall intelligence. This is achieved by constructing and applying a gradient boosting decision tree machine learning model. This model takes multiple parameters, including materials, processes, and environment, as input to predict the final compaction degree and final moisture content. Based on the predictions, it performs compaction degree compliance checks, moisture content window checks, and generates parameter optimization suggestions. Finally, through a feedback loop, the model continuously learns and iterates in engineering practice, forming a self-improving intelligent quality control ecosystem.

[0043] The present invention proposes a construction quality monitoring system for a method of controlling the compaction quality of aeolian sand roadbed, comprising: The sensor module includes temperature and humidity sensors and soil pressure sensors embedded in key sections of the roadbed.

[0044] The equipment monitoring module is integrated into the GNSS locator and status sensor on the vibratory roller equipment.

[0045] The data aggregation and processing platform is used to receive and process data from the sensor module and the equipment monitoring module, and to calculate and display the number of rolling passes, trajectory, moisture content distribution and compaction degree estimate in real time.

[0046] The early warning module is used to issue warnings when the compaction degree, moisture content, and rolling parameters deviate from preset thresholds.

[0047] Through the above technical solutions, in order to transform traditional discrete, post-event quality inspection into full-section, real-time, and continuous process monitoring, and to achieve visualized perception of construction status and proactive prevention and control of quality problems, an integrated system consisting of a sensor module, an equipment monitoring module, a data aggregation and processing platform, and an early warning module is constructed. This system can aggregate and fuse multi-source data in real time, generate visualized results such as the number of compaction passes, compaction degree estimation, and moisture content distribution, and automatically issue early warnings when key parameters deviate from preset thresholds.

[0048] Preferably, the surface of the roller of the vibratory roller is provided with a replaceable anti-skid texture module. The anti-skid texture module is composed of several groups of diamond-shaped protrusions. The surface of the protrusions is distributed with a diamond-shaped grid anti-skid texture. The status sensor includes a temperature and vibration integrated sensor. The temperature and vibration integrated sensor and GNSS locator are integrated and installed inside the roller. It is used to collect the position, vibration frequency, amplitude and surface temperature of the roller in real time during the rolling process, and send the data to the data aggregation and processing platform through a wireless transmission module.

[0049] The outer surfaces of both ends of the roller are respectively provided with grooves distributed in a ring. The inner surface of the groove is fixedly connected to the mounting shaft. The outer surface of the mounting shaft is rotatably sleeved with a translation wheel. The height of the convex surface of the translation wheel is consistent with the protrusion height of the protrusion block.

[0050] To significantly improve the initial compaction and kneading efficiency of vibratory rollers on loose aeolian sand fill, simultaneously acquire key parameters that directly reflect the interaction between the wheel and soil, and solve the wear and bumping problems of special textured rollers when transferring to non-operational surfaces, the vibratory roller rollers are designed with three integrated features. First, replaceable anti-skid textured modules composed of diamond-shaped protrusions are installed on the roller surface to enhance its shearing and embedding effect on loose fill. Second, an integrated temperature and vibration sensor and a GNSS locator are built into the roller cavity to achieve synchronous, in-situ, and high-precision acquisition of compaction position, vibration conditions, and wheel surface contact temperature. Third, freely rotating translation wheels are installed at both ends of the roller, with their convex surfaces aligned with the textured modules. This allows the translation wheels to bear the load during equipment transfer, effectively protecting the anti-skid texture and improving equipment passability. Simultaneously, the annularly distributed translation wheels effectively rotate with the rollers, and during rotation, the convex surfaces of the translation wheels compact the loose aeolian sand fill.

[0051] Preferably, the anti-slip texture module is connected to the body of the roller via a quick-integration interface. The quick-integration interface includes a base panel on the surface of the roller. The base panel is a diamond-shaped recessed groove. The inner bottom surface of the base panel has a mounting hole with internal threads and a tapered positioning pin hole. The back of the protrusion is connected to a fixing screw that is threaded to the inner surface of the mounting hole and a positioning pin that mates with the positioning pin hole.

[0052] To achieve rapid disassembly and replacement of the anti-slip texture module on the roller of a road roller, and to ensure its connection reliability and positioning accuracy under strong vibration conditions, thereby reducing maintenance time and improving the adaptability and utilization of construction equipment, a rapid integration interface is designed, which includes a diamond-shaped base panel, a tapered positioning pin, and threaded fasteners. Specifically, a diamond-shaped recessed base panel matching the contour of the module is machined on the roller surface. The tapered positioning pin hole at the bottom and the positioning pin on the back of the module achieve precise radial positioning and shear resistance. The internal threaded mounting hole on the panel cooperates with the fixing screw of the module to provide the main axial clamping force and fastening guarantee, thus forming a modular assembly system that separates positioning and locking, is easy to operate, and has a stable connection.

[0053] Preferably, the equipment monitoring module further includes a detachable multi-sensor integrated bracket. The integrated bracket is installed on the rigid frame of the vibratory roller by bolts and is located in front of the roller. The integrated bracket integrates an infrared moisture content sensor, a millimeter-wave radar thickness gauge, and a high-definition camera. Its sensing direction is pointed towards the roller compaction surface. The multiple sensors are used to simultaneously collect images of the moisture content distribution, compaction layer thickness, and surface smoothness of the filler surface before and after the compaction operation, and perform real-time data fusion and anomaly warning through an edge computing unit.

[0054] Through the above technical solution, in order to obtain information on the moisture content, thickness, and appearance of the material on the working surface in real time before and after compaction, and to achieve advanced perception and early warning of local anomalies, thereby further moving the quality control node forward and changing post-event inspection to pre-event prediction and in-event control, a detachable multi-sensor integrated bracket is installed on the rigid frame of the vibratory roller in front of the rollers. This bracket integrates an infrared moisture content sensor, a millimeter-wave radar thickness gauge, and a high-definition camera. Its sensing direction is pointed towards the compaction working surface, which can simultaneously collect multi-dimensional appearance data before and after operation, and use edge computing units to perform rapid data fusion and anomaly analysis, providing real-time and accurate on-site guidance for watering strategy adjustment and compaction operation.

[0055] The beneficial effects of this invention are as follows: 1. By setting up a complete methodology that includes precise watering and static curing, vibratory compaction and dynamic process control, final compaction sealing, and quality inspection and feedback, and by constructing a construction quality monitoring system that integrates machine learning prediction, intelligent compaction monitoring, and multi-source data fusion, a data-driven, closed-loop optimized aeolian sand subgrade compaction quality control system has been formed. This system overcomes the shortcomings of traditional methods, such as reliance on experience, disconnected processes, and lagging quality control. It achieves intelligent management and control of the entire process from post-inspection to pre-prediction, in-process control, and post-verification, significantly improving the consistency of construction quality and the long-term stability of the subgrade structure.

[0056] 2. By setting a settling time control formula based on ambient temperature, a precise watering calculation model that comprehensively considers material state and evaporation compensation, a dynamic prediction model that reflects the compaction attenuation law, and an intelligent compaction termination logic that integrates multiple criteria, the calculation and decision-making of key construction parameters are transformed from empirical qualitative to model quantitative. This effectively ensures that the moisture content of the filler is always within the optimal window, realizes the scientific prediction and dynamic optimization of the number of compaction passes, eliminates diseases such as springy soil, avoids energy waste and structural damage caused by excessive compaction, and achieves a balance between quality, efficiency and energy saving.

[0057] 3. By implementing a three-in-one integrated design for the vibratory roller roller, a replaceable anti-slip texture module composed of diamond-shaped protrusions is arranged on the roller surface to enhance its shearing and embedding effect on loose fill material. At the same time, a temperature and vibration integrated sensor and a GNSS locator are built into the roller cavity to achieve synchronous, in-situ, and high-precision acquisition of rolling position, vibration conditions, and wheel surface contact temperature. Furthermore, freely rotating translation wheels are set at both ends of the roller, with their convex surfaces aligned with the height of the texture module. This allows the translation wheels to bear the weight and move during equipment relocation, effectively protecting the anti-slip texture and improving equipment passability. Meanwhile, the annularly distributed translation wheels can effectively rotate with the roller, and during rotation, the convex surfaces of the translation wheels compact the loose aeolian sand fill material. Attached Figure Description

[0058] Figure 1 This is a schematic diagram of a method for controlling the compaction quality of aeolian sand roadbed proposed in this invention; Figure 2 This is a flowchart of the intelligent compaction termination decision-making process for a method for controlling the compaction quality of aeolian sand roadbed proposed in this invention. Figure 3 This is a block diagram of a machine learning prediction model system for a method for controlling the compaction quality of aeolian sand roadbed proposed in this invention. Figure 4 This is a block diagram of a construction quality monitoring system for a method for controlling the compaction quality of aeolian sand roadbed proposed in this invention; Figure 5 This is a three-dimensional view of the vibratory roller structure for a method of controlling the compaction quality of aeolian sand roadbed proposed in this invention. Figure 6 This is a three-dimensional view of the roller structure of a method for controlling the compaction quality of aeolian sand roadbed proposed in this invention. Figure 7 This is a three-dimensional view of the base panel structure of a method for controlling the compaction quality of aeolian sand roadbed proposed in this invention. Figure 8 This is a three-dimensional view of the translation wheel structure of a method for controlling the compaction quality of aeolian sand roadbed proposed in this invention; Figure 9This is a three-dimensional view of the protruding block structure in a method for controlling the compaction quality of aeolian sand roadbed proposed in this invention. Figure 10 This is a three-dimensional view of the integrated support structure for a method of controlling the compaction quality of aeolian sand roadbed proposed in this invention.

[0059] In the diagram: 1. Vibratory roller; 2. Roller; 3. Protrusion; 4. Groove; 5. Mounting shaft; 6. Translation wheel; 7. Base panel; 8. Mounting hole; 9. Locating pin hole; 10. Fixing screw; 11. Locating pin; 12. Integrated bracket; 13. Infrared moisture content sensor; 14. Millimeter-wave radar thickness gauge; 15. High-definition camera. Detailed Implementation

[0060] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0061] Reference Figures 1-10 A method for controlling the compaction quality of aeolian sand roadbed, the method comprising the following steps: S1. Layered filling and foundation treatment: Aeolian sand filler is laid according to the predetermined loose thickness, and edging soil is filled on both sides of the roadbed simultaneously.

[0062] S2. Precision watering and settling: The working surface is divided into grids. Based on the initial moisture content of the aeolian sand, a fixed amount of water is calculated and applied to bring the filler to the optimal moisture content range. Then, it is left to stand for a preset time.

[0063] S3. Vibratory compaction and process control: A vibratory roller 1 with front and rear wheel drive is used to vibrate and compact the matured fill material. The vibratory roller 1 operates in a high-frequency, low-amplitude mode. During the compaction process, the relationship data between the number of compaction passes and the increase in compaction degree is acquired in real time, and the number of compaction passes is dynamically adjusted based on the relationship data.

[0064] S4. Final compaction and sealing: Use a rubber-tired roller to perform 1-2 passes of static compaction on the sand layer after vibratory rolling to eliminate wheel tracks.

[0065] S5. Quality Inspection and Feedback: The compaction degree is determined by the ring cutter method and quickly verified by the nuclear density meter. The deflection value of the top surface of the subgrade is detected by the falling weight deflectometer. The test data is compared with the design standard. If it is not qualified, return to step S3 for additional compaction.

[0066] To ensure that aeolian sand filler receives sufficient and appropriate maturation time under different ambient temperatures, allowing for uniform moisture penetration to improve its engineering properties, thereby providing a uniform material base for subsequent compaction and increasing the first-pass compaction qualification rate, a preset settling time is specified in step S2. With ambient temperature Determined through the following empirical formula: By establishing an empirical formula linking the settling time and the environment, dynamic and quantitative control of the settling cycle is achieved, overcoming the problem of poor adaptability to environmental changes in the fixed-duration method. Calculation example: when the ambient temperature... hour, Therefore, it is recommended to let it stand for at least 4.5 hours.

[0067] To achieve precise control of the moisture content of aeolian sand filler, ensuring it remains within the optimal range during compaction and avoiding problems such as low compaction efficiency or springy soil caused by improper moisture content, the quantitative water content in step S2 is calculated using the following formula: ,in: Water replenishment per unit area, unit: .

[0068] and These are dimensionless empirical coefficients.

[0069] This represents the optimal moisture content.

[0070] Both represent the current moisture content and are expressed as decimals.

[0071] This refers to the dry density of the packing material, in units of... .

[0072] This represents the thickness of the compacted layer, in units of... .

[0073] Also referring to ambient temperature, the unit is... .

[0074] To estimate the operation time, the unit is... .

[0075] This is the wind speed correction factor, which is dimensionless.

[0076] This is a simplified term for the latent heat of vaporization of water, in units of... By adopting a comprehensive calculation model, not only is the basic requirement of adjusting the material to the target moisture content compensated, but also the estimated compensation for water evaporation loss during the operation is innovatively introduced, thereby realizing the scientific and precise calculation of water replenishment. Spring soil is an engineering disease state: when the soil is rolled or subjected to load when the moisture content is too high, it cannot be compacted to the density required by the design, but will instead exhibit obvious elasticity, plasticity and rebound phenomena.

[0077] In order to monitor the compaction degree growth pattern in real time and predict subsequent effects during compaction, thereby transforming the compaction operation from an extensive mode relying on a fixed number of passes or experience to a precise dynamic control mode based on data feedback, in step S3, the number of compaction passes... The dynamic prediction model for the relationship between compaction degree growth and the growth rate is an exponential decay growth model, assuming that three sets of data are available. , , and target compaction degree Its single-pass compaction increment satisfy: Its cumulative compaction model is: ,in: This represents the theoretical maximum single-pass compaction increment, and . Let be the compaction efficiency attenuation coefficient, and ,in: For crushing The cumulative compaction degree after each pass. This refers to the initial compaction degree of the filler layer before the rolling operation begins. In order to complete After compaction, the compaction degree of the filler is immediately measured. In order to complete After compaction, the compaction degree of the fill material is detected immediately. By establishing an exponential decay growth model, the model parameters can be quickly calibrated using a small amount of detection data from the early stage of construction, thereby achieving a quantitative description and prediction of the compaction process.

[0078] By establishing a complete methodology that includes precise watering and static curing, vibratory compaction and dynamic process control, final compaction sealing, and quality inspection and feedback, and by constructing a construction quality monitoring system that integrates machine learning prediction, intelligent compaction monitoring, and multi-source data fusion, a data-driven, closed-loop optimized quality control system for aeolian sand subgrade compaction has been formed. This system overcomes the shortcomings of traditional methods, such as reliance on experience, disconnected processes, and lagging quality control. It achieves intelligent management and control throughout the entire process, from post-inspection to pre-prediction, in-process control, and post-verification, significantly improving the consistency of construction quality and the long-term stability of the subgrade structure.

[0079] To scientifically predict the total number of compaction passes required to achieve the target compaction degree in the early stages of construction, thereby optimizing resource allocation, improving construction efficiency, and avoiding blind compaction, step S3 involves acquiring real-time data on the relationship between the number of compaction passes and the increase in compaction degree through an intelligent compaction detection system integrated on the vibratory roller 1. Furthermore, a preliminarily calibrated exponential decay growth model is used to predict the achievement of the target compaction degree. Total number of rolling passes required : By transforming the exponential decay growth model, a formula for predicting the total number of compaction passes is derived. This predicted value can be dynamically updated with new data from the intelligent compaction system, providing a forward-looking decision-making basis for process management.

[0080] To establish an intelligent and reliable compaction termination judgment mechanism that ensures compaction quality meets standards while preventing energy waste and potential structural damage caused by over-compaction, and can promptly identify abnormal working conditions, dynamic decision-making is based on the output results of an exponential decay growth model. After the first test data, i.e., the total number of tests. Compaction degree It includes at least one of the following judgment logics: Judgment condition 1: If If compaction fails, stop rolling immediately and ensure the compaction meets the standards.

[0081] Judgment condition 2: Calculate the compaction increment from the most recent measurement: ,when Stopping the rolling process indicates that the process has entered an inefficient zone; continuing to roll will yield minimal gains. It represents the minimum acceptable increment in compaction between two adjacent passes during continuous compaction.

[0082] Judgment condition 3: When Stop compaction, indicating that the predicted number of passes has been reached.

[0083] Judgment condition 4: To prevent excessive compaction, a maximum safe number of compaction passes is preset. ,when The rolling process is forcibly stopped to check for problems with the process and materials. By setting logical judgment conditions including stopping when the target is met, stopping when the efficiency is too low, stopping when the predicted number of passes is reached, and stopping when the safety limit is reached, a comprehensive decision-making system integrating target value, efficiency value, predicted value and safety threshold is constructed, realizing the coordinated optimization of quality, efficiency and safety.

[0084] To embed intelligent prediction and optimization layers into traditional construction processes, enabling accurate pre-construction prediction of compaction effects and moisture content outcomes, and providing specific suggestions for adjusting process parameters based on the prediction results, thereby improving the first-time construction pass rate and the level of intelligence, the method further includes step S0: constructing a machine learning prediction model. Between steps S1 and S2, the material characteristic parameters of aeolian sand, current construction parameters, and environmental parameters are input into the machine learning prediction model to predict the expected compaction degree and final moisture content under the preset compaction scheme, and the water spraying amount and compaction parameters are adjusted based on the prediction results.

[0085] The machine learning prediction model is a gradient boosting decision tree model, whose input feature vector consists of the following three types of parameters: a. Material characteristic parameters: percentage of particles with a particle size ≤ 0.075 mm, initial moisture content, and initial dry density.

[0086] b. Construction process parameters: loose layer thickness, planned number of vibratory compaction passes, and characteristics of the roller model after numerical conversion.

[0087] c. Environmental parameters: ambient temperature, ambient relative humidity, and wind speed.

[0088] Output target vector: the predicted final compaction degree and predicted final moisture content under the corresponding working conditions. Then, the data is cleaned, outliers are removed, missing values ​​are imputed, and all input features are normalized. The data is divided into training set, validation set and test set. The model is trained using the training set data and the mean square error between the predicted value and the true value as the loss function. Hyperparameters are tuned using the validation set to prevent overfitting. Finally, the model performance is evaluated on the test set.

[0089] The decision-making logic is based on the output target vector data of the gradient boosting decision tree model, including: Step 1: Compaction degree compliance judgment: Compare the predicted compaction degree with the design specification requirements. If the predicted value is significantly lower than the requirements, the model will issue an early warning.

[0090] Step 2, Moisture Content Window Judgment: Compare the predicted final moisture content with the dynamic optimal moisture content window. If the predicted value deviates from the window, it indicates that the current watering plan and environmental loss estimation are inaccurate.

[0091] Step 3: Parameter Optimization Suggestion Generation: Based on the prediction results, the model recommends adjustment schemes, including: When the predicted compaction degree is low and the predicted moisture content is low, it is recommended to increase the amount of water sprayed and increase the number of compaction passes. all over.

[0092] If the predicted moisture content is too high, it is recommended to extend the settling time and reduce the current watering volume.

[0093] If the forecast indicates that the existing equipment parameters are not good, it is recommended to adjust the vibration frequency of the road roller.

[0094] Execution and Feedback: Based on the model's recommendations, construction personnel adjust the water spraying amount in step S2 and the compaction parameters in step S3. After construction is completed, the actual measured compaction degree and moisture content are used as new data samples and fed back to the database for continuous iterative optimization of the model, forming a closed loop of data collection, model prediction, decision execution, and effect feedback. By constructing and applying a gradient boosting decision tree machine learning model, which takes multiple parameters such as materials, processes, and environment as inputs, the model predicts the final compaction degree and final moisture content. Based on the prediction results, it performs compaction degree compliance judgment, moisture content window judgment, and generates parameter optimization suggestions. Finally, through the execution feedback closed loop, the model can continuously learn and iterate in engineering practice, forming a self-improving intelligent quality control ecosystem.

[0095] By establishing a settling time control formula based on ambient temperature, a precise watering calculation model that comprehensively considers material state and evaporation compensation, a dynamic prediction model reflecting compaction attenuation laws, and an intelligent compaction termination logic integrating multiple criteria, the calculation and decision-making of key construction parameters have been transformed from empirical qualitative to model-based quantitative analysis. This effectively ensures that the moisture content of the fill material is always within the optimal window, achieving scientific prediction and dynamic optimization of the number of compaction passes. While eliminating defects such as springy soil, it avoids energy waste and structural damage caused by excessive compaction, achieving a balance between quality, efficiency, and energy saving.

[0096] To transform traditional discrete, post-construction quality inspection into full-section, real-time, and continuous process monitoring, and to achieve visualized perception of construction status and proactive prevention and control of quality problems, this invention proposes a construction quality monitoring system for a wind-blown sand subgrade compaction quality control method, comprising: The sensor module includes temperature and humidity sensors and soil pressure sensors embedded in key sections of the roadbed.

[0097] The equipment monitoring module is integrated with the GNSS locator and status sensor on the vibratory roller 1 equipment.

[0098] The data aggregation and processing platform is used to receive and process data from the sensor module and equipment monitoring module, and to calculate and display the number of compaction passes, trajectory, moisture content distribution and compaction degree estimate in real time.

[0099] The early warning module is used to issue warnings when compaction degree, moisture content, and rolling parameters deviate from preset thresholds. By constructing an integrated system consisting of a sensor module, an equipment monitoring module, a data aggregation and processing platform, and an early warning module, the system can aggregate and fuse multi-source data in real time, generate visualized results such as the number of rolling passes, compaction degree estimation, and moisture content distribution, and automatically issue warnings when key parameters deviate from preset thresholds.

[0100] To significantly improve the initial compaction and kneading efficiency of the vibratory roller 1 on loose aeolian sand fill, simultaneously acquire key parameters that directly reflect the interaction state between the roller and the soil, and solve the wear and bumping problems of the special textured roller 2 when transferring to non-operational surfaces, the surface of the roller 2 of the vibratory roller 1 is equipped with a replaceable anti-skid texture module. The anti-skid texture module consists of several groups of diamond-shaped protrusions 3, and the surface of the protrusions 3 is distributed with a diamond-shaped grid anti-skid texture. The status sensor includes a temperature and vibration integrated sensor. The temperature and vibration integrated sensor and GNSS locator are integrated and installed inside the roller 2 to collect the position, vibration frequency, amplitude and surface temperature of the roller 2 in real time during the compaction process, and send the data to the data aggregation and processing platform through a wireless transmission module.

[0101] The outer surfaces of both ends of the roller 2 are respectively provided with grooves 4 arranged in a ring. The inner surface of the grooves 4 is fixedly connected to the mounting shaft 5. The outer surface of the mounting shaft 5 is rotatably fitted with a translation wheel 6. The height of the convex surface of the translation wheel 6 is consistent with the protrusion height of the protrusion block 3. By implementing a three-in-one integrated design for the roller 2 of the vibratory roller 1, namely, arranging a replaceable anti-slip texture module composed of diamond-shaped protrusion blocks 3 on the surface of the roller 2 to enhance its shearing and embedding effect on loose fill, and simultaneously integrating a temperature and vibration sensor with GNSS positioning. The device is built into the roller 2 cavity to achieve synchronous, in-situ, and high-precision acquisition of the rolling position, vibration conditions, and wheel surface contact temperature. Moreover, free-rotating translation wheels 6 are set at both ends of the roller 2, so that their convex surfaces are aligned with the height of the texture module. This allows the translation wheels 6 to bear the weight and move when the equipment is moved, effectively protecting the anti-slip texture and improving the equipment's passability. At the same time, the ring-shaped translation wheels 6 can effectively rotate with the roller 2, and during the rotation, the convex surfaces of the translation wheels 6 can compact the loose aeolian sand filler.

[0102] To enable rapid installation, removal, and replacement of the anti-skid texture module on the roller 2 of the road roller, and to ensure its connection reliability and positioning accuracy under strong vibration conditions, thereby reducing maintenance time and improving the adaptability and utilization rate of the construction equipment, the anti-skid texture module is connected to the body of the roller 2 through a quick-integration interface. The quick-integration interface includes a base panel 7 on the surface of the roller 2. The base panel 7 has a diamond-shaped recessed groove, and its inner bottom surface has threaded mounting holes 8 and tapered positioning pin holes 9. The back of the protrusion 3 is connected to a fixing screw 10 that is threaded to the inner surface of the mounting hole 8. The system is connected to a positioning pin 11 that mates with the positioning pin hole 9. By designing a quick-integration interface that includes a diamond-shaped base panel 7, a tapered positioning pin 11, and threaded fasteners, a diamond-shaped recessed base panel 7 that matches the contour of the module is machined on the surface of the roller 2. The tapered positioning pin hole 9 at its bottom and the positioning pin 11 on the back of the module achieve precise radial positioning and shear resistance. The internal thread mounting hole 8 on the panel mates with the fixing screw 10 of the module to provide the main axial clamping force and fastening guarantee, thus forming a modular assembly system that separates positioning and locking, is easy to operate, and has a stable connection.

[0103] To obtain real-time information on the moisture content, thickness, and apparent condition of the material on the work surface before and after compaction, enabling proactive detection and early warning of local anomalies, and thus shifting quality control points further forward, transforming post-event inspection into pre-event prediction and in-event control, the equipment monitoring module also includes a detachable multi-sensor integrated bracket 12. The integrated bracket 12 is bolted to the rigid frame of the vibratory roller 1 and located in front of the roller 2. The integrated bracket 12 integrates an infrared moisture content sensor 13, a millimeter-wave radar thickness gauge 14, and a high-definition camera 15, with its sensing direction pointing towards the compaction surface of the roller 2. Multiple sensors are used to monitor the compaction process. Before and after compaction, images of the moisture content distribution, compaction layer thickness, and surface smoothness of the filler surface are collected simultaneously. Real-time data fusion and anomaly warning are performed through an edge computing unit. A detachable multi-sensor integrated bracket 12 is installed on the rigid frame of the vibratory roller 1 in front of the roller 2. This bracket integrates an infrared moisture content sensor 13, a millimeter-wave radar thickness gauge 14, and a high-definition camera 15. Its sensing direction is pointed towards the compaction surface. It can collect multi-dimensional appearance data simultaneously before and after the operation and perform rapid data fusion and anomaly analysis with the help of the edge computing unit, providing real-time and accurate on-site guidance for watering strategy adjustment and compaction operation.

[0104] By implementing a three-in-one integrated design for the roller 2 of the vibratory roller 1, a replaceable anti-slip texture module composed of diamond-shaped protrusions 3 is arranged on the surface of the roller 2 to enhance its shearing and embedding effect on loose fill. At the same time, a temperature and vibration integrated sensor and a GNSS locator are built into the cavity of the roller 2 to achieve synchronous, in-situ, and high-precision acquisition of the rolling position, vibration conditions, and wheel surface contact temperature. Furthermore, freely rotating translation wheels 6 are set at both ends of the roller 2, with their convex surfaces aligned with the height of the texture module. This allows the translation wheels 6 to bear the weight and move during equipment relocation, effectively protecting the anti-slip texture and improving the equipment's passability. Meanwhile, the annularly distributed translation wheels 6 can effectively rotate with the roller 2, and during the rotation, the convex surfaces of the translation wheels 6 compact the loose aeolian sand fill.

[0105] Working principle: In a specific embodiment of the present invention, the machine learning prediction model starts working before physical construction begins. It receives material properties, planned process and environmental parameters, and predicts the degree of compaction and final moisture content, thereby providing optimization suggestions for watering and preliminary compaction scheme, realizing prediction before construction begins. Construction workers carried out filling, watering, and settling based on prediction suggestions and model calculation results. Subsequently, a special vibratory roller 1 was put into operation. The anti-slip texture module on the surface of its roller 2 directly interacted with the fill material to enhance the shearing and compaction effect. At the same time, the temperature and vibration integrated sensor and GNSS locator integrated inside the roller 2 collected the core status ternary set of position, vibration, and temperature in real time. The infrared moisture content sensor 13, the millimeter-wave radar thickness gauge 14, and the high-definition camera 15 mounted on the multi-sensor integrated bracket 12 at the front of the frame simultaneously scanned the moisture content, thickness, and image information of the working surface. All collected data is wirelessly transmitted to the data aggregation and processing platform. The platform uses the data from the roller 2 sensor to calculate the current compaction degree in real time through an exponential decay growth model and dynamically predicts the total number of passes required to reach the target. The platform integrates the forward-looking data from the infrared moisture content sensor 13, the millimeter-wave radar thickness gauge 14, and the high-definition camera 15 to form a comprehensive perception of the working face status. The early warning module provides real-time alarms for situations such as moisture content deviation and excessively low compaction degree growth based on preset thresholds and model outputs. Based on the above information, the system executes intelligent rolling termination logic, namely stopping when the target is met, stopping when inefficient, stopping when the predicted number of passes is reached, and stopping when the safety limit is reached. It automatically or assists the operator in making decisions to stop rolling or adjust parameters, directly controlling the execution of step S3. After construction, traditional quality inspection results, along with a large amount of process data generated during construction, are fed back into the database as new samples. This data is used to continuously train and optimize machine learning prediction models and compaction dynamic models, making the system increasingly intelligent with use.

[0106] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for controlling the compaction quality of aeolian sand roadbed, characterized in that: The control method includes the following steps: S1. Layered filling and foundation treatment: Aeolian sand filler is laid according to the predetermined loose thickness, and edge soil is filled on both sides of the roadbed simultaneously; S2, Precision watering and settling: The working surface is divided into grids. Based on the initial moisture content of the aeolian sand, a fixed amount of water is calculated and applied to bring the filler to the optimal moisture content range, followed by settling for a preset time. S3. Vibratory compaction and process control: A vibratory roller (1) with front and rear wheel drive is used to vibrate and compact the matured fill material. The vibratory roller (1) operates in a high frequency and low amplitude mode. During the compaction process, the relationship data between the number of compaction passes and the increase in compaction degree is obtained in real time. The number of compaction passes is dynamically adjusted based on the relationship data. S4. Final compaction: Use a rubber-tired roller to perform 1-2 passes of static compaction on the sand layer after vibratory rolling to eliminate wheel tracks; S5. Quality Inspection and Feedback: The compaction degree is determined by the ring cutter method and quickly verified by the nuclear density meter. The deflection value of the top surface of the subgrade is detected by the falling weight deflectometer. The test data is compared with the design standard. If it is not qualified, return to step S3 for additional compaction.

2. The method for controlling the compaction quality of aeolian sand roadbed according to claim 1, characterized in that: In step S2, the preset resting time With ambient temperature Determined through the following empirical formula: .

3. The method for controlling the compaction quality of aeolian sand roadbed according to claim 2, characterized in that: In step S2, the amount of water is calculated using the following formula: ,in: Water replenishment per unit area, unit: ; and These are dimensionless empirical coefficients; The optimal moisture content; Both are current moisture contents, expressed as decimals; This refers to the dry density of the packing material, in units of... ; This represents the thickness of the compacted layer, in units of... ; Also referring to ambient temperature, the unit is... ; To estimate the operation time, the unit is... ; The wind speed correction factor is dimensionless. This is a simplified term for the latent heat of vaporization of water, in units of... .

4. The method for controlling the compaction quality of aeolian sand roadbed according to claim 3, characterized in that: In step S3, the number of rolling passes The dynamic prediction model for the relationship between compaction degree growth and the growth rate is an exponential decay growth model, assuming that three sets of data are available. , , and target compaction degree Its single-pass compaction increment satisfy: Its cumulative compaction model is: ,in: This represents the theoretical maximum single-pass compaction increment, and ; Let be the compaction efficiency attenuation coefficient, and ,in: For crushing Cumulative compaction degree after each pass; The initial compaction degree of the filler layer before the rolling operation begins; In order to complete The compaction degree of the filler was measured immediately after each compaction. In order to complete After compaction, the compaction degree of the filler is immediately measured.

5. The method for controlling the compaction quality of aeolian sand roadbed according to claim 4, characterized in that: In step S3, the real-time acquisition of the relationship data between the number of compaction passes and the increase in compaction degree is achieved through an intelligent compaction detection system integrated on the vibratory roller (1), and the target compaction degree is predicted using a preliminarily calibrated exponential decay growth model. Total number of rolling passes required : .

6. The method for controlling the compaction quality of aeolian sand roadbed according to claim 5, characterized in that: Dynamic decision-making is based on the output of the exponentially decaying growth model, and the result is obtained at the... After the first test data, i.e., the total number of tests. Compaction degree It includes at least one of the following judgment logics: Judgment condition 1: If If the compaction fails, immediately stop rolling and ensure the compaction meets the standards. Judgment condition 2: Calculate the compaction increment from the most recent measurement: ,when Stopping the rolling process indicates that the process has entered an inefficient zone; continuing to roll will yield minimal gains. This indicates the minimum acceptable increment in compaction between two adjacent passes during continuous compaction. Judgment condition 3: When Stop compaction to indicate that the predicted number of passes has been reached; Judgment condition 4: To prevent excessive compaction, a maximum safe number of compaction passes is preset. ,when The rolling process was forcibly stopped to check for problems with the process and materials.

7. The method for controlling the compaction quality of aeolian sand roadbed according to claim 6, characterized in that: It also includes step S0: constructing a machine learning prediction model, wherein, between steps S1 and S2, the material property parameters of aeolian sand, the current construction parameters and environmental parameters are input into the machine learning prediction model to predict the expected compaction degree and final moisture content under the preset compaction scheme, and the water spraying amount and compaction parameters are adjusted based on the prediction results. The machine learning prediction model is a gradient boosting decision tree model, and its input feature vector consists of the following three types of parameters: Material property parameters: percentage of particles with a diameter ≤0.075mm, initial moisture content, and initial dry density; b. Construction process parameters: loose layer thickness, planned number of vibratory compaction passes, and roller model characteristics converted from numerical data; c. Environmental condition parameters: ambient temperature, ambient relative humidity, wind speed; Output target vector: the predicted final compaction degree and predicted final moisture content under the corresponding working conditions. Then, the data is cleaned, outliers are removed, missing values ​​are imputed, and all input features are normalized. The data is divided into training set, validation set and test set. The model is trained using the training set data and the mean square error between the predicted value and the true value as the loss function. Hyperparameters are tuned using the validation set to prevent overfitting. Finally, the model performance is evaluated on the test set. The decision-making logic is based on the output target vector data of the gradient boosting decision tree model, including: Step 1: Compaction Degree Compliance Judgment: Compare the predicted compaction degree with the design specification requirements. If the predicted value is significantly lower than the requirements, the model will issue an early warning. Step 2, Moisture Content Window Judgment: Compare the predicted final moisture content with the dynamic optimal moisture content window. When the predicted value deviates from the window, it indicates that the current watering plan and environmental loss estimation are inaccurate. Step 3: Parameter Optimization Suggestion Generation: Based on the prediction results, the model recommends adjustment schemes, including: When the predicted compaction degree is low and the predicted moisture content is low, it is recommended to increase the amount of water sprayed and increase the number of compaction passes. all over; If the predicted moisture content is too high, it is recommended to extend the settling time and reduce the current watering amount. If the forecast indicates that the existing equipment parameters are unsatisfactory, it is recommended to adjust the vibration frequency of the road roller; Execution and Feedback: Based on the model's recommendations, the construction personnel adjust the water spraying amount in step S2 and the compaction parameters in step S3. After the construction is completed, the actual measured compaction degree and moisture content are used as new data samples and fed back to the database for continuous iterative optimization of the model, forming a closed loop of data collection, model prediction, decision execution, and effect feedback.

8. A construction quality monitoring system for implementing the aeolian sand roadbed compaction quality control method according to any one of claims 1-7, characterized in that, include: The sensor module includes temperature and humidity sensors and soil pressure sensors embedded in key sections of the roadbed. The equipment monitoring module is integrated with the GNSS locator and status sensor on the vibratory roller (1) equipment; The data aggregation and processing platform is used to receive and process data from the sensor module and the equipment monitoring module, and to calculate and display the number of rolling passes, trajectory, moisture content distribution and compaction degree estimate in real time. The early warning module is used to issue warnings when the compaction degree, moisture content, and rolling parameters deviate from preset thresholds.

9. The construction quality monitoring system for a method of controlling the compaction quality of aeolian sand roadbed according to claim 8, characterized in that: The roller (2) of the vibratory roller (1) is provided with a replaceable anti-skid texture module. The anti-skid texture module is composed of several groups of rhomboid protrusions (3). The surface of the protrusions (3) is distributed with a rhomboid grid anti-skid texture. The status sensor includes a temperature and vibration integrated sensor. The temperature and vibration integrated sensor and GNSS locator are integrated and installed inside the roller (2) to collect the position, vibration frequency, amplitude and surface temperature of the roller (2) in real time during the rolling process, and send the data to the data aggregation and processing platform through the wireless transmission module. The outer surfaces of the two ends of the roller (2) are respectively provided with grooves (4) arranged in a ring. The inner surface of the groove (4) is fixedly connected to the mounting shaft (5). The outer surface of the mounting shaft (5) is rotatably sleeved with a translation wheel (6). The height of the convex surface of the translation wheel (6) is consistent with the protrusion height of the protrusion block (3). The anti-slip texture module is connected to the body of the roller (2) through a quick integration interface. The quick integration interface includes a base panel (7) on the surface of the roller (2). The base panel (7) is a diamond-shaped recessed groove. The inner bottom surface of the base panel (7) has a mounting hole (8) with internal threads and a tapered positioning pin hole (9). The back of the protrusion (3) is connected to a fixing screw (10) that is threaded to the inner surface of the mounting hole (8) and a positioning pin (11) that mates with the positioning pin hole (9).

10. The construction quality monitoring system for a method of controlling the compaction quality of aeolian sand roadbed according to claim 9, characterized in that: The equipment monitoring module also includes a detachable multi-sensor integrated bracket (12). The integrated bracket (12) is installed on the rigid frame of the vibratory roller (1) by bolt fixing and is located in front of the roller (2). The integrated bracket (12) integrates an infrared moisture content sensor (13), a millimeter-wave radar thickness gauge (14) and a high-definition camera (15). Its sensing direction is pointed to the roller (2) compaction surface. The multiple sensors are used to simultaneously collect images of the moisture content distribution, compaction layer thickness and surface flatness of the filler surface before and after the compaction operation, and perform real-time data fusion and abnormal warning through the edge computing unit.

Images