Road soil condition detection equipment and method for road engineering

Through multi-agent reinforcement learning and deep reinforcement learning, optimize the sampling path, combined with robotic robots for adaptive sampling and cleaning, use cooling and lubrication modules to reduce soil interference, solve the problem of incomplete soil characteristic analysis in the existing technology, and achieve high efficiency , accurate soil characteristics detection and improvement suggestions are generated.

CN118932968AInactive Publication Date: 2024-11-12牡丹区公路事业发展中心

Patent Information

Application Number
CN202411411762.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-11
Publication Date
2024-11-12
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing road soil condition testing equipment only considers the soil moisture content, fails to fully understand the soil characteristics, and the cleaning process is cumbersome and lacks intelligent and automated processing.

Method used

Multi-agent reinforcement learning is used to optimize the sampling path, combine deep reinforcement learning and robotic manipulators for adaptive sampling and cleaning, use cooling and lubrication modules to reduce interference to soil samples, and display the distribution of soil characteristics through clustering algorithms and visual images.

Benefits of technology

Improve sampling efficiency and coverage, ensure sampling process stability and data accuracy, provide intuitive soil characteristic distribution support, help identify areas that do not meet the needs of highway construction and generate improvement recommendations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118932968A_ABST
    Figure CN118932968A_ABST
Patent Text Reader

Abstract

The invention provides highway soil condition detection equipment and method for highway engineering, and relates to the technical field of highway soil condition detection equipment. A plurality of modules such as a soil sampling module, a path selection module, a self-adaptive sampling module, a cooling and lubricating module, a self-cleaning module and a data analysis module are integrated. The equipment can efficiently collect soil samples on a construction site, optimizes a sampling path through multi-agent reinforcement learning, avoids repeated sampling, and improves the efficiency. The adaptive sampling module combines deep reinforcement learning, optimizes sampling parameters, and ensures that the sampling process is stable. The cooling and lubricating module reduces drill bit temperature and soil interference through a cooling system and a minimum quantity lubricating system. The self-cleaning module uses a vision module and a robot manipulator to automatically remove impurities on the surface of the sampler. And the data analysis module processes the soil data by using a DBSCAN clustering algorithm, generates a visual image and provides visual data support.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of highway soil condition detection equipment, and in particular to a highway soil condition detection equipment and method for highway engineering. Background Art

[0002] Highway engineering refers to the survey, measurement, design, construction, maintenance, and management of highway structures. Highway engineering structures include roadbeds, pavements, bridges, culverts, tunnels, drainage systems, safety protection facilities, greening and traffic monitoring facilities, as well as houses, workshops and other service facilities used for construction, maintenance and monitoring. In the process of highway construction, highway soil sampling and testing is required before paving to detect the moisture content of the soil in order to ensure the quality and stability of highway construction.

[0003] The document with the prior art publication number CN115575611A provides a highway soil condition detection equipment for highway engineering. By driving the reciprocating screw to drive the slider, the turning rod and the scraper to perform reciprocating motion, the accumulated soil is evenly turned over, the consistency of the moisture content detection of soils of different thicknesses is ensured, and the detection accuracy is improved. The diamond design of the turning rod helps to turn the soil smoothly and avoid blockage; the scraper can remove the soil sample attached to the inner wall of the detection box, which is convenient for subsequent cleaning. The piston tube, piston plate, and piston rod can discharge gas and draw air, which effectively prevents the feed port from being blocked and improves the smoothness of the feeding process. The fixed seat and the rotating plate are driven to engage or separate with the positioning groove of the top cover by rotating and lifting the adjustment rod, which simplifies the top cover disassembly process and facilitates the cleaning work inside the detection box.

[0004] Although the existing technology has made some progress in highway soil condition detection, there are still some shortcomings. The existing technology only considers the water content, but does not consider more detailed soil characteristics. The existing technology does not consider the different sampling parameters of different soil environments, lacks intelligent and automated processing in soil characteristic analysis and improvement suggestion generation, and the existing technology is relatively cumbersome when cleaning.

[0005] In view of this, we propose a highway soil condition detection device and method for highway engineering. Summary of the invention

[0006] The present invention provides a highway soil condition detection device and method for highway engineering, and the technical scheme includes the following aspects.

[0007] According to one aspect of an embodiment of the present application, a highway soil condition detection device for highway engineering is provided, and the device includes the following modules.

[0008] The soil sampling module is used to collect soil samples at a sampling point. The soil sampling module includes a drill bit, a sampler, a soil sensor and an equipment sensor. The soil sensor is used to collect soil data. The equipment sensor is used to collect equipment data.

[0009] The path selection module builds a path selection model based on multi-agent reinforcement learning to optimize the sampling path, avoid repeated sampling and maximize coverage.

[0010] Cooling and lubrication module, the drill bit adopts a closed cooling system inside and a micro-lubrication system outside to control the coolant flow rate and lubricant dosage to reduce the impact on soil samples.

[0011] The adaptive sampling module uses deep reinforcement learning to dynamically build a sampling optimization model based on equipment data to optimize the sampler's depth, rotation speed, coolant flow rate, and lubricant dosage.

[0012] The self-cleaning module, combined with the vision module and the robot manipulator, automatically detects the soil and impurities adhering to the surface of the sampler, and the robot manipulator removes the soil and impurities.

[0013] The data analysis module analyzes soil data through a clustering algorithm and displays the analysis results in a two-dimensional contour map or a three-dimensional surface map, wherein the analysis results are the distribution of soil characteristics.

[0014] As an optional solution of the present application, the soil sensor includes a temperature sensor, a soil density meter, and a moisture sensor; the soil data includes temperature, actual density and moisture content; the temperature sensor is installed inside the drill bit to monitor the temperature of the drill bit; the soil density meter is installed on one side of the equipment to obtain the actual density of the soil; the moisture sensor is installed at the front end of the sampler to accurately obtain the moisture content of the soil.

[0015] The equipment sensors include vibration sensors, displacement sensors, rotation speed sensors, flow rate sensors, water level sensors, and rotation speed sensors. The equipment data include vibration data, depth, rotation speed, coolant flow rate, and lubricant dosage of the sampler. The vibration sensor is installed on the outside of the drill bit to monitor the vibration data of the drill bit during the sampling process and identify abnormal vibrations that may be generated during the operation of the drill bit. The displacement sensor is located on the upper side of the drill bit to detect the depth of the rotary head. The rotation speed sensor is arranged on the motor to detect the rotation speed of the drill bit. The flow rate sensor is arranged inside the cooling system to detect the coolant flow rate. The water level sensor is located inside the micro-lubrication system to detect the amount of lubricant.

[0016] As an optional solution of the present application, the path selection module includes a device state and a path action: the device state is represented as: i =[x i ,yi ,z i ,CP,SP]; x i ,y i ,z i is the current position, CP is the current path, and SP is the sampled point information.

[0017] The path action includes selecting a new sampling point and adjusting the path. The path action is represented by a i =[Δx i ,Δy i ,Δz i ,new-SP]. Δx i ,Δy i ,Δz i is the new sampling point location, SP is the new sampling point information. Path optimization rewards the device to choose the best path and avoid repeated sampling, R path =w1PC i -w2RS i Among them, PC i represents the path coverage, RS i represents the number of repeated sampling, w1 and w2 are weight coefficients.

[0018] As an optional solution of this application, the adaptive sampling module; deep reinforcement learning constructs a sampling optimization model including action space, action and reward. The state space includes the equipment data and soil data of the current sampling point. The action includes adjusting the depth and speed of the sampler, controlling the flow rate of the coolant and the release amount of the external lubricant, and the action is represented by a i '=[Δd i ,Δv i ,Δc i ,Δl i ] Rewards include: Sampling quality reward R quality =αSQ i ; Lubricant cost bonus, R cost =-βLC i ; Comprehensive reward R i =R quality +R cost ; Where SQ is the sampling efficiency; LC is the lubrication cost; α and β are constants.

[0019] As an optional solution of the present application, the cooling and lubrication module includes a cooling unit and a lubrication unit; the cooling unit includes a cooling system and a cooling optimization model; the cooling system includes a water inlet, a water outlet and a high-efficiency heat exchanger, the high-efficiency heat exchanger is used to transfer the heat of the coolant to the external cooling system, and the flow rate of the coolant is adjusted by a flow control valve; based on the cooling optimization model, the coolant flow rate is adjusted to optimize the working temperature of the drill bit, and the flow rate control formula is Q=VΔT / R, where Q is the coolant flow rate, V is the flow rate, ΔT is the temperature difference between the coolant and the drill bit, and R is the thermal resistance; the lubrication unit includes a micro-lubrication system and a lubrication optimization model, the micro-lubrication system includes a plurality of micro-nozzles arranged on the outside of the drill bit, and when the soil is dry or the friction increases, the nozzles are used to accurately release the lubricant; based on the lubrication optimization model, the lubricant release amount is adjusted, and the lubricant control formula is L=MΔT / K, where L is the lubricant release amount, M is the friction of the drill bit, ΔT is the temperature difference between the lubricant and the drill bit, and K is the effectiveness coefficient of the lubricant.

[0020] As an optional solution of the present application, a self-cleaning module: the visual module uses a camera to capture images of the sampler surface in real time and pre-processes the images; uses the YOLO algorithm to detect targets, obtains the bounding box of the target area and its category probability, and extracts the image feature vector of the target area for use by the subsequent classification model; the image feature vector includes, f=[Area, Perimeter, MeanIntensity, TextureFeatures,…]; Area is the total number of pixels in the target area; Perimeter is the boundary length of the target area; MeanIntensity is the average brightness, the mean of the pixel brightness in the target area; TextureFeatures is the texture feature; the image feature vector f extracted by YOLO is input into the classification model to determine the most suitable cleaning method, and the classification result y*=argmax j P(y=j|f) is the optimal cleaning strategy. The cleaning effect is monitored in real time through cameras and sensors, and the cleaning strategy is continuously adjusted using feedback data to ensure that all adhesions are removed and the equipment is kept in optimal condition.

[0021] As an optional solution of the present application, the image is preprocessed: the RGB image is converted into a grayscale image to simplify the calculation, I(x,y)=0.299R(x,y)+0.587G(x,y)+0.114B(x,y), where I(x,y) is a grayscale image, R(x,y), G(x,y), and B(x,y) are the pixel values ​​of the RGB channels respectively; Gaussian filtering is used to reduce noise and improve the effect of edge detection; the Canny edge detection algorithm is used to extract the image edge to assist target detection, G=(G x 2 +Gy 2 ) 1 / 2 , where G x and G y is the change in image gradient in the x and y directions.

[0022] As an optional solution of the present application, according to the output result of the classification model, the optimal cleaning method is selected and the specific cleaning action of the robot manipulator is planned; the cleaning action type includes removing slight adhesions by a mechanical vibration device and controlling the vibration device parameters to generate a suitable vibration force; F v =k v Asin(ωt), where F v is the vibration force, k v is the vibration coefficient, A is the amplitude, ω is the angular frequency, and t is the time; scraping, using a scraper or brush to physically scrape off the attached soil, adjusting the pressure and friction of the manipulator, controlling the movement trajectory of the scraper, and ensuring the appropriate scraping force; F s =μ s N, where F s is the scraping force, μ s is the static friction coefficient, N is the positive pressure; airflow purge, use high-pressure airflow to remove attachments, adjust the direction and speed of the airflow nozzle according to the size and location of the target area to generate appropriate airflow force; F g =ρv 2 A / 2, where F g is the airflow force, ρ is the gas density, v is the airflow velocity, and A is the nozzle area.

[0023] As an optional solution of the present application, the data analysis module extracts soil data and standardizes the soil data of each sampling point to ensure that the soil data is on the same scale; the clustering algorithm DBSCAN is applied to cluster the standardized soil data, and DBSCAN can process irregularly shaped clusters and noise points to obtain the category label to which each sampling point belongs; after clustering, the soil data of each sampling point is expanded into a certain area to cover a certain range to display the distribution of soil characteristics; a two-dimensional or three-dimensional visualization image is generated to display the distribution of soil characteristics in the highway engineering area, and the visualization uses different colors or transparencies to represent different soil characteristics.

[0024] As an optional solution of the present application, according to the clustering results, the average distance or maximum distance from the center of each category and the sampling point to the center is calculated; this distance is used as the extended radius r, and each sampling point will be extended into a circular area with a radius of r, r=1 / n∑ i=1 n d(x i ,c); where d(x i,c) is the sampling point x i The distance to its cluster center c, n is the number of sampling points in the same category; interpolation methods are used to smoothly transition between adjacent areas; Kriging interpolation estimates the value of unknown points based on the soil data of the sampling points to generate a more continuous characteristic distribution map, Z(x)=∑ i=1 n λ i Z(x i ), where Z(x) is the soil data of the point x to be estimated, and Z(x i ) is the known sampling point x i Soil data 2, λ i is the weight coefficient, satisfying ∑λ i =1.

[0025] According to one aspect of an embodiment of the present application, a highway soil condition detection method for highway engineering is provided, and the method includes the following steps.

[0026] S1. Use soil sampling modules to collect soil samples at highway construction sites, and soil sensors to collect soil data in real time.

[0027] S2. Optimize the sampling path based on the multi-agent reinforcement learning algorithm to avoid repeated sampling and maximize the coverage, thereby improving sampling efficiency.

[0028] S3. The depth and speed of the sampler are adjusted through the adaptive sampling module to optimize the use strategy of coolant and lubricant in real time to ensure the stability of the sampling process.

[0029] S4. Use cooling and lubrication modules to maintain the temperature of the drill bit and reduce friction, thereby reducing disturbance to the soil sample.

[0030] S5. Use the self-cleaning module to automatically remove impurities on the surface of the sampler and sensor to ensure the accuracy of data collection and the normal operation of the equipment.

[0031] S6. The collected soil data is transmitted to the data analysis module, the soil data is clustered using the DBSCAN algorithm, and the analysis results are displayed through a graph structure and a visual image. The analysis results are the distribution of soil characteristics.

[0032] S7. Based on the analysis results, evaluate the key parameters of the soil and provide a scientific basis for highway construction.

[0033] Compared with the prior art, the present application has the following beneficial effects.

[0034] By optimizing the sampling path using the multi-agent reinforcement learning (MARL) algorithm, the sampling efficiency and coverage are significantly improved. The depth and speed of the sampler are adjusted through the adaptive sampling module to ensure the stability of the sampling process and the accuracy of the data.

[0035] The cooling and lubrication module maintains the temperature of the drill bit and reduces friction, which reduces interference with soil samples. The self-cleaning module ensures the accuracy of data collection and the normal operation of the equipment, avoiding interference of impurities on the sensor.

[0036] The collected soil data is clustered using the DBSCAN algorithm, and the distribution of soil characteristics is displayed through graph structures and visual images. This provides intuitive data support for engineers and decision makers, helps identify areas that do not meet highway construction needs, and plans further improvement and optimization measures.

[0037] By comparing the analysis results with highway engineering design standards, improvement recommendations are generated to ensure that the soil meets the quality and stability requirements for highway construction. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A structural diagram of a highway soil condition detection device for highway engineering provided in one embodiment of the present application.

[0039] Figure 2 A structural diagram of a soil sampling module of a highway soil condition detection equipment for highway engineering provided in one embodiment of the present application.

[0040] Figure 3 A structural diagram of a cooling and lubrication module for highway soil condition detection equipment used in highway engineering is provided as an embodiment of the present application.

[0041] Figure 4 A flow chart of a method for detecting highway soil conditions for highway engineering is provided as an embodiment of the present application. DETAILED DESCRIPTION

[0042] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below in conjunction with the accompanying drawings.

[0043] Please refer to Figure 1 , which shows a structural diagram of a highway soil condition detection device for highway engineering provided by an embodiment of the present application, and the device includes the following modules.

[0044] The soil sampling module is used to collect soil samples at a sampling point. The soil sampling module includes a drill bit, a sampler, a soil sensor and an equipment sensor. The soil sensor is used to collect soil data. The equipment sensor is used to collect equipment data.

[0045] The path selection module builds a path selection model based on multi-agent reinforcement learning to optimize the sampling path, avoid repeated sampling and maximize coverage.

[0046] Cooling and lubrication module, the drill bit adopts a closed cooling system inside and a micro-lubrication system outside to control the coolant flow rate and lubricant dosage to reduce the impact on soil samples.

[0047] The adaptive sampling module uses deep reinforcement learning to dynamically build a sampling optimization model based on equipment data to optimize the sampler's depth, rotation speed, coolant flow rate, and lubricant dosage.

[0048] The self-cleaning module, combined with the vision module and the robot manipulator, automatically detects the soil and impurities adhering to the surface of the sampler, and the robot manipulator removes the soil and impurities.

[0049] The data analysis module analyzes soil data through a clustering algorithm and displays the analysis results in a two-dimensional contour map or a three-dimensional surface map, wherein the analysis results are the distribution of soil characteristics.

[0050] Please refer to Figure 2 , which shows a structural diagram of a soil sampling module of a highway soil condition detection equipment for highway engineering provided by an embodiment of the present application, wherein the soil sensor includes a temperature sensor, a soil density meter, and a moisture sensor; the soil data includes temperature, actual density, and moisture content.

[0051] The temperature sensor is installed inside the drill bit to monitor the temperature of the drill bit; the soil density meter is installed on one side of the equipment to obtain the actual density of the soil; the moisture sensor is installed at the front end of the sampler to accurately obtain the moisture content of the soil.

[0052] The equipment sensors include vibration sensors, displacement sensors, rotation speed sensors, flow rate sensors, water level sensors, and rotation speed sensors. The equipment data include vibration data, depth, rotation speed, coolant flow rate, and lubricant usage of the sampler.

[0053] The vibration sensor is installed on the outside of the drill bit to monitor the vibration data of the drill bit during the sampling process and identify abnormal vibrations that may be generated during the operation of the drill bit; the displacement sensor is located on the upper side of the drill bit to detect the depth of the rotary head; the speed sensor is set on the motor to detect the speed of the drill bit, and the flow rate sensor is set inside the cooling system to detect the flow rate of the coolant; the water level sensor is located inside the micro-lubrication system to detect the amount of lubricant. The sampler is equipped with a sealing cover to prevent the sample from being contaminated or lost during transportation.

[0054] As an optional solution of the present application, the path selection module includes a device state and a path action: the device state is represented as: i =[x i ,y i ,z i ,CP,SP]; x i ,yi ,z i is the current position, CP is the current path, and SP is the sampled point information; the path action includes selecting a new sampling point and adjusting the path, and the path action is represented by a i =[Δx i ,Δy i ,Δz i ,new-SP];Δx i ,Δy i ,Δz i is the location of the new sampling point, SP is the information of the new sampling point; the path optimization rewards the device to select the optimal path and avoid repeated sampling, R path =w1PC i -w2RS i Among them, PC i represents the path coverage, RS i represents the number of repeated sampling, w1 and w2 are weight coefficients.

[0055] The training process is to initialize the policy network and value network of all devices, initialize the graph structure and soil data of sampling points; the devices perform actions, select sampling points and adjust paths, collect environmental feedback, calculate rewards, store experience and use the priority experience replay mechanism for training, and use the policy gradient method (such as MADDPG) to update the policy network and value network.

[0056] As an optional solution of this application, the adaptive sampling module: deep reinforcement learning constructs a sampling optimization model including action space, action and reward; the state space includes the equipment data and soil data of the current sampling point; the action includes adjusting the depth and speed of the sampler, controlling the coolant flow rate and the release amount of the external lubricant, and the action is represented by a i '=[Δd i ,Δv i ,Δc i ,Δl i ]; Rewards include: sampling quality reward R quality =αSQ i ; Lubricant cost bonus, R cost =-βLC i ; Comprehensive reward R i =R quality +R cost ; Where SQ is the sampling efficiency; LC is the lubrication cost; α and β are constants.

[0057] The training process is to initialize the device's policy network and value network; execute actions, adjust sampling depth, rotation speed, control coolant flow rate and the use of external lubricants, collect environmental feedback, and calculate rewards; store experience and conduct training; and use DDPG to update the policy network and value network.

[0058] Please refer to Figure 3 , which shows a structural diagram of a cooling and lubrication module of a highway soil condition detection equipment for highway engineering provided by an embodiment of the present application, wherein the cooling and lubrication module comprises a cooling unit and a lubrication unit, wherein the cooling unit installs a closed cooling system inside the drill bit, and removes the heat generated by the drill bit through the internal coolant cooling system to maintain the working temperature of the drill bit within an ideal range, specifically, a closed cooling system is designed inside the drill bit, comprising a water inlet, a water outlet and a high-efficiency heat exchanger, the high-efficiency heat exchanger is used to transfer the heat of the coolant to an external cooling system (such as a cooling tower or a radiator), and a flow control valve is used to adjust the flow rate of the coolant to adapt to different working conditions, and based on a sampling optimization model, the coolant flow rate is adjusted to optimize the working temperature of the drill bit, and the flow rate control formula is Q=VΔT / R, wherein Q is the coolant flow rate, V is the flow rate, ΔT is the temperature difference between the coolant and the drill bit, and R is the thermal resistance. The lubrication unit sets several micro-nozzles on the outside of the drill bit to evenly spray lubricant, which plays a lubricating role when the drill bit is difficult to drill and sample, while minimizing the impact on the soil sample. Special lubricants with little impact on the soil are selected, such as biodegradable oil. The lubricant should have good lubrication properties and have minimal impact on the physical and chemical properties of the soil sample. In special cases (such as when the soil is dry or friction increases), a nozzle is used to accurately release a small amount of lubricant. Based on the sampling optimization model, the lubricant release amount is adjusted. The lubricant control formula is L=MΔT / K, where L is the lubricant release amount, M is the friction of the drill bit, ΔT is the temperature difference between the lubricant and the drill bit, and K is the effectiveness coefficient of the lubricant. The cooling and lubrication module includes a cooling unit and a lubrication unit; the cooling unit includes a cooling system and a cooling optimization model; the heat generated by the drill bit is removed by an internal coolant cooling system to maintain the working temperature of the drill bit within an ideal range; the cooling system includes a water inlet, a water outlet and a high-efficiency heat exchanger, the high-efficiency heat exchanger is used to transfer the heat of the coolant to the external cooling system, and the flow rate of the coolant is adjusted by a flow control valve.

[0059] Based on the cooling optimization model, the coolant flow rate is adjusted to optimize the working temperature of the drill bit. The flow rate control formula is Q=VΔT / R, where Q is the coolant flow rate, V is the flow rate, ΔT is the temperature difference between the coolant and the drill bit, and R is the thermal resistance.

[0060] The lubrication unit includes a minimal lubrication system and a lubrication optimization model.

[0061] The micro-lubrication system includes a plurality of micro-nozzles arranged outside the drill bit, and the nozzles are used to accurately release lubricant when the soil is dry or friction increases; the lubricant should have good lubrication properties and have minimal effect on the physical and chemical properties of the soil sample.

[0062] Based on the lubrication optimization model, the lubricant release amount is adjusted, and the lubricant control formula is L=MΔT / K, where L is the lubricant release amount, M is the friction of the drill bit, ΔT is the temperature difference between the lubricant and the drill bit, and K is the effectiveness coefficient of the lubricant.

[0063] As an optional solution of the present application, a self-cleaning module: the self-cleaning module automatically detects soil and impurities adhering to the sampler, and achieves self-cleaning through a visual module and a robot manipulator.

[0064] The visual module uses a camera to capture images of the sampler surface in real time and preprocesses the images. The YOLO algorithm is used for target detection to obtain the bounding box of the target area and its category probability, and to extract the image feature vector of the target area for use in the subsequent classification model.

[0065] The image feature vector includes f=[Area,Perimeter,MeanIntensity,TextureFeatures,…]; Area is the total number of pixels in the target area; Perimeter is the boundary length of the target area; MeanIntensity is the average brightness, the mean of the pixel brightness in the target area; TextureFeatures is the texture feature.

[0066] The image feature vector f extracted by YOLO is input into the classification model to determine the most suitable cleaning method. The classification result y*=argmax j P(y=j|f) is the optimal cleaning strategy. The cleaning effect is monitored in real time through cameras and sensors, and the cleaning strategy is continuously adjusted using feedback data to ensure that all adhesions are removed and the equipment is kept in optimal condition.

[0067] Based on the output of the classification model, the optimal cleaning method is selected and the specific cleaning actions of the robot manipulator are planned. The cleaning action types include removing light adhesions through a mechanical vibration device and controlling the vibration device parameters to generate appropriate vibration force. v =k v Asin(ωt), where F v is the vibration force, k v is the vibration coefficient, A is the amplitude, ω is the angular frequency, and t is the time.

[0068] Scraping, use a scraper or brush to physically scrape off the attached soil, adjust the pressure and friction of the manipulator, control the movement trajectory of the scraper, and ensure the scraping force is appropriate. s =μ s N, where F s is the scraping force, μ s is the static friction coefficient and N is the normal pressure.

[0069] Air purge uses high-pressure air to remove attachments. According to the size and location of the target area, the direction and speed of the air nozzle are adjusted to generate appropriate air flow force. g =ρv 2 A / 2, where F g is the airflow force, ρ is the gas density, v is the airflow velocity, and A is the nozzle area.

[0070] The visual module and the robot manipulator realize self-cleaning; the camera is used to capture the image of the sampler surface in real time, and the image is preprocessed, including converting the RGB image into a grayscale image to simplify the calculation, I(x,y)=0.299R(x,y)+0.587G(x,y)+0.114B(x,y), where I(x,y) is a grayscale image, R(x,y), G(x,y), and B(x,y) are the pixel values ​​of the RGB channels respectively; Gaussian filtering reduces noise and improves the effect of edge detection; Canny edge detection and other algorithms are used to extract image edges to assist target detection, G=(G x 2 +G y 2 ) 1 / 2 , where G x and G y It is the change of image gradient in x and y directions. YOLO algorithm is used for target detection. YOLO divides the image into S×S grids, and each grid predicts several bounding boxes and their corresponding category probabilities. Through the YOLO model, the bounding box of the target area and its category probability are obtained, and the feature vectors of these areas are extracted for use by the subsequent classification model. The feature vector includes f=[Area,Perimeter,MeanIntensity,TextureFeatures,…], area Area, the total number of pixels in the target area; perimeter Perimeter, the boundary length of the target area; mean brightness MeanIntensity, the mean value of the pixel brightness in the target area; texture features TextureFeatures, such as contrast, homogeneity, entropy, etc. extracted from the grayscale co-occurrence matrix. The image feature vector f extracted by YOLO is input into the classification model to determine the most suitable cleaning method. The classification model is Softmax classifier, support vector machine, decision tree, random forest, and the classification result y*=argmax j P(y=j|f) is the optimal cleaning strategy. According to the output of the classification model, the optimal cleaning method is selected and the specific cleaning actions of the robot manipulator are planned.

[0071] The cleaning action types include, vibration, removing light adhesions by mechanical vibration device, controlling the vibration device parameters to generate appropriate vibration force, F v =kv Asin(ωt), where F v is the vibration force, k v is the vibration coefficient, A is the amplitude, ω is the angular frequency, and t is the time; scraping, using a scraper or brush to physically scrape off the attached soil, adjusting the pressure and friction of the manipulator, controlling the movement trajectory of the scraper, and ensuring the appropriate scraping force, F s =μ s N, where F s is the scraping force, μ s is the static friction coefficient, N is the positive pressure; airflow purge, use high-pressure airflow to remove attachments, adjust the direction and speed of the airflow nozzle according to the size and location of the target area, and generate the appropriate airflow force, F g =ρv 2 A / 2, where F g is the airflow force, ρ is the gas density, v is the airflow velocity, and A is the nozzle area. The cleaning effect is monitored in real time through cameras and equipment sensors, and the cleaning strategy is continuously adjusted using feedback data to ensure that all adhesions are removed and the equipment is kept in optimal condition.

[0072] As an optional solution of the present application, the data analysis module extracts soil moisture content, density, organic matter content, particle distribution and other characteristics, and divides the sampling points into different categories according to the extracted characteristics through the clustering algorithm DBSCAN. Each category represents an area with similar soil characteristics. Specifically, the soil data of each sampling point is standardized to ensure that the soil data is on the same scale, and Z-score standardization is used, x′=(x-μ) / σ, where x is the original data, μ is the mean of the feature, and σ is the standard deviation of the feature; the clustering algorithm DBSCAN is used to cluster the standardized soil data. DBSCAN can handle irregularly shaped clusters and noise points to obtain the category label to which each sampling point belongs. After clustering is completed, the soil data of each sampling point is expanded into a certain area to cover a certain range, and the point data is converted into surface data to more clearly display the distribution of soil characteristics. Specifically, according to the clustering results, the average distance or maximum distance from the center (center of mass) of each category and the sampling point to the center is calculated. This distance is used as the extended radius r. Each sampling point will be extended into a circular area with a radius of r, r=1 / n∑ i=1 n d(x i ,c), where d(x i ,c) is the sampling point x iThe distance to its cluster center c, n is the number of sampling points in the same category. In order to avoid abrupt changes in adjacent areas, smooth transitions are made between regions. The interpolation method Kriging interpolation is used to smoothly transition between adjacent regions. Kriging interpolation estimates the value of unknown points based on the soil data of the sampling points to generate a more continuous characteristic distribution map, Z(x)=∑ i=1 n λ i Z(x i ), where Z(x) is the soil data of the point x to be estimated, and Z(x i ) is the known sampling point x i Soil data 2, λ i is the weight coefficient, satisfying ∑λ i =1. The regional data obtained by clustering and interpolation expansion is used to generate two-dimensional or three-dimensional visualization images to show the distribution of soil properties in the highway engineering area. Visualization uses different colors or transparency to represent different soil properties or their changing gradients. Specifically, a two-dimensional contour map uses a contour map to display the distribution of soil properties, and the color depth or transparency represents the different levels of soil properties; a three-dimensional surface map uses a three-dimensional surface map to represent the spatial distribution of soil properties, and more intuitively displays the trend of property changes. The final visualization images and data analysis results are used to identify areas that do not meet the needs of highway construction (such as excessive soil moisture content, insufficient density, etc.), provide intuitive data support for engineers and decision makers, and plan further improvement and optimization measures.

[0073] As an optional solution of the present application, the improvement suggestion module proposes soil improvement measures based on the expert knowledge graph and data analysis results. Construct a knowledge graph, define nodes and edges, nodes represent soil characteristics, improvement measures, engineering requirements, etc., and edges represent the relationship and influence between these nodes; knowledge representation, organize expert knowledge into a graph structure, covering soil types, common problems, recommended improvement methods, etc. The specific knowledge graph structure is that the nodes include soil characteristics (water content, density,) improvement measures (adding stabilizers, adjusting moisture, etc.), and engineering requirements (bearing capacity, stability, etc.); edges are relationships connecting nodes, including "influence", "suggestions", and "requirements". The knowledge graph query and reasoning process is to query the relevant information in the knowledge graph according to the analysis results, obtain improvement measures related to the current soil problem, and use the rule-based reasoning engine to derive suitable improvement suggestions from the graph. The reasoning formula is, Suggestion = IF (Condition) THENAction, where Condition is the condition of the soil state and Action is the corresponding improvement measure. The process of generating improvement suggestions is to analyze soil characteristics, compare the actual characteristics of the soil with the engineering design standards, and determine whether the requirements are met. The comparison formula is Deviation = (Xactual -X standard ) / X standard , where Deviation is the deviation, X actual is the actual measured soil property value, X standard is the design standard value. According to the soil deviation and the rules in the knowledge graph, specific improvement suggestions are generated. Considering the difficulty and cost of implementing the improvement measures, the most appropriate improvement plan is recommended. The improvement suggestion formula is Improvement=Optimization(Deviation,Cost,Effectiveness), where Improvement is the improvement suggestion, Deviation is the deviation, Cost is the implementation cost, and Effectiveness is the effect evaluation. According to the generated improvement suggestions, specific soil improvement measures are implemented, such as adding stabilizers, adjusting water content, etc. After implementation, the soil properties are retested to evaluate the improvement effect. Effectiveness after =(X improved -X baseline ) / (X target -X baseline ), among which, Effectiveness after This is the improved effect, X improved is the improved soil property value, X baseline is the soil property value before improvement, X target It is the target soil characteristic value. According to the feedback of implementation effect, the improvement strategy is adjusted, the soil treatment process is optimized, the improvement suggestions and implementation effects are sorted out, and a report is generated for subsequent decision-making.

[0074] Please refer to Figure 4 , which shows a flow chart of a highway soil condition detection method for highway engineering provided by an embodiment of the present application, and the method includes the following steps.

[0075] S1. Use soil sampling modules to collect soil samples at highway construction sites, and sensors to collect soil data in real time.

[0076] S2. Optimize the sampling path based on the multi-agent reinforcement learning (MARL) algorithm to avoid repeated sampling and maximize coverage, thereby improving sampling efficiency.

[0077] S3. The depth and speed of the sampler are adjusted through the adaptive sampling module to optimize the use strategy of coolant and lubricant in real time to ensure the stability of the sampling process.

[0078] S4. Use cooling and lubrication modules to maintain the temperature of the drill bit and reduce friction, thereby reducing disturbance to the soil sample.

[0079] S5. Use the self-cleaning module to automatically remove impurities on the surface of the sampler and sensor to ensure the accuracy of data collection and the normal operation of the equipment.

[0080] S6. The collected soil data is transmitted to the data analysis module, the DBSCAN algorithm is used to cluster the soil data, and the soil characteristic distribution is displayed through graph structure and visual images.

[0081] S7. Based on the cluster analysis results, key soil parameters are evaluated to provide a scientific basis for highway construction.

[0082] S8. Compare the analysis results with the highway engineering design standards to determine whether the soil meets the construction requirements. If not, generate improvement suggestions.

[0083] S9. Treat the soil according to the improvement suggestions to ensure that it meets the quality and stability requirements of highway construction, and organize the test data and analysis reports to provide reference for subsequent highway design, construction and maintenance.

Claims

1. A road soil condition detection device for highway engineering, characterized in that: The device comprises the following modules: A soil sampling module is used to collect soil samples at a sampling point. The soil sampling module includes a drill, a sampler, a soil sensor and an equipment sensor. The soil sensor is used to collect soil data. The equipment sensor is used to collect equipment data. The path selection module builds a path selection model based on multi-agent reinforcement learning to optimize the sampling path, avoid repeated sampling and maximize coverage; Cooling and lubrication module: the drill bit adopts a closed cooling system inside and a micro-lubrication system outside to control the coolant flow rate and lubricant dosage to reduce the impact on soil samples; The adaptive sampling module uses deep reinforcement learning to dynamically build a sampling optimization model based on equipment data to optimize the sampler's depth, rotation speed, coolant flow rate, and lubricant dosage.

2. A highway soil condition detection device for highway engineering according to claim 1, characterized in that: The device also includes the following modules: The self-cleaning module, combined with the vision module and the robot manipulator, automatically detects soil and impurities adhering to the surface of the sampler, and the robot manipulator removes the soil and impurities; The data analysis module analyzes soil data through a clustering algorithm and displays the analysis results in a two-dimensional contour map or a three-dimensional surface map, wherein the analysis results are the distribution of soil characteristics.

3. A highway soil condition detection device for highway engineering according to claim 1, characterized in that: The soil sensor includes a temperature sensor, a soil density meter, and a moisture sensor; the soil data includes temperature, actual density, and moisture content; The temperature sensor is installed inside the drill bit to monitor the temperature of the drill bit; the soil density meter is installed on one side of the equipment to obtain the soil density; the moisture sensor is installed at the front end of the sampler to accurately obtain the moisture content of the soil; The equipment sensors include vibration sensors, displacement sensors, rotation speed sensors, flow rate sensors, water level sensors, and rotation speed sensors. The equipment data include vibration data, depth, rotation speed, coolant flow rate, and lubricant usage of the sampler. The vibration sensor is installed on the outside of the drill bit to monitor the vibration data of the drill bit during the sampling process and identify abnormal vibrations that may be generated during the operation of the drill bit; the displacement sensor is located on the upper side of the drill bit to detect the depth of the rotary head; the speed sensor is set on the motor to detect the drill bit speed, and the flow rate sensor is set inside the cooling system to detect the coolant flow rate; the water level sensor is located inside the micro-lubrication system to detect the amount of lubricant used.

4. A road soil condition detection device for road engineering according to claim 1, characterized in that: The path selection module includes device status and path action: The device status is represented by: i =[x i ,y i ,z i ,CP,SP]; x i ,y i ,z i is the current position, CP is the current path, and SP is the sampled point information; The path action includes selecting a new sampling point and adjusting the path. The path action is represented by a i =[Δx i ,Δy i ,Δz i ,new-SP]; Δx i ,Δy i ,Δz i is the new sampling point position, SP is the new sampling point information; Path optimization rewards devices for choosing the best path and avoiding repeated sampling, R path =w1PC i -w2RS i ; Among them, PC i represents the path coverage, RS i represents the number of repeated sampling, w1 and w2 are weight coefficients.

5. The highway soil condition detection equipment for highway engineering according to claim 1 is characterized in that: Adaptive Sampling Module: Deep reinforcement learning constructs a sampling optimization model including action space, actions, and rewards; The state space includes the equipment data and soil data of the current sampling point; The actions include adjusting the depth and speed of the sampler, controlling the coolant flow rate, and the release amount of the external lubricant. The action is represented by a i '=[Δd i ,Δv i ,Δc i ,Δl i ]; Rewards include: Sampling quality reward R quality =αSQ i ; Lubricant cost bonus, R cost =-βLC i ; Comprehensive reward R i =R quality +R cost ; Where SQ is the sampling efficiency; LC is the lubrication cost; α and β are constants.

6. A road soil condition detection device for road engineering according to claim 1, characterized in that: The cooling and lubrication module includes a cooling unit and a lubrication unit; The cooling unit includes a cooling system and a cooling optimization model; The cooling system comprises a water inlet, a water outlet and a high-efficiency heat exchanger, the high-efficiency heat exchanger is used to transfer the heat of the coolant to the external cooling system, and the flow rate of the coolant is adjusted by a flow control valve; Based on the cooling optimization model, the coolant flow rate is adjusted to optimize the working temperature of the drill bit. The flow rate control formula is Q=VΔT / R, where Q is the coolant flow rate, V is the flow rate, ΔT is the temperature difference between the coolant and the drill bit, and R is the thermal resistance; The lubrication unit includes a micro-lubrication system and a lubrication optimization model. The micro-lubrication system includes a plurality of micro-nozzles disposed outside the drill bit, and the nozzles are used to precisely release lubricant when the soil is dry or friction increases; Based on the lubrication optimization model, the lubricant release amount is adjusted, and the lubricant control formula is L=MΔT / K, where L is the lubricant release amount, M is the friction of the drill bit, ΔT is the temperature difference between the lubricant and the drill bit, and K is the effectiveness coefficient of the lubricant.

7. A road soil condition detection device for road engineering according to claim 2, characterized in that: Self-cleaning module: The visual module uses a camera to capture images of the sampler surface in real time and pre-processes the images; Use the YOLO algorithm to detect targets, obtain the bounding box of the target area and its category probability, and extract the image feature vector of the target area for use in subsequent classification models; The image feature vector includes, f=[Area,Perimeter,MeanIntensity,TextureFeatures,…]; Area is the total number of pixels in the target area; Perimeter is the boundary length of the target area; MeanIntensity is the average brightness, the mean value of the pixel brightness in the target area; TextureFeatures is the texture features; The image feature vector f extracted by YOLO is input into the classification model to determine the most suitable cleaning method. The classification result y*=argmax j P(y=j|f) is the optimal cleaning strategy.

8. A road soil condition detection device for road engineering according to claim 7, characterized in that: Preprocess the image: Convert the RGB image to a grayscale image to simplify the calculation, I(x,y)=0.299R(x,y)+0.587G(x,y)+0.114B(x,y), where I(x,y) is the grayscale image, R(x,y), G(x,y), and B(x,y) are the pixel values ​​of the RGB channels respectively; Use Gaussian filtering to reduce noise; Use the Canny edge detection algorithm to extract image edges and assist target detection. x 2 +G y 2 ) 1 / 2 , where G x and G y is the change in image gradient in the x and y directions.

9. A road soil condition detection device for road engineering according to claim 7, characterized in that: Cleaning strategy: According to the output of the classification model, the optimal cleaning method is selected and the specific cleaning actions of the robot manipulator are planned; The types of cleaning actions include removing lightly adhered matter by means of a mechanical vibration device, controlling the parameters of the vibration device to generate a combined vibration force; F v =k v Asin(ωt), where F v is the vibration force, k v is the vibration coefficient, A is the amplitude, ω is the angular frequency, and t is the time; Scraping, using a scraper or brush to physically scrape off the attached soil, adjusting the pressure and friction of the manipulator to control the movement trajectory of the scraper; F s =μ s N, where F s is the scraping force, μ s is the coefficient of static friction, N is the normal pressure; Air purge uses high-pressure air to remove attachments. The direction and speed of the air nozzle are adjusted according to the size and location of the target area to generate air flow force. F g =ρv 2 A / 2, where F g is the airflow force, ρ is the gas density, v is the airflow velocity, and A is the nozzle area.

10. The highway soil condition detection equipment for highway engineering according to claim 2, characterized in that: Data Analysis Module: The data analysis module extracts soil data and standardizes the soil data of each sampling point to ensure that the soil data are on the same scale; The clustering algorithm DBSCAN is used to cluster the standardized soil data. DBSCAN can handle irregular clusters and noise points and obtain the category label of each sampling point. After clustering, the soil data of each sampling point were expanded into regions to show the distribution of soil characteristics; Generate 2D or 3D visualizations of the distribution of soil properties within the highway project area, using different colors or transparencies to represent different soil properties.

11. A road soil condition detection device for road engineering according to claim 10, characterized in that: According to the clustering results, calculate the center of each category and the average distance or maximum distance from the sampling point to the center; This distance is used as the extended radius r, and each sampling point will be expanded into a circular area with a radius of r, r=1 / n∑ i= 1 n d(x i ,c); Among them, d(x i ,c) is the sampling point x i The distance to its cluster center c, n is the number of sampling points in the same category; Use interpolation methods to smoothly transition between adjacent regions; Kriging interpolation estimates the value of unknown points based on the soil data at the sampling points, generating a more continuous characteristic distribution map, Z(x)=∑ i=1 n λ i Z(x i ); Among them, Z(x) is the soil data of the point x to be estimated, Z(x i ) is the known sampling point x i Soil data 2, λ i is the weight coefficient, satisfying ∑λ i =1.

12. The detection method of a highway soil condition detection device for highway engineering according to claim 1, characterized in that: The method comprises the following steps: S1. Use soil sampling modules to collect soil samples at highway construction sites, and soil sensors to collect soil data in real time; S2, optimize the sampling path based on multi-agent reinforcement learning algorithm, avoid repeated sampling and maximize coverage, and improve sampling efficiency; S3, adjust the depth and speed of the sampler through the adaptive sampling module, optimize the use strategy of coolant and lubricant in real time, and ensure the stability of the sampling process; S4, using cooling and lubrication modules to maintain the temperature of the drill bit and reduce friction, thereby reducing disturbance to the soil sample; S5. Use the self-cleaning module to automatically remove impurities on the surface of the sampler and sensor to ensure the accuracy of data collection and the normal operation of the equipment; S6. The collected soil data is transmitted to the data analysis module, the soil data is clustered using the DBSCAN algorithm, and the analysis results are displayed through a graph structure and a visual image, where the analysis results are the distribution of soil characteristics; S7. Based on the analysis results, evaluate the key parameters of the soil and provide a scientific basis for highway construction.

Citation Information

Patent Citations

  • Road soil condition detection equipment for road engineering

    CN115575611A

  • Intelligent soil identification sampling system and method

    CN116990074A

  • Multi-station automatic hydraulic sliding table drilling and tapping machine and method thereof

    CN117161764A

  • Vehicle-mounted soil sampling operation system and application method

    CN118258636A

  • Soil detection sampling equipment capable of automatically planning path

    CN216594255U

Cited By

  • Punch forming method and system for steel plate part, intelligent terminal and storage medium

    CN119702821A

  • A highway soil condition monitoring and analysis method for highway engineering

    CN119779748A

  • Roadbed compaction quality detection method and system based on multi-dimensional data

    CN120009307A