Mobile rapid detection system and method for roadbed compactness

Abstract

The invention provides a mobile rapid detection system and method for roadbed compactness. The detection system comprises a data acquisition module, a data processing module, a parameter solving module, an intelligent inversion module, a quality control module and a user interaction module, the data acquisition module acquires vibration signals, and the data processing module performs spectral analysis to generate a surface wave frequency dispersion curve; the parameter solving module is used for calculating longitudinal wave and surface wave speeds, dynamic stiffness and other dynamic parameters of a rock-fill body, the intelligent inversion module is used for performing intelligent inversion on the compaction degree of the earth and rock materials, the quality control module is used for evaluating the construction quality of the earth and rock materials and generating a quality report, and the user interaction module is used for providing a user interface. And the collected vibration signals can be efficiently and accurately processed. Through machine learning and a deep learning model, in combination with an optimization algorithm, intelligent inversion is carried out on the compaction degree of the soil and stone materials, and the accuracy and reliability of a detection result are remarkably improved.

Description

Technical Field

[0001] The invention relates to the technical field of road construction quality detection, and in particular to a mobile roadbed compaction rapid detection system. Background Art

[0002] With the rapid development of highway construction, the detection and control of roadbed compaction as one of the important indicators of highway quality is particularly important. Traditional roadbed compaction detection methods mainly rely on static tests, such as nuclear density meter and sand filling method. Although these methods have high accuracy, they are complicated to operate, time-consuming and labor-intensive, and have high requirements on the construction site environment. In recent years, dynamic detection technologies have gradually attracted people's attention, such as surface wave method and free field wave (FWD) method. These methods infer the dynamic parameters of foundation materials by analyzing vibration signals, but in practical applications, there are still problems such as complex data processing and insufficient intuitive results.

[0003] The existing technology has the following technical problems: First, most traditional detection equipment is fixed or handheld, which cannot meet the needs of rapid detection in large-scale construction sections and lacks mobility, resulting in low detection efficiency. Secondly, the existing vibration signal analysis and data processing methods are complex and rely on highly skilled operators. In addition, the detection results are easily affected by environmental noise and the accuracy needs to be improved. In addition, the existing technology has deficiencies in real-time processing and intelligent analysis of data. It is impossible to perform intelligent inversion through machine learning and deep learning algorithms, and it is difficult to quickly and accurately evaluate the compaction degree of soil and stone materials. Finally, the traditional method lacks an effective management and sharing mechanism for detection data and quality reports, and it is difficult for construction personnel to obtain and use detection results and quality control recommendations in a timely manner. Summary of the invention

[0004] In response to the problems existing in the prior art, the present invention provides a mobile roadbed compaction rapid detection system and method. The system can improve the detection efficiency, and through machine learning and deep learning models, combined with optimization algorithms, intelligent inversion of the compaction of soil and stone materials can be performed, thereby improving the accuracy and reliability of the detection results.

[0005] In order to achieve the above technical objectives, the present invention provides a mobile roadbed compaction rapid detection system, the detection system includes a data acquisition module, a data processing module, a parameter acquisition module, an intelligent inversion module, a quality control module and a user interaction module;

[0006] The data acquisition module is used to simultaneously collect vibration signals of the highway subgrade and the field plane construction section, and conduct joint detection using the surface wave method and the free field wave method;

[0007] The data processing module receives the measured vibration signal from the data acquisition module, performs spectrum analysis, and generates a surface wave dispersion curve;

[0008] The parameter obtaining module calculates multiple dynamic parameters of the highway subgrade and the rockfill body of the field plane construction section based on the surface wave dispersion curve, and the dynamic parameters specifically include longitudinal wave velocity, surface wave velocity, dynamic stiffness, damping coefficient and shear modulus;

[0009] The intelligent inversion module uses a machine learning algorithm to establish a mapping relationship between the compaction degree of soil and rock and the dynamic parameters calculated by the parameter acquisition module, and uses the Ridge regression algorithm to train a regression model for predicting the compaction degree of soil and rock; and combines the deep learning model and the optimization algorithm to perform intelligent inversion on the compaction degree of soil and rock to predict the compaction degree of the highway subgrade and the field plane construction section;

[0010] The quality control module evaluates the construction quality of earth and stone materials in the highway subgrade and field plane construction sections based on the compaction prediction results of the intelligent inversion module;

[0011] The user interaction module provides a user interface for construction personnel to view the test results in real time, and supports data storage and export functions.

[0012] A further technical solution of the present invention is as follows: the data acquisition module specifically comprises a control unit and a mobile platform, on which a surface wave sensor, a free surface wave sensor, an excitation device, a signal amplifier, and a data collector are integrated;

[0013] The surface wave sensor is used to be installed on the highway subgrade and field plane construction section to capture the surface vibration signal. The surface wave sensor obtains vibration data by measuring the propagation speed and frequency characteristics of the surface wave;

[0014] The free-field surface wave sensor is used to be installed in a free-field environment to measure the surface wave signal of the foundation material;

[0015] The excitation device is used to generate vibration signals on the highway subgrade and field plane construction section, and excite the ground surface fluctuations through mechanical impact or vibration to ensure that the sensor receives effective vibration signals;

[0016] The signal amplifier is connected to the surface wave sensor and the free surface wave sensor to amplify the collected vibration signal to ensure that the signal strength and quality meet the data processing requirements;

[0017] The data collector is used to convert the amplified analog signal into a digital signal, and to store and transmit the collected vibration data in real time to ensure the integrity and accuracy of the data;

[0018] The control unit is used to control the working states of the surface wave sensor, the free surface wave sensor and the excitation device, coordinate the data collection process, and ensure the synchronous collection and efficient transmission of the vibration signal.

[0019] A further technical solution of the present invention: the data processing module includes a spectrum analysis unit, a dispersion curve generation unit and a data filtering unit;

[0020] The spectrum analysis unit uses the wavelet transform method to perform spectrum analysis on the vibration signal collected by the data acquisition module. The wavelet transform selects time windows of different widths for different frequencies, selects narrow time windows for high-frequency signals, and selects wide time windows for low-frequency signals. The continuous wavelet transform of the known signal x(t) is defined as:

[0021]

[0022] Among them, a is the scaling scale, b is the translation factor, is the mother wavelet function that meets the conditions;

[0023] Its inverse transform is defined as:

[0024]

[0025] in,

[0026] Wavelet transform calculation obtains the time-scale domain result, and the relationship between the scale domain and the center frequency is:

[0027]

[0028] Among them, a is the scale, Δ is the sampling interval, F e is the center frequency of the wavelet, F a is the frequency of the scale. From this, we can get the result of Rayleigh wave in the time-frequency domain;

[0029] The dispersion curve generating unit generates a surface wave dispersion curve using the frequency components and energy distribution characteristics extracted by the spectrum analyzing unit; the data filtering unit performs filtering processing on the vibration signal by applying high-pass and low-pass filters to remove noise and interference signals, thereby ensuring the accuracy and reliability of the data.

[0030] A preferred technical solution of the present invention: the parameter obtaining module specifically includes a speed calculation unit, a stiffness calculation unit and a dynamic parameter calculation unit;

[0031] The velocity calculation unit calculates the longitudinal wave velocity and the surface wave velocity of the rockfill body based on the surface wave dispersion curve, and the specific calculation process is as follows:

[0032] P-wave velocity V p Calculated as:

[0033] Where f is the frequency, R is the dispersion relation in the surface wave dispersion curve;

[0034] Surface wave velocity V s Calculated as:

[0035] Where h is the layer thickness, V p is the longitudinal wave velocity;

[0036] The stiffness calculation unit calculates the dynamic stiffness of the rockfill body based on the velocity calculation result. The specific calculation process is as follows:

[0037] Dynamic stiffness K d Calculated as: K d =ρV p 2 ,

[0038] Where ρ is the density of the rockfill;

[0039] The dynamic parameter calculation unit is used to calculate the damping coefficient and the shear modulus. The specific calculation process is as follows:

[0040] The damping coefficient D is calculated as:

[0041] Among them, ΔE is energy loss and E is stored energy;

[0042] The shear modulus is calculated as: G = ρV s 2 ,

[0043] Among them, V s is the surface wave velocity, and ρ is the density of the rockfill body.

[0044] Further technical solution of the present invention: the intelligent inversion module includes a machine learning unit, a deep learning unit and an optimization algorithm unit, the machine learning unit is used to establish a mapping relationship between the compaction degree of soil and stone materials and the dynamic parameters; the deep learning unit is used to train historical data to improve the inversion accuracy; the optimization algorithm unit is used to optimize the calculation parameters in the inversion process to improve the inversion efficiency and accuracy.

[0045] A further technical solution of the present invention is as follows: the quality control module includes a quality assessment unit, a report generation unit and a construction adjustment suggestion unit; the quality assessment unit receives the compaction prediction results of the highway subgrade and the field plane construction section and the related dynamic parameters from the intelligent inversion module, and the related dynamic parameters include longitudinal wave velocity, surface wave velocity, dynamic stiffness, damping coefficient and shear modulus; the quality index is calculated according to the compaction prediction value of the highway subgrade and the field plane construction section provided by the intelligent inversion module and the preset construction quality standard value; the quality of the construction section is evaluated by the quality index to see whether it meets the standard, and the calculation process of the quality index is as follows:

[0046]

[0047] Among them, the predicted compaction degree is the predicted compaction degree value of the highway subgrade and field plane construction section provided by the intelligent inversion module;

[0048] The standard compaction degree is the preset construction quality standard value, and N is the number of samples;

[0049] The evaluation results of whether the quality of the construction section meets the standards are divided into qualified, unqualified and critical states; when the quality index>0, the evaluation result is qualified; when the quality index=0, the evaluation result is critical; when the quality index<0, the evaluation result is unqualified;

[0050] The report generation unit is used to generate a detailed construction quality report; the construction quality report includes basic information of the construction section, the compaction prediction value of the highway subgrade and field plane construction section provided by the intelligent inversion module and related dynamic parameters, the evaluation results of the quality evaluation unit, problems and risks found during the construction process, statistical analysis and trend analysis;

[0051] The construction adjustment suggestion unit is used to make construction adjustment suggestions according to the quality assessment result.

[0052] The present invention also provides a mobile roadbed compaction rapid detection method, which uses the mobile roadbed compaction rapid detection system to perform detection, and the specific steps are as follows:

[0053] S1. The data acquisition module collects vibration signals of the highway subgrade and the field plane construction section; specifically, the vibration signals are generated on the highway subgrade and the field plane construction section by an excitation device, the surface vibration signals are captured by a surface wave sensor installed on the highway subgrade and the field plane construction section, and the surface wave signals of the foundation material are measured by a free field surface wave sensor installed in a free field environment;

[0054] S2. The data processing module receives the measured vibration signal of the data acquisition module, performs spectrum analysis on the vibration signal collected by the data acquisition module using the wavelet transform method, and generates a surface wave dispersion curve using the analyzed frequency components and energy distribution characteristics;

[0055] S3. The parameter obtaining module calculates the dynamic parameters of the highway subgrade and the rockfill body in the field plane construction section based on the surface wave dispersion curve, wherein the dynamic parameters specifically include longitudinal wave velocity, surface wave velocity, dynamic stiffness, damping coefficient and shear modulus;

[0056] S4. Use machine learning algorithms to establish a mapping relationship between soil-rock compaction and the dynamic parameters calculated in step S3, use the Ridge regression algorithm to train a regression model for predicting soil-rock compaction, and combine the deep learning model and the optimization algorithm to perform intelligent inversion on the soil-rock compaction to predict the compaction of the highway subgrade and field plane construction section;

[0057] S5. Evaluate the construction quality of soil and stone materials based on the prediction results of the intelligent inversion module, and view, store and export them through the user interaction module.

[0058] A further technical solution of the present invention is as follows: In the step S4, a machine learning algorithm is used to establish a mapping relationship between the compaction degree of soil and rock and the dynamic parameters calculated in the step S3; the mapping relationship between the compaction degree of soil and rock y and a set of dynamic parameter characteristics X is based on linear regression, expressed as: y = β0 + β1V p +β2V s +β3K d +β4D+β5G+∈,

[0059] Among them, y is the output compaction degree of soil and rock, β0 is the intercept of the model, β1, β2, β3, β4, β5 are the regression coefficients of the model, indicating the influence of each dynamic parameter on the compaction degree, V p is the longitudinal wave velocity, V s is the surface wave velocity, K d is the dynamic stiffness, D is the damping coefficient, G is the shear modulus, ∈ is the model error term, representing the unexplained random error; the model training is specifically as follows:

[0060] (1) collecting historical data, wherein the historical data includes known compaction degrees and corresponding dynamic parameters, and normalizing the historical data;

[0061] (2) Select key features, including longitudinal wave velocity, surface wave velocity, dynamic stiffness, damping coefficient, and shear modulus;

[0062] (3) Model training: A regression model for predicting soil-rock compaction is trained using the Ridge regression algorithm. The training calculation is:

[0063] Among them, y i is the actual compaction degree, x i is the eigenvector, β is the regression coefficient,

[0064] α is the regularization parameter, N is the number of samples, p is the number of features,

[0065] The training process optimizes the regression model parameters by minimizing the loss function.

[0066] A further technical solution of the present invention is as follows: the deep learning model in step S4 is used to train historical data to improve inversion accuracy, which specifically includes the following steps:

[0067] (1) Neural network architecture design: Design a deep neural network model, including an input layer, multiple hidden layers, and an output layer. Each hidden layer uses a ReLU activation function, and the output layer uses a linear activation function.

[0068] (2) Model training: Use historical data for training and adjust network weights and biases through the back-propagation algorithm to minimize the mean square error loss function;

[0069] (3) Model validation: Use cross-validation methods to evaluate model performance, prevent overfitting, and ensure the generalization ability of deep neural network models;

[0070] The optimization algorithm in step S4 uses a particle swarm optimization algorithm, and adjusts the parameters of the regression model and the deep neural network model through multiple iterations of the particle swarm optimization algorithm to minimize the objective function value; the particle swarm optimization algorithm is an optimization algorithm based on swarm intelligence, which finds the optimal solution by simulating information sharing between individuals in a swarm, and optimizes the calculation parameters in the inversion process through the particle swarm optimization algorithm to improve the inversion efficiency and accuracy.

[0071] A further technical solution of the present invention is as follows: In the step S5, the quality of soil and stone construction is evaluated based on the prediction results of the intelligent inversion module, and the results are viewed, stored and exported through the user interaction module. The specific process is as follows:

[0072] S501 receives the compaction prediction results of the highway subgrade and the field plane construction section and the related dynamic parameters from the intelligent inversion module, and the related dynamic parameters include longitudinal wave velocity, surface wave velocity, dynamic stiffness, damping coefficient and shear modulus;

[0073] S502 calculates the quality index based on the predicted compaction values ​​of the highway subgrade and the field plane construction section provided by the intelligent inversion module and the preset construction quality standard values:

[0074]

[0075] Among them, the predicted compaction degree is the predicted compaction degree value of the highway subgrade and field plane construction section provided by the intelligent inversion module;

[0076] The standard compaction degree is the preset construction quality standard value, and N is the number of samples;

[0077] S503 evaluates whether the quality of the construction section meets the standards through quality indicators, and displays, stores and views the evaluation results through a user interaction module; the evaluation results are divided into qualified, unqualified and critical states; when the quality indicator>0, the evaluation result is qualified; when the quality indicator=0, the evaluation result is critical state; when the quality indicator<0, the evaluation result is unqualified.

[0078] The mobile roadbed compaction rapid detection system of the present invention can be quickly moved and deployed between different construction sections by integrating a mobile device; the system integrates a surface wave sensor, a free surface wave sensor, an excitation device and a data acquisition device into a mobile platform, so that it can flexibly adapt to various construction environments, greatly improving the detection efficiency, solving the problems of poor mobility and low efficiency of traditional fixed or handheld detection equipment, and meeting the rapid detection needs of large-area construction sections.

[0079] The present invention uses spectrum analysis technology and intelligent inversion algorithms to process the collected vibration signals efficiently and accurately; through machine learning and deep learning models, combined with optimization algorithms, the compaction degree of soil and stone materials is intelligently inverted, significantly improving the accuracy and reliability of the test results. The system overcomes the problems of traditional methods that rely on highly skilled operators and are susceptible to environmental noise interference, realizes real-time data processing and intelligent analysis, and provides a scientific basis for construction quality assessment.

[0080] The present invention provides an intuitive user interface, which allows construction personnel to view test results and quality control suggestions in real time, and supports data storage and export functions; the user interface displays test data in charts and graphics, provides detailed quality control suggestions, and supports a variety of interactive operations, making it convenient for construction personnel to obtain and use test information. In addition, the data storage unit and the data export unit ensure the safe storage and convenient sharing of test data and quality reports, realize the effective management and sharing of test data, and improve the quality control efficiency of the construction site. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] Figure 1 A schematic diagram of a system flow of an embodiment of the present invention;

[0082] Figure 2 Schematic diagram of the intelligent inversion module process of an embodiment of the present invention. DETAILED DESCRIPTION

[0083] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0084] like Figure 1-Figure 2 As shown, an embodiment provides a mobile roadbed compaction rapid detection system, including a data acquisition module, a data processing module, a parameter acquisition module, an intelligent inversion module, a quality control module and a user interaction module.

[0085] The data acquisition module is used to collect vibration signals of highway subgrade and field plane construction sections, and adopts surface wave method and free field wave method for joint detection; the data acquisition module specifically includes a control unit and a mobile platform, and the mobile platform is integrated with a surface wave sensor, a free field wave sensor, an excitation device, a signal amplifier, and a data collector. The surface wave sensor is used to be installed on the highway subgrade and field plane construction section to capture the surface vibration signal. The surface wave sensor obtains vibration data by measuring the propagation speed and frequency characteristics of the surface wave; the free field wave sensor is used to be installed in a free field environment to measure the surface wave signal of the foundation material; the excitation device is used to generate vibration signals on the highway subgrade and field plane construction section, and excite the surface fluctuations through mechanical impact or vibration to ensure that the sensor receives effective vibration signals; the signal amplifier is connected to the surface wave sensor and the free field wave sensor, and is used to amplify the collected vibration signal to ensure that the signal strength and quality meet the data processing requirements; the data collector is used to convert the amplified analog signal into a digital signal, and store and transmit the collected vibration data in real time to ensure the integrity and accuracy of the data; the control unit is used to control the working status of the surface wave sensor, the free field wave sensor and the excitation device, coordinate the data collection process, and ensure the synchronous collection and efficient transmission of the vibration signal. Through the collaborative work of the above components, the data acquisition module can efficiently and accurately collect vibration signals of the highway subgrade and field plane construction sections, and use the surface wave method and free field wave method for joint detection, providing high-quality raw data for subsequent data processing and parameter determination.

[0086] The data processing module receives the measured vibration signal from the data acquisition module, performs spectrum analysis, and generates a surface wave dispersion curve; the data processing module includes a spectrum analysis unit, a dispersion curve generation unit, and a data filtering unit.

[0087] The spectrum analysis unit uses the wavelet transform method to perform spectrum analysis on the vibration signal collected by the data acquisition module. The wavelet transform selects time windows of different widths for different frequencies, selects narrow time windows for high-frequency signals, and selects wide time windows for low-frequency signals. The continuous wavelet transform of the known signal x(t) is defined as:

[0088]

[0089] Among them, a is the scaling scale, b is the translation factor, is the mother wavelet function that meets the conditions. Its inverse transform is defined as:

[0090]

[0091] in,

[0092] Wavelet transform calculation obtains the time-scale domain result, and the relationship between the scale domain and the center frequency is:

[0093]

[0094] Among them, a is the scale, Δ is the sampling interval, F e is the center frequency of the wavelet, F a is the frequency of the scale. From this, we can get the result of Rayleigh wave in the time-frequency domain;

[0095] The dispersion curve generating unit generates a surface wave dispersion curve using the frequency components and energy distribution characteristics extracted by the spectrum analyzing unit; the data filtering unit performs filtering processing on the vibration signal by applying high-pass and low-pass filters to remove noise and interference signals, thereby ensuring the accuracy and reliability of the data.

[0096] The parameter obtaining module calculates multiple dynamic parameters of the highway subgrade and the rockfill body of the field plane construction section based on the surface wave dispersion curve. The dynamic parameters specifically include longitudinal wave velocity, surface wave velocity, dynamic stiffness, damping coefficient and shear modulus. The parameter obtaining module specifically includes a velocity calculation unit, a stiffness calculation unit and a dynamic parameter calculation unit.

[0097] The velocity calculation unit calculates the longitudinal wave velocity and the surface wave velocity of the rockfill body based on the surface wave dispersion curve. The stiffness calculation unit calculates the dynamic stiffness of the rockfill body based on the velocity calculation result. The dynamic parameter calculation unit is used to calculate other dynamic parameters related to compaction, including damping coefficient and shear modulus. The specific calculation is as follows:

[0098] P-wave velocity calculation: Based on the frequency and wave velocity data in the surface wave dispersion curve, the P-wave velocity V p Calculated as: Where f is the frequency, R is the dispersion relation in the surface wave dispersion curve;

[0099] Surface wave velocity calculation: Surface wave velocity V s Calculated as: Where h is the layer thickness, V p is the longitudinal wave velocity;

[0100] Dynamic stiffness calculation: based on longitudinal wave velocity V p and density ρ, dynamic stiffness K d Calculated as: K d =ρV p 2 , where ρ is the density of the rockfill;

[0101] Damping coefficient calculation: Based on experimental data and material properties, the damping coefficient D is calculated as: Among them, ΔE is energy loss and E is stored energy;

[0102] Shear modulus calculation: The shear modulus G is based on the surface wave velocity V s and density ρ, calculated as: G = ρV s 2 , where V s is the surface wave velocity, ρ is the density of the rockfill body;

[0103] The parameter acquisition module can extract the longitudinal wave velocity, surface wave velocity, dynamic stiffness and other dynamic parameters of the rockfill from the surface wave dispersion curve. These parameters provide basic data for the intelligent inversion module for further intelligent inversion of soil and rock compaction.

[0104] The intelligent inversion module uses machine learning algorithms and deep learning models, combined with optimization algorithms, to perform intelligent inversion on soil and rock compaction; the intelligent inversion module includes a machine learning unit, a deep learning unit, and an optimization algorithm unit. The machine learning unit is used to establish a mapping relationship between soil and rock compaction and dynamic parameters. The mapping relationship between the soil and rock compaction y and a set of dynamic parameter characteristics X is based on linear regression and is expressed as:

[0105] y=β0+β1V p +β2V s +β3K d +β4D+β5G+∈, where y is the output compaction degree of soil and rock, β0 is the intercept of the model, β1, β2, β3, β4, β5 are the regression coefficients of the model, indicating the influence of each dynamic parameter on the compaction degree, V p is the longitudinal wave velocity, V s is the surface wave velocity, K d is the dynamic stiffness, D is the damping coefficient, G is the shear modulus, ∈ is the model error term, representing the unexplained random error; the model training is specifically as follows:

[0106] Data preprocessing: collect historical data, including known compaction degree and corresponding dynamic parameters (such as longitudinal wave velocity, surface wave velocity, dynamic stiffness, damping coefficient and shear modulus), and normalize the historical data;

[0107] Feature selection: Select key features, including longitudinal wave velocity, surface wave velocity, dynamic stiffness, damping coefficient and shear modulus;

[0108] Model training: Use the Ridge regression algorithm to train a regression model for predicting soil compaction. Ridge regression is a linear regression with L2 regularization that can handle multicollinearity problems and improve the robustness of the model. The calculation is: Among them, y i is the actual compaction degree, x iis the feature vector, β is the regression coefficient, α is the regularization parameter, N is the number of samples, and p is the number of features. The training process optimizes the regression model parameters by minimizing the loss function, which is calculated as:

[0109]

[0110] The deep learning unit is used to train historical data to improve inversion accuracy, specifically including:

[0111] Neural network architecture design: Design a deep neural network model, including an input layer, multiple hidden layers and an output layer. Each hidden layer uses the ReLU activation function, and the output layer uses a linear activation function. The ReLU activation function is calculated as: h i =ReLU(W i ·h i-1 +b i )=max(0,W i ·h i-1 +b i ), where h i is the activation value of the i-th layer, W i and b i are the weight and bias of the i-th layer, h i-1 is the activation value of the previous layer, and the linear activation function is calculated as: o = W o ·h n +b o , where o is the output value of the output layer, W o and b o are the weight and bias of the output layer, respectively, h n is the activation value of the last hidden layer;

[0112] Model training: Use historical data for training, and adjust network weights and biases through the back-propagation algorithm to minimize the mean square error loss function. The back-propagation algorithm adjusts network weights and biases by calculating the gradient of the loss function relative to each weight;

[0113] Forward propagation, calculate the output of the network, calculated as: a (l) =σ(W (l) a (l-1) +b (l) ), where a (l) is the activation value of layer l, W (l) and b (l) are the weight and bias of the lth layer respectively, and σ is the activation function;

[0114] The loss is calculated using the mean squared error loss function, which is calculated as: Among them, y i is the actual value, is the predicted value, N is the number of samples;

[0115] The gradient of the loss with respect to the output layer activation is calculated as: Among them, δ (L) is the error of the output layer;

[0116] By using the chain rule, the error is propagated backwards and calculated as:

[0117] δ (l) =((W (l+1) ) T δ (l+1) )⊙σ′(z (l) ), where z (l) =W (l) a (l-1) +b (l) , σ′ is the derivative of the activation function, ⊙ represents the element-wise product;

[0118] Update weights and biases using gradient descent: W (l) =W (l) -ηδ (l) (a (l-1) ) T , b (l) =b (l) -ηδ (l) , where η is the learning rate;

[0119] Model Validation: Use cross-validation methods to evaluate model performance, prevent overfitting, and ensure the generalization ability of deep neural network models.

[0120] The optimization algorithm unit is used to optimize the calculation parameters in the inversion process to improve the inversion efficiency and accuracy, specifically including:

[0121] Optimization method selection: Use particle swarm optimization algorithm to optimize the calculation parameters in the inversion process to improve the inversion efficiency and accuracy;

[0122] Particle Swarm Optimization Algorithm Description: Particle Swarm Optimization Algorithm is an optimization algorithm based on swarm intelligence. It finds the optimal solution by simulating information sharing between individuals in a swarm. Each individual moves in the solution space and updates its position based on its own and the swarm's experience.

[0123] Objective function definition: Define the objective function in the optimization process, usually the error between the predicted compaction and the actual compaction, calculated as:

[0124] Iterative optimization: Through multiple iterations of the particle swarm optimization algorithm, the parameters of the regression model and the deep neural network model are adjusted to minimize the objective function value.

[0125] Through the above steps, the intelligent inversion module can use the Ridge regression algorithm and deep neural network, combined with the particle swarm optimization algorithm, to perform intelligent inversion of soil and stone compaction, and achieve fast and accurate compaction evaluation.

[0126] The quality control module includes a quality assessment unit, a report generation unit and a construction adjustment suggestion unit. The quality assessment unit is used to assess the construction quality of soil and stone materials based on the results of the intelligent inversion module, specifically including:

[0127] Data receiving: receiving compaction prediction results and related dynamic parameters from the intelligent inversion module, including longitudinal wave velocity, surface wave velocity, dynamic stiffness, damping coefficient and shear modulus;

[0128] Standard comparison: Compare the compaction prediction results provided by the intelligent inversion module with the preset construction quality standards, which can be determined based on industry specifications, project requirements or historical data;

[0129]

[0130] Among them, the predicted compaction is the compaction value provided by the intelligent inversion module, the standard compaction is the construction quality standard, and N is the number of samples;

[0131] Quality assessment: When the predicted value ≥ standard value, that is, the quality index ≥ 0, the quality of the construction section meets the standard; based on the comparison results, evaluate whether the quality of the construction section meets the standard, and the evaluation results can be divided into qualified, unqualified and critical status; when the quality index > 0, the evaluation result is qualified; when the quality index = 0, the evaluation result is critical; when the quality index < 0, it is unqualified; the evaluation standard can be whether the average compaction degree of a certain section exceeds a certain threshold or whether the compaction degree of all sections meets the standard;

[0132] The report generating unit is used to generate a detailed construction quality report, specifically including:

[0133] Report content: Generate a construction quality report, which includes basic information of the construction section, compaction prediction results and dynamic parameters provided by the intelligent inversion module, evaluation results of the quality evaluation unit, problems and risks found during the construction process, statistical analysis and trend analysis;

[0134] Report formats: Reports are generated in different formats, including PDF, Word and Excel, for easy archiving and sharing;

[0135] Data Visualization: The data in the report is visualized, including compaction distribution graphs, quality assessment results charts, and trend analysis graphs, which facilitates quick understanding and analysis.

[0136] The construction adjustment suggestion unit is used to make construction adjustment suggestions based on the quality assessment results, including:

[0137] Problem identification: Based on the quality assessment results, identify problems in the construction process. For example, if the compaction degree of a certain section does not meet the standard, it may be due to the failure of the compaction equipment to fully compact or the nature of the soil and rock itself.

[0138] Adjustment plan: Propose targeted construction adjustment suggestions, including specific operation steps and improvement measures, such as increasing the number of compaction times or extending the compaction time, replacing or adjusting the parameters of the compaction equipment, or improving the physical properties of soil and stone materials (such as adjusting the water content);

[0139] Suggested implementation: Construction adjustment suggestions are operational and executable, and can be fed back to the construction site in a timely manner to facilitate construction personnel to make adjustments;

[0140] Through the above steps, the quality control module can comprehensively evaluate the construction quality of earth and stone materials based on the results of the intelligent inversion module, generate a detailed quality report, and put forward practical and feasible construction adjustment suggestions, thereby ensuring that the construction quality meets the standards and the project proceeds smoothly.

[0141] The user interaction module includes a user interface unit, a data storage unit and a data export unit. The user interface unit provides a user interface for construction personnel to view quality assessment results and construction adjustment suggestions in real time. The data storage unit is used to store test data and construction quality reports. The test data includes vibration signal data, spectrum analysis results, dynamic parameters, compaction prediction results and quality assessment data. The data export unit is used to export the test data and construction quality reports into files of different formats.

[0142] Through the above functions, the user interaction module can provide construction personnel with an intuitive and easy-to-use user interface, display test results and quality control recommendations in real time, and support data storage and export, ensuring the efficiency and accuracy of construction quality management.

[0143] The second embodiment provides a method for rapid detection of compaction degree of a mobile roadbed, which uses the rapid detection system for compaction degree of a mobile roadbed described in the first embodiment for detection, and the specific steps are as follows:

[0144] S1. The data acquisition module collects vibration signals of the highway subgrade and the field plane construction section; specifically, the vibration signals are generated on the highway subgrade and the field plane construction section by an excitation device, the surface vibration signals are captured by a surface wave sensor installed on the highway subgrade and the field plane construction section, and the surface wave signals of the foundation material are measured by a free field surface wave sensor installed in a free field environment;

[0145] S2. The data processing module receives the measured vibration signal of the data acquisition module, performs spectrum analysis on the vibration signal collected by the data acquisition module using the wavelet transform method, and generates a surface wave dispersion curve using the analyzed frequency components and energy distribution characteristics;

[0146] S3. The parameter obtaining module calculates the dynamic parameters of the highway subgrade and the rockfill body in the field plane construction section based on the surface wave dispersion curve, wherein the dynamic parameters specifically include longitudinal wave velocity, surface wave velocity, dynamic stiffness, damping coefficient and shear modulus;

[0147] S4. Use machine learning algorithms to establish a mapping relationship between soil-rock compaction and the dynamic parameters calculated in step S3, use the Ridge regression algorithm to train a regression model for predicting soil-rock compaction, and combine the deep learning model and the optimization algorithm to perform intelligent inversion on the soil-rock compaction to predict the compaction of the highway subgrade and field plane construction section;

[0148] A mapping relationship between the compaction degree of soil and rock and the dynamic parameters calculated in step S3 is established using a machine learning algorithm; the mapping relationship between the compaction degree of soil and rock y and a set of dynamic parameter characteristics X is based on linear regression and is expressed as: y = β0 + β1V p +β2V s +β3K d +β4D+β5G+∈,

[0149] Among them, y is the output compaction degree of soil and rock, β0 is the intercept of the model, β1, β2, β3, β4, β5 are the regression coefficients of the model, indicating the influence of each dynamic parameter on the compaction degree, V p is the longitudinal wave velocity, V s is the surface wave velocity, K d is the dynamic stiffness, D is the damping coefficient, G is the shear modulus, ∈ is the model error term, representing the unexplained random error; the model training is specifically as follows:

[0150] (1) collecting historical data, wherein the historical data includes known compaction degrees and corresponding dynamic parameters, and normalizing the historical data;

[0151] (2) Select key features, including longitudinal wave velocity, surface wave velocity, dynamic stiffness, damping coefficient, and shear modulus;

[0152] (3) Model training: A regression model for predicting soil-rock compaction is trained using the Ridge regression algorithm. The training calculation is:

[0153] Among them, y i is the actual compaction degree, x i is the eigenvector, β is the regression coefficient,

[0154] α is the regularization parameter, N is the number of samples, p is the number of features,

[0155] The training process optimizes the regression model parameters by minimizing the loss function.

[0156] The deep learning model is used to train historical data to improve inversion accuracy, and specifically includes the following steps:

[0157] (1) Neural network architecture design: Design a deep neural network model, including an input layer, multiple hidden layers, and an output layer. Each hidden layer uses a ReLU activation function, and the output layer uses a linear activation function.

[0158] (2) Model training: Use historical data for training and adjust network weights and biases through the back-propagation algorithm to minimize the mean square error loss function;

[0159] (3) Model validation: Use cross-validation methods to evaluate model performance, prevent overfitting, and ensure the generalization ability of deep neural network models;

[0160] The optimization algorithm uses a particle swarm optimization algorithm, and adjusts the parameters of the regression model and the deep neural network model through multiple iterations of the particle swarm optimization algorithm to minimize the objective function value; the particle swarm optimization algorithm is an optimization algorithm based on swarm intelligence, which finds the optimal solution by simulating information sharing between individuals in a swarm, and optimizes the calculation parameters in the inversion process through the particle swarm optimization algorithm to improve the inversion efficiency and accuracy.

[0161] S5. Based on the prediction results of the intelligent inversion module, the construction quality of soil and stone materials is evaluated and viewed, stored and exported through the user interaction module; the specific process is as follows:

[0162] S501 receives the compaction prediction results of the highway subgrade and the field plane construction section and the related dynamic parameters from the intelligent inversion module, and the related dynamic parameters include longitudinal wave velocity, surface wave velocity, dynamic stiffness, damping coefficient and shear modulus;

[0163] S502 calculates the quality index based on the predicted compaction values ​​of the highway subgrade and the field plane construction section provided by the intelligent inversion module and the preset construction quality standard values:

[0164]

[0165] Among them, the predicted compaction degree is the predicted compaction degree value of the highway subgrade and field plane construction section provided by the intelligent inversion module;

[0166] The standard compaction degree is the preset construction quality standard value, and N is the number of samples;

[0167] S503 evaluates whether the quality of the construction section meets the standards through the calculated quality indicators, and displays, stores and views the evaluation results through the user interaction module: the evaluation results can be divided into qualified, unqualified and critical status; when the quality indicator>0, the evaluation result is qualified; when the quality indicator=0, the evaluation result is critical status; when the quality indicator<0, it is unqualified.

[0168] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present invention is limited to these examples. Under the concept of the present invention, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the present invention as described above, which are not provided in detail for the sake of simplicity.

[0169] The present invention is intended to cover all such substitutions, modifications and variations that fall within the broad scope of the claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A mobile roadbed compaction rapid detection system, characterized by: The detection system includes a data acquisition module, a data processing module, a parameter acquisition module, an intelligent inversion module, a quality control module and a user interaction module; The data acquisition module is used to simultaneously collect vibration signals of the highway subgrade and the field plane construction section, and adopts the surface wave method and the free field wave method for joint detection; The data processing module receives the measured vibration signal from the data acquisition module, performs spectrum analysis, and generates a surface wave dispersion curve; The parameter obtaining module calculates multiple dynamic parameters of the highway subgrade and the rockfill body of the field plane construction section based on the surface wave dispersion curve, and the dynamic parameters specifically include longitudinal wave velocity, surface wave velocity, dynamic stiffness, damping coefficient and shear modulus; The intelligent inversion module uses a machine learning algorithm to establish a mapping relationship between the compaction degree of soil and rock and the dynamic parameters calculated by the parameter acquisition module, and uses the Ridge regression algorithm to train a regression model for predicting the compaction degree of soil and rock; and combines the deep learning model and the optimization algorithm to perform intelligent inversion on the compaction degree of soil and rock to predict the compaction degree of the highway subgrade and the field plane construction section; The quality control module evaluates the construction quality of soil and stone materials of the highway subgrade and the field plane construction section based on the compaction prediction results of the highway subgrade and the field plane construction section by the intelligent inversion module; The user interaction module provides a user interface for construction personnel to view the test results in real time, and supports data storage and export functions.

2. A mobile roadbed compaction rapid detection system according to claim 1, characterized in that: The data acquisition module specifically includes a control unit and a mobile platform, on which a surface wave sensor, a free surface wave sensor, an excitation device, a signal amplifier and a data collector are integrated; The surface wave sensor is used to be installed on the roadbed and the field plane construction section to capture the surface vibration signal. The surface wave sensor obtains vibration data by measuring the propagation speed and frequency characteristics of the surface wave; The free-field surface wave sensor is used to be installed in a free-field environment to measure the surface wave signal of the foundation material; The excitation device is used to generate vibration signals on the highway subgrade and field plane construction section, and excite the ground surface fluctuations through mechanical impact or vibration to ensure that the sensor receives effective vibration signals; The signal amplifier is connected to the surface wave sensor and the free surface wave sensor to amplify the collected vibration signal to ensure that the signal strength and quality meet the data processing requirements; The data collector is used to convert the amplified analog signal into a digital signal, and to store and transmit the collected vibration data in real time to ensure the integrity and accuracy of the data; The control unit is used to control the working states of the surface wave sensor, the free surface wave sensor and the excitation device, coordinate the data collection process, and ensure the synchronous collection and efficient transmission of the vibration signal.

3. A mobile roadbed compaction rapid detection system according to claim 1 or 2, characterized in that: The data processing module includes a spectrum analysis unit, a dispersion curve generation unit and a data filtering unit; The spectrum analysis unit uses the wavelet transform method to perform spectrum analysis on the vibration signal collected by the data acquisition module to extract different frequency components. The wavelet transform selects time windows of different widths for different frequency components, selects a narrow time window for high-frequency signals, and selects a wide time window for low-frequency signals; and obtains the time-scale domain result through wavelet transform calculation, and the relationship between the scale domain and the center frequency is: Among them, a is the scale, Δ is the sampling interval, F e is the center frequency of the wavelet, F a is the frequency of the scale; The dispersion curve generating unit generates a surface wave dispersion curve using the frequency components extracted by the spectrum analyzing unit; The data filtering unit performs filtering processing on the vibration signal by applying high-pass and low-pass filters to remove noise and interference signals.

4. A mobile roadbed compaction rapid detection system according to claim 1 or 2, characterized in that: The parameter obtaining module specifically includes a speed calculation unit, a stiffness calculation unit and a dynamic parameter calculation unit; The velocity calculation unit calculates the longitudinal wave velocity and the surface wave velocity of the rockfill body based on the surface wave dispersion curve, and the specific calculation process is as follows: P-wave velocity V p Calculated as: Where f is the frequency, R is the dispersion relation in the surface wave dispersion curve; Surface wave velocity V s Calculated as: Where h is the layer thickness, V p is the longitudinal wave velocity; The stiffness calculation unit calculates the dynamic stiffness of the rockfill body based on the velocity calculation result. The specific calculation process is as follows: Dynamic stiffness K d Calculated as: Where ρ is the density of the rockfill; The dynamic parameter calculation unit is used to calculate the damping coefficient and the shear modulus. The specific calculation process is as follows: The damping coefficient D is calculated as: Among them, ΔE is energy loss and E is stored energy; The shear modulus is calculated as: G = ρV s 2 , Among them, V s is the surface wave velocity, and ρ is the density of the rockfill body.

5. A mobile roadbed compaction rapid detection system according to claim 1 or 2, characterized in that: The intelligent inversion module includes a machine learning unit, a deep learning unit and an optimization algorithm unit. The machine learning unit is used to establish a mapping relationship between soil and stone compaction degree and dynamic parameters; the deep learning unit is used to train historical data to improve inversion accuracy; the optimization algorithm unit is used to optimize the calculation parameters in the inversion process to improve inversion efficiency and accuracy.

6. A mobile roadbed compaction rapid detection system according to claim 1 or 2, characterized in that: The quality control module includes a quality assessment unit, a report generation unit and a construction adjustment suggestion unit; the quality assessment unit receives the compaction prediction results of the highway subgrade and the field plane construction section and the related dynamic parameters from the intelligent inversion module, and the related dynamic parameters include longitudinal wave velocity, surface wave velocity, dynamic stiffness, damping coefficient and shear modulus; the quality index is calculated according to the compaction prediction value of the highway subgrade and the field plane construction section provided by the intelligent inversion module and the preset construction quality standard value; the quality of the construction section is evaluated by the quality index. The calculation process of the quality index is as follows: Among them, the predicted compaction degree is the predicted compaction degree value of the highway subgrade and field plane construction section provided by the intelligent inversion module; The standard compaction degree is the preset construction quality standard value, and N is the number of samples; The evaluation results of whether the quality of the construction section meets the standards are divided into qualified, unqualified and critical states; when the quality index>0, the evaluation result is qualified; when the quality index=0, the evaluation result is critical; when the quality index<0, the evaluation result is unqualified; The report generation unit is used to generate a detailed construction quality report; the construction quality report includes basic information of the construction section, the compaction prediction value of the highway subgrade and field plane construction section provided by the intelligent inversion module and related dynamic parameters, the evaluation results of the quality evaluation unit, problems and risks found during the construction process, statistical analysis and trend analysis; The construction adjustment suggestion unit is used to make construction adjustment suggestions according to the quality assessment result.

7. A mobile roadbed compaction rapid detection method, using the mobile roadbed compaction rapid detection system according to claim 2 for detection, and the specific steps are as follows: S1. The data acquisition module collects vibration signals of the highway subgrade and the field plane construction section; specifically, the vibration signals are generated on the highway subgrade and the field plane construction section by an excitation device, the surface vibration signals are captured by a surface wave sensor installed on the highway subgrade and the field plane construction section, and the surface wave signals of the foundation material are measured by a free field surface wave sensor installed in a free field environment; S2. The data processing module receives the measured vibration signal of the data acquisition module, performs spectrum analysis on the vibration signal collected by the data acquisition module using the wavelet transform method, and generates a surface wave dispersion curve using the analyzed frequency components and energy distribution characteristics; S3. The parameter obtaining module calculates the dynamic parameters of the highway subgrade and the rockfill body in the field plane construction section based on the surface wave dispersion curve, wherein the dynamic parameters specifically include longitudinal wave velocity, surface wave velocity, dynamic stiffness, damping coefficient and shear modulus; S4. Use machine learning algorithms to establish a mapping relationship between soil-rock compaction and the dynamic parameters calculated in step S3, use the Ridge regression algorithm to train a regression model for predicting soil-rock compaction, and combine the deep learning model and the optimization algorithm to perform intelligent inversion on the soil-rock compaction to predict the compaction of the highway subgrade and field plane construction section; S5. Evaluate the construction quality of soil and stone materials based on the prediction results of the intelligent inversion module, and view, store and export them through the user interaction module.

8. A mobile roadbed compaction rapid detection method according to claim 7, characterized in that: In step S4, a machine learning algorithm is used to establish a mapping relationship between the compaction degree of soil and rock and the dynamic parameters calculated in step S3; the mapping relationship between the compaction degree of soil and rock y and a set of dynamic parameter characteristics X is based on linear regression and is expressed as: y = β0 + β1V p +β2V s +β3K d +β4D+β5G+∈, Among them, y is the output compaction degree of soil and rock, β0 is the intercept of the model, β1, β2, β3, β4, β5 are the regression coefficients of the model, indicating the influence of each dynamic parameter on the compaction degree, V p is the longitudinal wave velocity, V s is the surface wave velocity, K d is the dynamic stiffness, D is the damping coefficient, G is the shear modulus, ∈ is the model error term, representing the unexplained random error; the model training is specifically as follows: (1) collecting historical data, wherein the historical data includes known compaction degrees and corresponding dynamic parameters, and normalizing the historical data; (2) Select key features, including longitudinal wave velocity, surface wave velocity, dynamic stiffness, damping coefficient, and shear modulus; (3) Model training: A regression model for predicting soil-rock compaction is trained using the Ridge regression algorithm. The training calculation is: Among them, y i is the actual compaction degree, x i is the eigenvector, β is the regression coefficient, α is the regularization parameter, N is the number of samples, p is the number of features, The training process optimizes the regression model parameters by minimizing the loss function.

9. A mobile roadbed compaction rapid detection method according to claim 7, characterized in that: The deep learning model in step S4 is used to train historical data to improve inversion accuracy, and specifically includes the following steps: (1) Neural network architecture design: Design a deep neural network model, including an input layer, multiple hidden layers, and an output layer. Each hidden layer uses a ReLU activation function, and the output layer uses a linear activation function. (2) Model training: Use historical data for training and adjust network weights and biases through the back-propagation algorithm to minimize the mean square error loss function; (3) Model validation: Use cross-validation methods to evaluate model performance, prevent overfitting, and ensure the generalization ability of deep neural network models; The optimization algorithm in step S4 uses a particle swarm optimization algorithm, and adjusts the parameters of the regression model and the deep neural network model through multiple iterations of the particle swarm optimization algorithm to minimize the objective function value; the particle swarm optimization algorithm is an optimization algorithm based on swarm intelligence, which finds the optimal solution by simulating information sharing between individuals in a swarm, and optimizes the calculation parameters in the inversion process through the particle swarm optimization algorithm to improve the inversion efficiency and accuracy.

10. A mobile roadbed compaction rapid detection method according to claim 7, characterized in that: In step S5, the construction quality of soil and stone materials is evaluated based on the prediction results of the intelligent inversion module, and is viewed, stored and exported through the user interaction module. The specific process is as follows: S501 receives the compaction prediction results of the highway subgrade and the field plane construction section and the related dynamic parameters from the intelligent inversion module, and the related dynamic parameters include longitudinal wave velocity, surface wave velocity, dynamic stiffness, damping coefficient and shear modulus; S502 calculates the quality index based on the predicted compaction values ​​of the highway subgrade and the field plane construction section provided by the intelligent inversion module and the preset construction quality standard values: Among them, the predicted compaction degree is the predicted compaction degree value of the highway subgrade and field plane construction section provided by the intelligent inversion module; The standard compaction degree is the preset construction quality standard value, and N is the number of samples; S503 evaluates whether the quality of the construction section meets the standards through quality indicators, and displays, stores and views the evaluation results through a user interaction module; the evaluation results are divided into qualified, unqualified and critical states; when the quality indicator>0, the evaluation result is qualified; when the quality indicator=0, the evaluation result is critical state; when the quality indicator<0, the evaluation result is unqualified.

Images