Roadbed compaction monitoring system

Abstract

The invention relates to a roadbed compaction monitoring system which comprises a data processing module and a visualization module. The data processing module is used for receiving real-time coordinate data of the road roller and instantaneous vibration signals of a vibration wheel; extracting the frequency domain characteristics of the instantaneous vibration signals by using FFT (Fast Fourier Transform); a rolling track coordinate sequence is generated through the RTK resolving technology; constructing spatio-temporal features based on the coordinate sequence and the frequency domain features; constructing an LSTM neural network model by using the spatial-temporal characteristics and calculating the compactness; and meanwhile, sensitive characteristics of the compaction degree are analyzed and screened by using a grey relational degree, and a compaction degree report is generated. The visualization module is used for generating a compaction track thermodynamic diagram according to the compaction degree report; a color difference-based compactness gradient map is generated by using a color gradation coding technology; and generating a three-dimensional visual compaction map by using a three-dimensional modeling and data matching technology. On the basis, the roadbed compaction monitoring system can realize accurate monitoring and visual display of the roadbed compaction condition.

Description

Technical Field

[0001] The present invention belongs to the field of subgrade compaction, and particularly relates to a subgrade compaction monitoring system. Background Art

[0002] With the large-scale advancement of transportation infrastructure construction, traditional subgrade compaction quality detection generally relies on manual sampling inspection, sand replacement method or nuclear density gauge and other means, resulting in low detection efficiency, long cycle and large destructiveness, and it is difficult to meet the real-time requirements of modern mechanized construction. Although vibration sensing and satellite positioning technologies have been gradually applied to the field of online monitoring in recent years, they are still limited by technical bottlenecks such as insufficient multi-source data fusion, low accuracy of compaction degree calculation model and single visualization representation: single data is difficult to coordinate the coupling relationship between the frequency domain characteristics of vibration signals, the dynamic changes of the temperature field and the spatio-temporal correlation of rolling trajectories. Traditional spectrum analysis is easily affected by noise interference and does not correct the thermal expansion and contraction effect of materials; the prediction of compaction degree still uses static empirical formulas, lacking time-varying coupling modeling of the number of rolling passes, vibration energy attenuation and material property differences; existing visualization technologies are mostly limited to two-dimensional planes or simple heat maps, and cannot stereoscopically present the three-dimensional spatial distribution and dynamic evolution law of compaction degree. Summary of the Invention

[0003] Based on this, it is necessary to provide a subgrade compaction monitoring system that can solve the above problems.

[0004] In a first aspect, the present application provides a subgrade compaction monitoring system, including a data processing module and a visualization module:

[0005] The data processing module is used for:

[0006] Receiving the real-time coordinate data of the roller and the instantaneous vibration signal of the vibrating wheel;

[0007] Extracting the frequency domain characteristics of the instantaneous vibration signal by using FFT transformation;

[0008] Generating a rolling trajectory coordinate sequence by using RTK solution technology based on the coordinate data;

[0009] Constructing spatio-temporal characteristics based on the coordinate sequence and the frequency domain characteristics;

[0010] Constructing an LSTM neural network model by using the spatio-temporal characteristics and calculating the compaction degree;

[0011] Screening the sensitive characteristics of the compaction degree by using grey relational analysis and generating a compaction degree report;

[0012] The visualization module is used for:

[0013] Generating a heat map of the compaction trajectory according to the compaction degree report;

[0014] Based on the compaction trajectory heat map, use the color scale encoding technology to generate a compaction degree gradient map based on color difference;

[0015] Based on the compaction degree gradient map, use 3D modeling and data matching technology to generate a 3D visual compaction map.

[0016] In one embodiment, the data processing module is further configured to extract the frequency domain features of the instantaneous vibration signal according to the following steps:

[0017] Perform FFT transformation on the instantaneous vibration signal using the following formula:

[0018]

[0019] where X[k] represents the frequency domain representation, k represents the index of the frequency component, n represents the index of the sampling point, N represents the total number of collected samples, j represents the imaginary unit, represents the complex exponential function, which is used to convert the time domain signal into the frequency domain signal;

[0020] Calculate the frequency component corresponding to each X[k] using the following formula:

[0021]

[0022] where f k represents the frequency of the k-th frequency component, and F s represents the sampling frequency;

[0023] Calculate the amplitude of each frequency component using the following formula:

[0024]

[0025] where |X[k]| represents the amplitude of each frequency component, Re(X[k]) represents the real part of X[k], and Im(X[k]) represents the imaginary part of X[k];

[0026] Calculate the phase of each frequency component using the following formula:

[0027]

[0028] where ∠X[k] represents the phase angle of each frequency component, Re(X[k]) represents the real part of X[k], and Im(X[k]) represents the real part of X[k];

[0029] Calculate the power of each frequency component using the following formula:

[0030] P[k] = |X[k]| 2

[0031] Among them, P[k] represents the power of each frequency component, and X[k] represents the frequency-domain representation;

[0032] The frequency centroid of the signal is calculated using the following formula:

[0033]

[0034] Among them, C f represents the weighted average of the frequency components with the amplitudes of the frequency components as weights;

[0035] The frequency kurtosis of the signal is calculated using the following formula:

[0036]

[0037] Among them, K f represents the frequency kurtosis of the signal, and C f represents the weighted average of the frequency components;

[0038] The rate of change of the signal is calculated using the following formula:

[0039]

[0040] Among them, R f represents the rate of change of the signal, and X[k] represents the frequency-domain representation;

[0041] The following formula is used to construct a comprehensive feature vector:

[0042] F = [|X[0]|, |X[1]|, …, |X[N / 2]|, ∠X[0], ∠X[1], ∠X[N / 2], P[0], P[1], …, P[N / 2], C f , K f , R f

[0043] Among them, the amplitudes |X[k]| of all frequency components are arranged in order to form the first sub-vector, the phases ∠X[k] of all frequency components are arranged in order to form the second sub-vector, the powers P[k] of all frequency components are arranged in order to form the third sub-vector, the frequency centroid C f , the frequency kurtosis K f and the rate of change of frequency R f are used as additional features.

[0044] In one embodiment, the data processing module is further configured to:

[0045] Construct spatio-temporal features using the following formula:

[0046]

[0047] Among them, a​v (τ) represents the vertical acceleration of the vibrating wheel at time τ, N(τ) represents the cumulative number of compaction passes, T represents the ambient temperature, T0 represents the reference temperature, α represents the temperature attenuation coefficient, and δ(x - x(τ)) represents the spatial position matching function.

[0048] In one of the embodiments, the data processing module is further configured to:

[0049] Calculate the degree of compaction using the following formula:

[0050]

[0051] where DCI(t) represents the real-time dynamic compaction degree, a v(τ) represents the vertical acceleration speed of the vibrating wheel at time t, N(τ) represents the cumulative number of compaction passes, α represents the temperature attenuation coefficient, T represents the ambient temperature, T0 represents the standard working condition temperature, H S represents the designed loose paving temperature, H d represents the actual compaction thickness, Δv represents the rolling speed deviation, and η represents the material property correction coefficient.

[0052] In one of the embodiments, the data processing module is further configured to:

[0053] Analyze and screen the sensitive features of the compaction degree using the following formula:

[0054]

[0055] where ξ i represents the correlation degree of the i-th factor, x'0(j) represents the reference sequence, represents the normalized comparison sequence, min i min j |x'0(j) - x' i (j)| represents the global minimum difference, max i max j |x'0(j) - x' i (j)| represents the global maximum difference, ρ represents the resolution coefficient, and j represents the time point or data point index.

[0056] In one of the embodiments, the data processing module is further configured to:

[0057] Train the LSTM neural network model using the following loss function:

[0058]

[0059] where L combined represents the multi-task loss function combining attention and adaptive weights, ω idenotes the adaptive weight of the i-th sample, and λ1, λ2, and λ3 denote task weight coefficients, and denotes the true value of the i-th sample, and MSE denotes the mean square error loss function.

[0060] In one embodiment, the visualization module is further configured to:

[0061] generate a three-dimensional visualization compaction map using the following formula:

[0062]

[0063] where h(x, y, z, t) denotes the time-varying compaction degree distribution function, denotes the fractional-order time derivative, S denotes the stress tensor, β and γ denote material property weight coefficients, δ denotes the temperature decay coefficient, T denotes the ambient temperature, V denotes the volume of the detection area, denotes the topological complexity index, V i denotes the i-th sub-region, denotes the average sub-region volume, χ(S i ) denotes the Euler number of the sub-region surface.

[0064] In one embodiment, the visualization module is further configured to:

[0065] generate a compaction degree gradient map using the following formula:

[0066]

[0067] where H(x, y, t) denotes the spatio-temporal thermal map of the compaction degree, x denotes the spatial horizontal coordinate, y denotes the spatial vertical coordinate, and t denotes the time dimension, and denotes the spatial gradient operator, denotes the second-order time derivative operator.

[0068] In one embodiment, an experimental detection module is further included, which is configured to:

[0069] receive multi-source heterogeneous detection data of the compaction track, compaction passes, loose paving thickness, and compaction thickness after the experiment of the roller and manual detection, extract features and integrate and match them to form an experimental database;

[0070] calculate the compaction degree detection deviation according to the data in the experimental database;

[0071] adjust the corresponding parameters for compaction degree calculation according to the compaction degree detection deviation value.

[0072] In one embodiment, the experimental detection module is further configured to:

[0073] Receive the evolution law information of the compaction degree varying with time, temperature, and humidity, and use multi-source data fusion technology to construct a compaction prediction model;

[0074] Take the time, temperature, and humidity information received in real time as inputs, and use the compaction prediction model to generate compaction quality prediction information;

[0075] According to the compaction quality prediction information, compare the warning thresholds and standards in the warning rule base to generate corresponding warning information.

[0076] In a second aspect, the present application also provides a subgrade compaction monitoring method, including:

[0077] Receive the real-time coordinate data of the roller and the instantaneous vibration signal of the vibrating wheel;

[0078] Use FFT transformation to extract the frequency-domain characteristics of the instantaneous vibration signal;

[0079] Based on the coordinate data, use RTK solution technology to generate a rolling trajectory coordinate sequence;

[0080] Based on the coordinate sequence and frequency-domain characteristics, construct spatio-temporal characteristics;

[0081] Use the spatio-temporal characteristics to construct an LSTM neural network model and calculate the compaction degree;

[0082] Use grey relational analysis to screen the sensitive characteristics of the compaction degree and generate a compaction degree report;

[0083] Generate a compaction trajectory heat map according to the compaction degree report;

[0084] Based on the compaction trajectory heat map, use color scale coding technology to generate a compaction degree gradient map based on color difference;

[0085] Based on the compaction degree gradient map, use 3D modeling and data matching technology to generate a 3D visualization compaction map.

[0086] In a third aspect, the present application also provides a computer device, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, it realizes the functions of the above-mentioned subgrade compaction monitoring system.

[0087] In a fourth aspect, the present application also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it realizes the functions of the above-mentioned subgrade compaction monitoring system.

[0088] The above-mentioned subgrade compaction monitoring system. The data processing module receives the real-time coordinate data of the roller and the instantaneous vibration signal of the vibrating wheel to grasp the working state of the roller; uses the FFT transform to extract the frequency-domain characteristics of the instantaneous vibration signal, analyzes key information such as the frequency, amplitude, and phase of the vibration signal, and provides accurate vibration characteristic parameters for compaction degree calculation; based on the coordinate data, uses the RTK solution technology to generate a rolling trajectory coordinate sequence and restore the rolling path of the roller; based on the coordinate sequence and frequency-domain characteristics, constructs spatio-temporal characteristics, combines time and space factors with vibration characteristics to form a comprehensive description of the compaction process, and provides comprehensive data support. Uses the spatio-temporal characteristics to construct an LSTM neural network model to calculate the compaction degree, learns complex patterns and rules in the compaction process through the deep learning model, and realizes the prediction of the compaction degree; uses grey relational analysis to screen the sensitive characteristics of the compaction degree, generates a compaction degree report to identify key factors that have a greater impact on the compaction degree, and provides optimization suggestions. The visualization module generates a compaction trajectory heat map according to the compaction degree report to display the distribution of the compaction area and the difference in compaction degree. Based on the compaction trajectory heat map, uses the color scale coding technology to generate a compaction degree gradient map based on color difference to refine the visualization effect of the compaction degree, and distinguishes the areas with low to high compaction degree through different colors, which helps to discover weak links in the compaction process. Based on the compaction degree gradient map, uses 3D modeling and data matching technology to combine the compaction information with the actual terrain and landform to form a 3D model with a sense of space and reality, and provides a panoramic view of the compaction situation. Brief Description of the Drawings

[0089] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for use in the description of the embodiments or related technologies. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0090] Figure 1 It is a structural diagram of a subgrade compaction monitoring system of the present invention;

[0091] Figure 2 It is a flowchart of a subgrade compaction monitoring method of the present invention; Detailed Embodiments

[0092] In order to make the purpose, technical solutions and advantages of the present application more clear, the following further details the present application in conjunction with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0093] A subgrade compaction monitoring system of the present application, the implementation environment includes sensors, a communication network, and a data processing terminal / server. The application scenarios cover real-time monitoring, quality assessment, and early warning decision-making of subgrade compaction. When it is necessary to monitor the subgrade compaction quality, the sensors collect data, which is transmitted through the communication network to the data processing terminal / server for analysis, and then the results are fed back to the construction equipment for construction personnel to optimize the compaction process.

[0094] In one embodiment, as Figure 1 shown, a subgrade compaction monitoring system is provided. In this embodiment, it is exemplified by the deployment of the system on the terminal. It can be understood that this embodiment is exemplified by the integration of the method into the subgrade ballast terminal. It can be understood that the system can also be deployed on the server, and the two-way transmission of data and instructions is realized through the interaction between the terminal and the server. In this embodiment, the system includes a data processing module 103 and a visualization module 104:

[0095] The data processing module 103 is used for:

[0096] Receiving the real-time coordinate data of the roller and the instantaneous vibration signal of the vibration wheel.

[0097] Among them, the real-time coordinate data of the roller can be obtained through the Beidou positioning module 101, and the real-time coordinate data contains the current position information of the roller. The instantaneous vibration signal of the vibration wheel is obtained through the vibration sensing module 102 installed on the vibration wheel. The instantaneous vibration signal of the vibration wheel contains information such as the frequency, amplitude, and phase of the vibration, which can reflect the working state and compaction effect of the vibration wheel. By receiving the real-time coordinate data of the roller and the instantaneous vibration signal of the vibration wheel, through analysis, it can be judged whether the compaction process meets the expected compaction standard.

[0098] Using the FFT transform to extract the frequency-domain characteristics of the instantaneous vibration signal.

[0099] Among them, FFT (Fast Fourier Transform) is an efficient algorithm for calculating the discrete Fourier transform. Using the FFT transform can extract the amplitude, phase, and other characteristics of each frequency component of the instantaneous vibration signal. These frequency-domain characteristics reflect the vibration characteristics of the vibration wheel, such as the main frequency, frequency distribution, and energy concentration.

[0100] Based on the coordinate data, using the RTK solution technology to generate a rolling trajectory coordinate sequence.

[0101] Among them, RTK (Real-time kinematic) is a differential method for real-time processing of carrier phase observations at two measurement stations. By setting up a reference station near the construction site, the reference station and the Beidou positioning module 101 on the roller receive satellite signals simultaneously. The reference station calculates the error correction values of the satellite signals based on its known precise coordinates and sends these correction values to the Beidou positioning module 101 in real time. After receiving the correction data from the reference station, the Beidou positioning module 101 uses the RTK algorithm to perform differential processing on the original coordinate data, enabling the positioning accuracy to reach the centimeter level or even higher. The generated rolling trajectory coordinate sequence can record the position information of each roll during the construction process. Arranged in chronological order, it forms continuous trajectory points, which not only contain the longitude and latitude information of the roller, but also can obtain the rolling path in three-dimensional space by combining with elevation data.

[0102] Based on the coordinate sequence and frequency domain features, spatio-temporal features are constructed.

[0103] Among them, the vibration characteristics of the roller at different times and spaces are integrated to form a comprehensive feature vector. Each coordinate point in the coordinate sequence is associated with the corresponding frequency domain feature. For example, for each coordinate point on the rolling trajectory, frequency domain features such as the main frequency, frequency distribution, and energy concentration of the vibration wheel are recorded, so that the coordinate point carries the vibration information in the time dimension, forming spatio-temporal features, enabling the analysis of the compaction process from both time and space dimensions. In the time dimension, the vibration changes of the vibration wheel at different time periods can be understood to judge the stability of the compaction process; in the space dimension, the uniformity and coverage of the compaction operation can be evaluated.

[0104] An LSTM neural network model is constructed using spatio-temporal features, and the compaction degree is calculated.

[0105] Among them, LSTM (Long Short-Term Memory) is a time recurrent neural network that can effectively process and predict time series data. The LSTM neural network model captures the long-term dependencies in the compaction process and learns the spatio-temporal feature patterns at different positions and times. The trained LSTM neural network model can output the corresponding compaction degree prediction value according to the input spatio-temporal feature sequence. The LSTM neural network model realizes the prediction of the compaction degree of new data by learning the mapping relationship between the spatio-temporal features and the compaction degree in the historical data.

[0106] Grey relational analysis is used to screen the sensitive features of the compaction degree and generate a compaction degree report.

[0107] Among them, grey relational analysis is a method used to measure the degree of association between factors. In compaction degree evaluation, there are various influencing factors, such as frequency domain characteristics like the frequency and amplitude of the vibrating wheel, and spatio-temporal characteristics such as the traveling speed and number of rolling passes of the roller. Through grey relational analysis, it is possible to determine which characteristics have a high degree of association with the compaction degree, that is, sensitive characteristics. The steps are as follows: Determine the reference sequence, usually the standard value or ideal value of the compaction degree; standardize the characteristic data of the influencing factors to form a comparison sequence to eliminate the influence of dimension and unit of dimension; calculate the difference between the comparison sequence and the reference sequence, and find the global minimum difference and global maximum difference in combination with the distinguishing coefficient; use the formula to calculate the correlation degree of each factor. The greater the correlation degree, the stronger the association between the factor and the compaction degree, that is, the sensitive characteristic. After selecting the sensitive characteristics of the compaction degree, generate a compaction degree report according to the data and analysis results of the sensitive characteristics. The content of the report can include: the calculated value of the compaction degree based on the sensitive characteristics, reflecting the compaction situation at different positions; the specific values of the sensitive characteristics and their influence degree on the compaction degree, showing the role of key factors; the comparison result with the compaction degree standard, showing the areas where the compaction degree meets the standard and does not meet the standard; records of abnormal situations during construction, such as mutation points of the compaction degree or areas with abnormally low values, etc. According to the provided compaction quality information, construction personnel can timely adjust the construction strategy, optimize the compaction process, and ensure that the subgrade compaction quality meets the design requirements.

[0108] The visualization module 104 is used for:

[0109] Generate a compaction trajectory heat map according to the compaction degree report.

[0110] Among them, the compaction degree report contains the compaction degree data of the grid or coordinate points in the construction area, and the rolling trajectory of the roller records the traveling path and coverage range. Match the compaction degree data with the rolling trajectory data of the roller to form a data set containing position and compaction degree information. Divide the construction area into grids, and count the compaction degree data in each grid unit and calculate the average value as the representative value of the compaction degree of the unit. According to the compaction degree value range, set the color mapping rule, map the compaction degree value of each grid unit to the corresponding color to generate a heat map, visually display the spatial distribution of the compaction degree, and help construction personnel discover areas with insufficient or excessive compaction.

[0111] Based on the compaction trajectory heat map, use the color scale coding technology to generate a compaction degree gradient map based on color difference.

[0112] Among them, it is possible to analyze the mapping relationship between the colors of the thermal map and the compaction degree, subdivide the color levels, form an accurate color scale coding scheme, divide the compaction degree into 20 levels at 5% intervals, and each level corresponds to a different color according to the color scale. Calculate the difference in compaction degree between each grid cell in the thermal map and its neighboring grid cells, and use a numerical value to represent the degree of compaction change. Map the spatial gradient value to the color difference to generate a compaction degree gradient map.

[0113] Based on the compaction degree gradient map, use 3D modeling and data matching technology to generate a 3D visual compaction map.

[0114] Among them, according to the topographic and geomorphic data, a 3D terrain model of the construction area can be constructed using 3D modeling software or algorithms to restore the topographic features of the construction site. Integrate the compaction degree data in the compaction degree gradient map with the 3D terrain model. Through spatial coordinate matching, map each compaction degree data point to a specific position on the 3D terrain model. Based on the 3D terrain model, according to the matched compaction degree data, assign corresponding colors and textures to each area of the model. The colors and textures are based on the color mapping rules in the compaction degree gradient map to visually display the distribution of the compaction degree in 3D space.

[0115] The above-mentioned subgrade compaction monitoring system, through the data processing module 103 and the visualization module 104, uses the real-time coordinate data and instantaneous vibration signals of the roller, combines FFT transformation, RTK solution technology and LSTM model to realize the real-time monitoring and accurate evaluation of the subgrade compaction process. Analyze the compaction process from the time and space dimensions, use grey relational analysis to screen out the sensitive features of the compaction degree, generate a compaction degree report, and generate a compaction trajectory thermal map, a compaction degree gradient map and a 3D visual compaction map through the visualization module 104. Improve the efficiency and accuracy of compaction monitoring, help construction personnel discover areas with insufficient or excessive compaction, optimize the construction plan, reduce construction costs and time, and effectively solve the problems of low efficiency, long cycle, large destructiveness of traditional compaction detection methods, and insufficient multi-source data fusion, low accuracy and single visualization representation in monitoring technologies.

[0116] In one of the embodiments, the data processing module 103 is further configured to extract the frequency domain features of the instantaneous vibration signal according to the following steps:

[0117] Perform FFT transformation on the instantaneous vibration signal using the following formula:

[0118]

[0119] Among them, X[k] represents the frequency domain representation, k represents the index of the frequency component, n represents the index of the sampling point, N represents the total number of collected samples, j represents the imaginary unit, represents the complex exponential function, which is used to convert the time domain signal into the frequency domain signal.

[0120] Specifically, the FFT decomposes the time-domain waveform of the vibration signal, i.e., the vibration intensity varying with time, into a superposition of a series of sine waves, and each sine wave corresponds to a specific frequency component, enabling the frequency characteristics of the signal to be quantified, which is conducive to the calculation of subsequent steps.

[0121] Use the following formula to calculate the frequency component corresponding to each X[k]:

[0122]

[0123] where f k represents the frequency of the k-th frequency component, and F s represents the sampling frequency.

[0124] Specifically, map the discrete FFT result to the actual physical frequency. Subgrade materials with different compaction degrees have different response frequencies to vibration, providing frequency-domain characteristics closely related to the compaction state for analysis.

[0125] Use the following formula to calculate the amplitude of each frequency component:

[0126]

[0127] where |X[k]| represents the amplitude of each frequency component, Re(X[k]) represents the real part of X[k], and Im(X[k]) represents the imaginary part of X[k].

[0128] Exemplarily, this formula can extract the amplitude |X[k]| characteristics of the rate components in the FFT result, and the amplitude |X[k]| characteristics can explain the dynamic response characteristics of the subgrade material. During the compaction process, if the low-frequency amplitude increases significantly, it may reflect the improvement of material density. The amplitude can be used to identify the main frequency, and the frequency component with the largest amplitude can be identified as the main frequency of the vibration source.

[0129] Use the following formula to calculate the phase of each frequency component:

[0130]

[0131] where ∠X[k] represents the phase angle of each frequency component, Re(X[k]) represents the real part of X[k], and Im(X[k]) represents the real part of X[k].

[0132] Specifically, through this formula, the system can extract the phase characteristics of the vibration signal and explain the dynamic response law of the subgrade material. If the subgrade material is evenly compacted, the phase distribution of the vibration signal should be concentrated and stable. If the phase is dispersed or mutated, it represents loose materials, abnormal equipment, or thermal expansion and contraction of materials caused by temperature changes.

[0133] Calculate the power of each frequency component using the following formula:

[0134] P[k] = |X[k]| 2

[0135] where P[k] represents the power of each frequency component, and X[k] represents the frequency-domain representation.

[0136] Exemplarily, P[k] represents the energy density of the signal at frequency f k The more concentrated the energy, the more significant the role of vibration at this frequency. Since power is the square of the amplitude, this square relationship makes power more sensitive to amplitude changes and is suitable for capturing the strength differences of vibration signals.

[0137] Calculate the frequency centroid of the signal using the following formula:

[0138]

[0139] where C f represents the weighted average of the frequency components with the amplitudes of each frequency component as weights;

[0140] Specifically, the frequency centroid C f , through the method of weighted average, reflects the energy concentration degree of different frequency components in the signal. The frequency component with a larger amplitude |X[k]| has a greater impact on the centroid position.

[0141] Calculate the frequency kurtosis of the signal using the following formula:

[0142]

[0143] where K f represents the frequency kurtosis of the signal, and C f represents the weighted average of the frequency components.

[0144] Specifically, the frequency kurtosis of the signal is used to measure the concentration degree of the signal frequency components. High kurtosis indicates that the energy is highly concentrated in a few high-frequency components, such as near K f ≈ C f , and low kurtosis indicates that the energy is dispersed in a wide frequency range. When the subgrade material is insufficiently compacted, the vibration energy is distributed in the low-frequency band and the frequency kurtosis is low. When the subgrade material is moderately compacted, the proportion of high-frequency energy increases and the frequency kurtosis rises. When the subgrade material is over-compacted, the high-frequency energy decays and the frequency kurtosis may decrease or tend to be stable.

[0145] Calculate the rate of change of the signal using the following formula:

[0146]

[0147] where R frepresents the change rate of the signal, and X[k] represents the frequency-domain representation.

[0148] Exemplarily, the change rate of the signal is a key indicator for measuring the dynamic characteristics of the frequency-domain distribution of the vibration signal. It reflects the degree of fluctuation of the signal in the frequency domain by analyzing the amplitude changes of adjacent frequency components, and further evaluates the uniformity of the compaction process and the stability of the material response. When the compaction is insufficient, the material is loose, and the vibration energy distribution is dispersed, and R f is relatively high. When the compaction is moderate, the energy concentrates in the high frequency, and the frequency-domain distribution tends to be stable, and R f decreases. When the compaction is excessive, the high-frequency energy decays, and the frequency-domain distribution is dispersed again, and R f may rise again.

[0149] The following formula is used to construct the comprehensive feature vector:

[0150] F = [|X[0]|, |X[1]|, …, |X[N / 2]|, ∠X[0], ∠X[1], ∠X[N / 2], P[0], P[1], …, P[N / 2], C f , K f , R f

[0151] wherein, the amplitudes |X[k]| of all frequency components are arranged in sequence to form the first sub-vector, the phases ∠X[k] of all frequency components are arranged in sequence to form the second sub-vector, the powers P[k] of all frequency components are arranged in sequence to form the third sub-vector, the frequency centroid C f , the frequency kurtosis K f , and the frequency change rate R f are used as additional features.

[0152] Specifically, the comprehensive feature vector F comprehensively characterizes the physical characteristics of the vibration signal by fusing the amplitude, phase, power in the frequency domain and the centroid, kurtosis, and change rate of statistics. The value lies in: multi-dimensional perception, simultaneously capturing the energy distribution, dynamic response, and stability. Optimization of the model input, providing highly discriminative features for deep learning models such as LSTM. In practical applications, RTK positioning data such as rolling trajectories, speeds, and temperature sensor data can be combined to construct a spatio-temporal correlation feature matrix to improve the accuracy of compaction quality assessment.

[0153] In one of the embodiments, the data processing module 103 is further configured to:

[0154] Use the following formula to construct the spatio-temporal features:

[0155]

[0156] wherein, a v ​$(\tau)$ represents the vertical acceleration of the vibrating wheel at time $\tau$, $N(\tau)$ represents the cumulative number of compaction passes, $T$ represents the ambient temperature, $T_0$ represents the reference temperature, $\alpha$ represents the temperature attenuation coefficient, and $\delta(x - x(\tau))$ represents the spatial position matching function.

[0157] Specifically, through the coupling of vibration energy and the number of compaction passes, that is, by using v $a(\tau)\cdot N(\tau)$ to represent the cumulative effect of vibration energy over time and the number of rolling passes, and using the dynamic correction of temperature on the compaction effect, that is , and controlling the correction amplitude according to the temperature attenuation coefficient $\alpha$, using the spatial position matching function $\delta(x - x(\tau))$, integrating time dynamics such as vibration energy and compaction times, ambient temperature factors, and spatial coordinate information into a single eigenvalue to reflect the comprehensive compaction state at a certain position under specific time and environment.

[0158] In one embodiment, the data processing module 103 is further configured to:

[0159] Calculate the compaction degree using the following formula:

[0160]

[0161] where $DCI(t)$ represents the real-time dynamic compaction degree, $a$ v(τ) represents the vertical acceleration speed of the vibrating wheel at time $t$, $N(\tau)$ represents the cumulative number of compaction passes, $\alpha$ represents the temperature attenuation coefficient, $T$ represents the ambient temperature, $T_0$ represents the standard working condition temperature, $H$ S represents the designed loose paving temperature, $H$ d represents the actual compaction thickness, $\Delta v$ represents the rolling speed deviation, and $\eta$ represents the material property correction coefficient.

[0162] Specifically, through integral operation, $a$ v(τ) reflects the dynamic energy exerted by the vibrating wheel on the roadbed, $N(\tau)$ reflects the number of compaction passes, and the temperature correction factor accumulates in the time dimension, reflecting the energy accumulation effect of the dynamic compaction process. Using $(H$ S $- H$ d ) to reflect the deviation between the designed thickness and the actual thickness. If $H$ S $> H$ d it means under-compaction, otherwise it may be over-compaction. $\Delta v$ reflects the stability of the rolling speed, and quantifies the deviation degree between the construction process and the design goal through the weighted combination of geometry and speed errors . At the same time, referring to the material property correction coefficient $\eta$, a dynamic compaction degree evaluation with multi-physical quantity coupling is realized.

[0163] In one embodiment, the data processing module 103 is further configured to:

[0164] Use the following formula to analyze and screen the sensitive features of compaction degree:

[0165]

[0166] where ξ i represents the correlation degree of the i-th factor, x'0(j) represents the reference sequence, represents the standardized comparison sequence, min i min j |x'0(j)-x' i (j)| represents the global minimum difference, max i max j |x'0(j)-x' i (j)| represents the global maximum difference, ρ represents the resolution coefficient, and j represents the time point or data point index.

[0167] Specifically, standardize the reference sequence x'0(j) and the comparison sequence x' i (j) to eliminate the dimension and order of magnitude differences. For each time point j, calculate the absolute difference |x'0(j)-x' i (j)| between the comparison sequence and the reference sequence, determine the global minimum difference min i min j |x'0(j)-x' i (j)| and the global maximum difference max i max j |x'0(j)-x' i (j)|, and calculate the correlation degree ξ i ,

[0168] If |x'0(j)-x' i (j)| is close to the global minimum difference min i min j |x'0(j)-x' i (j)|, it indicates that this factor changes highly synchronously with the compaction degree and has a high correlation degree;

[0169] If |x'0(j)-x' i (j)| is close to the global maximum difference max i max j |x'0(j)-x' i (j)|, it indicates that the correlation between this factor and the compaction degree is weak. The advantage of using grey correlation analysis to quantify the dynamic correlation between factors is that it is applicable to engineering data with non-linear and non-normal distributions; reliable conclusions can be drawn with only a small amount of construction data.

[0170] In one of the embodiments, the data processing module 103 is further configured to:

[0171] Train the LSTM neural network model using the following loss function:

[0172]

[0173] where L combined represents the multi-task loss function that combines attention and adaptive weights, ω i represents the adaptive weight of the i-th sample, λ1, λ2, and λ3 represent the task weight coefficients, and represent the true value of the i-th sample, and MSE represents the mean squared error loss function.

[0174] Specifically, this loss function can balance the contributions of each task to the total loss by jointly optimizing multiple related tasks such as compaction degree prediction, vibration signal analysis, and temperature compensation, introducing the task weight coefficients λ1, λ2, and λ3. According to the difficulty or importance of the samples, use ω i to adaptively adjust the loss contribution, solve the data imbalance problem, calculate the error MSE separately in the three subtasks, and integrate it into the total loss through weighted summation to improve the generalization ability and robustness of the model.

[0175] In one embodiment, the visualization module 104 is further configured to:

[0176] Generate a three-dimensional visualization compaction map using the following formula:

[0177]

[0178] where h(x, y, z, t) represents the time-varying compaction degree distribution function, represents the fractional-order time derivative, S represents the stress tensor, β and γ represent the material property weight coefficients, δ represents the temperature decay coefficient, T represents the ambient temperature, V represents the volume of the detection area, represents the topological complexity index, V i represents the i-th sub-region, represents the average sub-region volume, χ(S i ) represents the Euler number of the sub-region surface.

[0179] Exemplarily, describe the non-linear variation characteristics of the compaction degree h(x, y, z, t) with time. The fractional-order derivative captures the memory effect of the compaction process such as the viscous response of the material, which is more in line with the slow-varying characteristics in actual construction compared to the integer-order derivative. The square term can amplify the amplitude of the compaction degree fluctuation in the time dimension and highlight the abnormal change area. reflects the degree of concentration of the internal stress field of the material. The square term Enhance the weight of the stress anomaly region to detect potential uneven compaction or material damage. In, the sub-region volume ratio is used to measure the deviation of the volume of the i-th sub-region from the average volume. The surface Euler number χ(S i ) characterizes the topological connectivity of the sub-region such as the number of holes. By weighted summation, comprehensively evaluate the geometric and topological anomalies of the sub-region. The temperature correction term e δT In, the temperature decay coefficient δ represents the influence of the ambient temperature T on the material compaction characteristics. High temperature may cause the material to soften and reduce the compaction efficiency. The exponential term increases with the increase of temperature, thus highlighting the abnormal risk of the high temperature region in the thermal map. Through triple integration of the detection region V, weighted fusion of time dynamics, stress distribution, topological complexity and temperature effect, a three-dimensional visual compaction map is generated to present the spatial distribution anomaly of the roadbed compaction degree.

[0180] In one of the embodiments, the visualization module 104 is further configured to:

[0181] Generate a compaction degree gradient map using the following formula:

[0182]

[0183] where H(x, y, t) represents the spatio-temporal thermal map of the compaction degree, x represents the spatial horizontal coordinate, y represents the spatial vertical coordinate, and t represents the time dimension, and represents the spatial gradient operator, represents the second-order time derivative operator.

[0184] Specifically, use spatial echelon logarithmic operation to measure the severity of the local change of the compaction degree on the two-dimensional spatial plane (x-y), highlighting the contour of the abnormal region. Calculate the square root of the second-order time derivative Through square root normalization, the dynamic characteristics of the compaction degree change can be reflected. Through spatio-temporal coupling analysis, multiply the result of the spatial gradient logarithmic operation by the square root of the second-order time derivative to obtain the two-dimensional gradient map C diff (x, y) to realize the dynamic evaluation of the construction quality.

[0185] In one of the embodiments, it further includes an experimental detection module 105 for:

[0186] Receive multi-source heterogeneous detection data of the compaction trajectory, compaction passes, loose paving thickness and compaction thickness after the experiment of the roller and manual detection, extract features and integrate and match to form an experimental database;

[0187] Calculate the detection deviation of the compaction degree according to the data in the experimental database;

[0188] Adjust the corresponding parameters for compaction degree calculation according to the deviation value of compaction degree detection.

[0189] Exemplarily, associate the characteristic data with the compaction degree results detected manually and the actual detection results to construct an annotation database. Compare the model prediction value with the manually detected value, quantify the model error, and optimize the key parameters of the compaction degree calculation model according to the deviation analysis results to improve the real-time prediction accuracy. Through the closed-loop process of data fusion - feature extraction - deviation analysis - parameter optimization, ensure the accuracy of the compaction degree calculation model under different working conditions; through dynamic parameter adjustment, guide the optimization of construction parameters; combine manual detection and model prediction to form an intelligent monitoring system of perception - analysis - decision-making.

[0190] In one of the embodiments, the experimental detection module 105 is further configured to:

[0191] Receive the evolution law information of the compaction degree changing with time, temperature and humidity, and use the multi-source data fusion technology to construct a compaction prediction model;

[0192] Take the time, temperature and humidity information received in real time as inputs, and use the compaction prediction model to generate compaction quality prediction information;

[0193] According to the compaction quality prediction information, compare the warning thresholds and standards in the warning rule library to generate corresponding warning information.

[0194] Specifically, integrate the time-series compaction degree data, environmental temperature and humidity data, and construction parameter data such as rolling speed and number of passes to construct a dynamic compaction prediction model. The input data can be the change curve of the compaction degree with time, and an integrated model that uses LSTM to process time series + random forest to capture non-linear relationships to generate compaction quality prediction information. Through the preset multi-level warning rule library, based on the compaction quality prediction information, calculate the risk coefficient, compare the warning thresholds and standards, and generate corresponding warning information.

[0195] To further illustrate the solution of the embodiments of the present application, a specific embodiment is described below for illustration.

[0196] A Beidou time service server mushroom head antenna, a vibration sensor, a data processing unit providing data calculation capabilities, and an intelligent display terminal are installed on a certain roller. At the same time, a subgrade compaction monitoring system of the present invention is deployed, powered by the roller, starts synchronously with the roller, and the device runs automatically after startup. The startup of the software and the connection of the network do not require manual operation. The user only needs to view the intelligent display terminal to know the compaction quality indicators during construction. A subgrade compaction monitoring system of this embodiment includes a data processing module 103 and a visualization module 104:

[0197] The data processing module 103 is used for:

[0198] Receiving the real-time coordinate data of the roller and the instantaneous vibration signal of the vibration wheel.

[0199] Specifically, through the mushroom head antenna of the Beidou time service server, the Beidou satellite signal is received to obtain the real-time coordinate data of the roller. Through the vibration sensor, the acceleration of the vibration wheel is collected to obtain the instantaneous vibration signal of the vibration wheel.

[0200] Using the FFT transform to extract the frequency domain characteristics of the instantaneous vibration signal.

[0201] Exemplarily, as a dynamic loading device, during the rolling process, the vibration wheel of the roller is simultaneously subjected to the exciting force from the roller itself and the reaction force of the roadbed. The change of mechanical parameters is closely related to the compaction degree of the roadbed filling. The data processing module can use the FFT transform based on the instantaneous vibration signal of this vertical vibration response to realize the real-time monitoring and feedback control of the compaction quality during the rolling process.

[0202] Based on the coordinate data, using the RTK solution technology to generate the rolling trajectory coordinate sequence.

[0203] Exemplarily, by setting up a reference station near the construction site, the Beidou positioning module 101 on the reference station and the mushroom head antenna of the Beidou time service server on the roller receive satellite signals simultaneously. The reference station calculates the error correction value of the satellite signal according to its own known accurate coordinates and sends the correction value to the roller. After receiving the correction data, the roller uses the RTK solution algorithm to perform differential processing on the coordinate data to generate the rolling trajectory coordinate sequence. In this example, the rolling trajectory coordinate sequence contains information about the compaction trajectory and the number of compaction passes.

[0204] Based on the coordinate sequence and the frequency domain characteristics, construct the spatio-temporal characteristics.

[0205] Specifically, based on the information about the compaction trajectory and the number of compaction passes in the rolling trajectory coordinate sequence, and the frequency domain characteristic information of the instantaneous vibration signal, construct the spatio-temporal characteristics. This makes the coordinate points not only have spatial attributes but also carry vibration information in the time dimension, which is beneficial to the use of the calculation model in the subsequent steps.

[0206] Use the spatio-temporal characteristics to construct an LSTM neural network model and calculate the compaction degree.

[0207] Exemplarily, the LSTM model can learn spatio-temporal feature patterns at different positions and times. The LSTM model trained with spatio-temporal features outputs corresponding compaction degree prediction values according to the input spatio-temporal feature sequence. In this embodiment, calculating the compaction degree using the LSTM model includes identifying new layers and old layers, calculating the number of compaction passes, compaction thickness, and loose paving thickness. When the difference between the elevation data and the previous elevation data exceeds 12 cm, it is considered that a new layer has been entered. Based on the same layer: when the roller moves forward once and then backs up, it is regarded as two passes, and 4 to 8 passes are qualified passes; if the compaction thickness is less than 15 cm, it is over-compacted, if it is 15 - 25 cm, it is a qualified thickness, and if it is higher than 25 cm, it is under-compacted; if the loose paving thickness is less than 18 cm, it is under-paved, if it is 15 - 25 cm, it is a qualified thickness, and if it is higher than 25 cm, it is over-paved.

[0208] Use grey relational analysis to screen sensitive features of compaction degree and generate a compaction degree report.

[0209] Specifically, in this embodiment, grey relational analysis is used to screen sensitive features of compaction degree, analyze loose paving thickness, compaction thickness, and the number of compaction passes, and generate a compaction degree report. For example, when the roller is operating, calculate the compaction degree according to the LSTM model, use grey relational analysis to screen sensitive features, generate a comprehensive compaction degree report, and feedback layer information; when on the same layer, feedback the qualified information of loose paving thickness, compaction thickness, and the number of compaction passes, and give a graded alarm for abnormal compaction thickness, loose paving thickness, or the number of compaction passes among the three indicators.

[0210] The visualization module 104 is used for:

[0211] According to the compaction degree report, use the KDE algorithm and the Matplotlib plotting library to generate a compaction trajectory heat map.

[0212] Exemplarily, KDE (Kernel Density Estimation) can convert discrete compaction degree measurement points in the compaction degree report into a continuous density distribution. In this embodiment, the trajectory and compaction degree data in the compaction degree report are used as inputs, the Gaussian kernel function is selected as the calculation parameter for KDE, and the KDE algorithm is used to generate a density estimation map reflecting the density of the compaction trajectory. Through Matplotlib (a visualization plotting library), a compaction trajectory heat map is generated using a color gradient scheme from red to yellow to green.

[0213] Based on the compaction trajectory heat map, use color scale coding technology to generate a compaction degree gradient map based on color difference.

[0214] Specifically, in this embodiment, the number of compaction passes: from red to yellow to green indicates an increase in the number of rolling passes, and green represents the qualified number of passes; compaction thickness: the compaction thickness represented by red is less than 15 cm, the compaction thickness represented by yellow is 15 - 25 cm, and the compaction thickness represented by green is above 25 cm; loose paving thickness: the loose paving thickness represented by red is less than 18 cm; the loose paving thickness represented by yellow is 18 - 30 cm, and the compaction thickness represented by green is above 30 cm.

[0215] A subgrade compaction monitoring system of the present application uses the real-time coordinate data and instantaneous vibration signals of the vibrating wheel of a roller, combines FFT transformation, RTK solution technology, and LSTM model to monitor and evaluate the real-time process of subgrade compaction. Analyze the compaction process from the time and space dimensions, analyze and screen out the sensitive features of the compaction degree, and generate a compaction degree report. Generate a compaction trajectory heat map, a compaction degree gradient map, and a three-dimensional visual compaction map through the visualization module 104. Improve the efficiency and accuracy of compaction monitoring, help construction personnel discover areas with insufficient or excessive compaction, optimize the construction plan, reduce construction costs and time, solve the problems of low efficiency, long cycle, large destructiveness of traditional compaction detection methods, insufficient multi-source data fusion, low model accuracy, and single visual representation in monitoring technologies, and improve the intelligent level of subgrade compaction construction.

[0216] It should be understood that although the steps in the flowcharts involved in the above-described embodiments are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-described embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same moment, but can be executed at different moments. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least a part of other steps or steps in other steps.

[0217] Based on the same inventive concept, the embodiments of the present application also provide a method for implementing a subgrade compaction monitoring system as described above. The solution provided by the device to solve the problem is similar to the solution described in the above method. Therefore, the specific limitations in one or more of the following embodiments of the subgrade compaction monitoring method can refer to the limitations on the subgrade compaction monitoring system in the above text, and will not be repeated here.

[0218] In an exemplary embodiment, as Figure 2 shown, a subgrade compaction monitoring method is provided, including:

[0219] S1, Receive the real-time coordinate data of the roller and the instantaneous vibration signal of the vibrating wheel;

[0220] S2, Use FFT transformation to extract the frequency-domain characteristics of the instantaneous vibration signal;

[0221] S3, Based on the coordinate data, use RTK solution technology to generate a compaction trajectory coordinate sequence;

[0222] S4, Based on the coordinate sequence and the frequency-domain characteristics, construct spatio-temporal characteristics;

[0223] S5, Use the spatio-temporal characteristics to construct an LSTM neural network model and calculate the compaction degree;

[0224] S6, Use grey relational analysis to screen the sensitive characteristics of the compaction degree and generate a compaction degree report;

[0225] S7, According to the compaction degree report, generate a compaction trajectory heat map;

[0226] S8, Based on the compaction trajectory heat map, use color scale encoding technology to generate a compaction degree gradient map based on color difference;

[0227] S9, Based on the compaction degree gradient map, use 3D modeling and data matching technology to generate a 3D visualization compaction map.

[0228] The above embodiments only represent several implementation manners of the embodiments of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the embodiments of the application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the embodiments of the present application, several deformations and improvements can still be made, and these all belong to the protection scope of the embodiments of the present application.

Claims

1. A roadbed compaction monitoring system, characterized in that: Including data processing module and visualization module: The data processing module is used to: Receive the real-time coordinate data of the roller and the instantaneous vibration signal of the vibrating wheel; Use FFT transform to extract the frequency domain characteristics of instantaneous vibration signals; Based on the coordinate data, the RTK solution technology is used to generate the rolling track coordinate sequence; Based on the coordinate sequence and frequency domain features, constructing spatiotemporal features; Using the spatiotemporal features, a LSTM neural network model is constructed and the degree of compaction is calculated; Using grey correlation analysis to screen the sensitive features of the compaction and generate a compaction report; The visualization module is used to: Generating a compaction trajectory heat map according to the compaction report; Based on the compaction trajectory heat map, a color-difference-based compaction degree gradient map is generated using a color-scale encoding technique; Based on the compaction gradient map, a three-dimensional visual compaction map is generated using three-dimensional modeling and data matching technology.

2. The system according to claim 1, characterized in that The data processing module is also used to extract the frequency domain features of the instantaneous vibration signal according to the following steps: Use the following formula to perform FFT transformation on the instantaneous vibration signal: Among them, X[k] represents the frequency domain representation, k represents the index of the frequency component, n represents the index of the sampling point, N represents the total number of samples collected, and j represents the imaginary unit. represents the complex exponential function, which is used to convert the time domain signal into the frequency domain signal; Use the following formula to calculate the frequency component corresponding to each X[k]: Among them, f k represents the frequency of the kth frequency component, F s Indicates the sampling frequency; Calculate the magnitude of each frequency component using the following formula: Where |X[k]| represents the amplitude of each frequency component, Re(X[k]) represents the real part of X[k], and Im(X[k]) represents the imaginary part of X[k]; Calculate the phase of each frequency component using the following formula: Wherein, ∠X[k] represents the phase angle of each frequency component, Re(X[k]) represents the real part of X[k], and Im(X[k]) represents the real part of X[k]; Calculate the power of each frequency component using the following formula: P[k]=|X[k]| 2 Where P[k] represents the power of each frequency component, and X[k] represents the frequency domain representation; Calculate the frequency center of gravity of the signal using the following formula: Among them, C f It represents the weighted average of the frequency components with the amplitude of each frequency component as the weight; The frequency kurtosis of a signal is calculated using the following formula: Among them, K f represents the frequency kurtosis of the signal, C f represents the weighted average of frequency components; The rate of change of a signal is calculated using the following formula: Among them, R f represents the rate of change of the signal, X[k] represents the frequency domain representation; The comprehensive feature vector is constructed using the following formula: F=[|X[0]|,|X[1]|,…,|X[N / 2]|,∠X[0],∠X[1],∠X[N / 2],P[0],P[1],…,P[N / 2],C f ,K f ,R f ] Among them, the amplitudes of all frequency components |X[k]| are arranged in order to form the first sub-vector, the phases of all frequency components ∠X[k] are arranged in order to form the second sub-vector, the powers of all frequency components P[k] are arranged in order to form the third sub-vector, and the frequency center of gravity C f , frequency kurtosis K f and the frequency change rate R f As an additional feature.

3. The system according to claim 1, characterized in that The data processing module is also used for: The spatiotemporal features are constructed using the following formula: Among them, a v (τ) represents the vertical acceleration of the vibrating wheel at time τ, N(τ) represents the cumulative number of compaction passes, T represents the ambient temperature, T0 represents the reference temperature, α represents the temperature attenuation coefficient, and δ(xx(τ)) represents the spatial position matching function.

4. The system according to claim 1, characterized in that The data processing module is also used for: Calculate the degree of compaction using the following formula: Where DCI(t) represents the real-time dynamic compaction degree, a v(τ) represents the vertical acceleration speed of the vibrating wheel at time t, N(τ) represents the cumulative number of compaction passes, α represents the temperature attenuation coefficient, T represents the ambient temperature, T0 represents the standard operating temperature, H S Indicates the design loose laying temperature, H d It represents the actual compaction thickness, Δv represents the rolling speed deviation, and η represents the material property correction coefficient.

5. The system according to claim 3, characterized in that The data processing module is also used for: The following formula was used to analyze and screen the sensitive characteristics of the compaction degree: Among them, ξ i represents the correlation of the i-th factor, x'0(j) represents the reference sequence, Represents a standardized comparison sequence, min i min j |x'0(j)-x' i (j)| represents the global minimum difference, max i max j |x'0(j)-x' i (j)| represents the global maximum difference, ρ represents the resolution coefficient, and j represents the time point or data point index.

6. The system according to claim 1, characterized in that The data processing module is also used for: The LSTM neural network model is trained using the following loss function: Among them, L combined represents the multi-task loss function combining attention and adaptive weights, ω i represents the adaptive weight of the i-th sample, λ1, λ2 and λ3 represent the task weight coefficients, and represents the true value of the i-th sample, and MSE represents the mean square error loss function.

7. The system according to claim 1, characterized in that The visualization module is also used for: The following formula is used to generate a 3D visualization of the compaction map: Where h(x, y, z, t) represents the time-varying compaction distribution function, represents the fractional time derivative, S represents the stress tensor, β and γ represent the material property weight coefficients, δ represents the temperature attenuation coefficient, T represents the ambient temperature, and V represents the volume of the detection area. represents the topological complexity index, V i represents the i-th sub-region, represents the average sub-region volume, χ(S i ) represents the Euler number of the sub-region surface.

8. The system according to claim 1, characterized in that The visualization module is also used for: The compaction gradient map is generated using the following formula: Among them, H(x, y, t) represents the compaction spatiotemporal heat map, x represents the spatial horizontal coordinate, y represents the spatial vertical coordinate, and t represents the time dimension. and represents the spatial gradient operator, represents the time second derivative operator.

9. The system according to claim 1, characterized in that Also included are experimental detection modules for: Receive multi-source heterogeneous detection data of compaction trajectory, compaction passes, loose paving thickness and compaction thickness detected by rollers and humans after the experiment, extract features and integrate and match them to form an experimental database; Calculating the compaction degree detection deviation according to the experimental database data; The corresponding parameters of the compaction calculation are adjusted according to the compaction detection deviation value.

10. The system according to claim 9, characterized in that The experimental detection module is also used for: Receive information on the evolution of compaction with time, temperature and humidity, and use multi-source data fusion technology to build a compaction prediction model; Using the time, temperature and humidity information received in real time as input, and using the compaction prediction model to generate compaction quality prediction information; According to the compaction quality prediction information, the warning thresholds and standards in the warning rule library are compared to generate corresponding warning information.

Images