A method and device for detecting the bearing capacity of a road engineering construction material

Abstract

This invention relates to the field of intelligent sensing systems, and discloses a method and device for detecting the bearing capacity of road construction materials. The method involves simultaneously acquiring mixed signals and clean noise signals using a main measurement sensor and a reference noise sensor to construct a multi-dimensional observation matrix. After centering and whitening preprocessing, blind source separation is performed using an independent component analysis algorithm based on maximizing non-Gaussianity. The target bearing capacity response signal with ultra-Gaussian characteristics is identified based on the kurtosis value of each source signal component, and the dynamic deformation modulus is then calculated. The system includes a sensor array module, a synchronous acquisition module, a data processing module based on FPGA hardware acceleration, and a result output and storage module. This invention effectively suppresses strong coherent vibration noise, fully preserves signal details, and achieves high-precision, real-time bearing capacity detection.

Description

Technical Field

[0001] This invention belongs to the field of intelligent sensing system technology, specifically relating to a method and device for detecting the bearing capacity of road construction materials. Background Technology

[0002] With the continuous advancement of infrastructure construction, the demand for real-time and accurate testing of the load-bearing capacity of road construction materials is becoming increasingly urgent. Load-bearing capacity, as a core indicator for measuring the compaction quality of subgrade and base course materials, directly relates to the long-term stability and service life of road structures. Traditional testing methods mainly rely on static load tests or vibration response analysis based on fixed thresholds. Their theoretical basis is built upon ideal working condition assumptions, making them difficult to adapt to the complex and ever-changing dynamic environment of construction sites. Especially during the operation of large machinery such as heavy rollers, strong mechanical vibrations can couple into the testing sensors, resulting in noise amplitudes in the acquired signals that can be more than three times the effective load-bearing capacity response signal, severely masking the true mechanical state of the materials.

[0003] The load-bearing capacity of road construction materials is typically tested using acceleration, displacement, or stress sensors mounted on rollers or independent testing platforms. The aim is to invert the material's stiffness and density through vibration response characteristics. This approach relies on the overall structural response information contained in the low-frequency band of the signal, while high-frequency components are often considered interference.

[0004] However, existing detection systems generally use low-pass filters with fixed cutoff frequencies for signal preprocessing. While this can partially suppress high-frequency noise, it inevitably filters out weak low-frequency characteristic components corresponding to minute changes in bearing capacity, resulting in a persistently high rate of missed detections in areas with insufficient early compaction. Such missed detections are difficult to detect during the construction phase but may lead to uneven settlement, rutting, or even structural damage in later stages.

[0005] Existing technologies lack the ability to dynamically understand and adaptively process the noise spectrum characteristics when dealing with the contradiction between strong vibration interference and weak effective signals. On the one hand, vibration noise has non-stationary characteristics that vary with the type of machinery, rotational speed, material state, and environmental conditions, making it impossible for fixed filtering strategies to specifically suppress the actual interference frequency band. On the other hand, subtle degradation of load-bearing capacity often manifests as slow energy attenuation within a specific frequency band, requiring high-fidelity extraction while preserving the complete effective spectrum.

[0006] Current methods suffer from rigid filtering mechanisms, making it impossible to construct a mapping relationship between noise and working conditions, or to dynamically adjust filtering parameters based on real-time spectral characteristics. This results in a difficulty in balancing detection sensitivity and robustness, making it hard to reliably identify early compaction defects in complex construction scenarios. There is an urgent need for an intelligent detection method that can integrate noise feature recognition and signal preservation strategies. Summary of the Invention

[0007] To address the aforementioned technical problems, this invention provides a method for detecting the bearing capacity of road construction materials. This method involves constructing a distributed sensor array that includes a main measurement sensor and a reference noise sensor. It simultaneously acquires a mixed signal coupled with material response and environmental vibration, as well as a pure environmental vibration reference signal. The method uses a blind source separation algorithm based on higher-order statistics to demix the multiple signals, separates the independent dynamic response source signal of the material bearing capacity from the strong noise, and accurately calculates the bearing capacity index based on the pure signal.

[0008] This invention provides a method for testing the bearing capacity of road construction materials, which includes the following steps:

[0009] The first mixed signal sequence is formed by the linear superposition of the dynamic response signal of bearing capacity and the vibration and noise signal of construction machinery by the main measurement sensor set on the surface of the material to be tested in real time.

[0010] A reference noise sensor, installed on the main structure of the construction machinery and having no direct mechanical contact with the material under test, is used to synchronously collect a second reference noise signal sequence generated by the vibration of the construction machinery. The acquisition process of the first mixed signal sequence and the second reference noise signal sequence is based on a unified high-precision clock source to achieve synchronous sampling.

[0011] The first mixed signal sequence and the second reference noise signal sequence are used to construct a multidimensional observation signal matrix. Preprocessing operations are performed on the multidimensional observation signal matrix, including centering and whitening. The centering process involves subtracting the time average value of each signal sequence in the observation signal matrix. The whitening process involves performing a linear transformation on the centered observation signal matrix using a whitening matrix, so that the transformed signal components are pairwise uncorrelated and have a variance of 1.

[0012] An independent component analysis algorithm based on maximizing non-Gaussianity is used to iteratively solve the preprocessed multidimensional observation signal matrix, and a source signal matrix containing multiple statistically independent source signal components is obtained. The source signal components include the target bearing capacity dynamic response source signal and the vibration noise source signal.

[0013] For each independent source signal component in the source signal matrix, calculate the higher-order statistical parameters of its signal waveform, where the higher-order statistical parameters are the kurtosis values ​​of the signal.

[0014] Based on the preset signal physical characteristic criteria, the kurtosis values ​​are compared and analyzed, and the independent source signal component with the largest positive kurtosis value is identified as the target bearing capacity dynamic response source signal. The signal physical characteristic criteria are that the dynamic response signal generated by the material when subjected to external load impact exhibits transient pulse characteristics, and its probability density distribution has a super Gaussian distribution feature with sharp peaks and heavy tails.

[0015] Based on the pure time series waveform of the identified target load-bearing capacity dynamic response source signal, the peak value of its signal amplitude is extracted, and the dynamic deformation modulus characterizing the load-bearing capacity of the material under test is calculated according to the peak value, the known externally applied load value, and the geometric dimensional parameters of the loaded structure.

[0016] In one embodiment of the present invention, the main measuring sensor is specifically a high-sensitivity piezoresistive thin-film pressure sensor, which is encapsulated at the bottom of a rigid bearing plate and directly attached to the compacted surface of the road material to be tested. The reference noise sensor is specifically a triaxial microelectromechanical system accelerometer, which is mounted on the metal frame of the road roller or paver that generates vibration noise by magnetic adsorption or bolt fixing, and its measuring axis is consistent with the main mechanical vibration direction.

[0017] Furthermore, the synchronous sampling is achieved through a multi-channel synchronous data acquisition card. The time offset between each channel of the data acquisition card is less than 1 microsecond, its built-in analog-to-digital converter has a resolution of not less than 24 bits, and the sampling frequency is set to not less than 20 kHz to ensure complete capture of the high-frequency components of the material's dynamic response signal and the broadband spectral characteristics of vibration noise.

[0018] As one embodiment of the present invention, the whitening process is implemented as follows: First, the covariance matrix of the centered observation signal matrix is ​​calculated, and then the covariance matrix is ​​decomposed into eigenvalues ​​to obtain the eigenvalue diagonal matrix and the eigenvector orthogonal matrix. The whitening matrix is ​​the product of the eigenvector orthogonal matrix and the eigenvalue diagonal matrix raised to the power of -1 / 2.

[0019] Furthermore, the independent component analysis algorithm based on maximizing non-Gaussianity is specifically a fast fixed-point iterative algorithm. This algorithm initializes a randomized solution mixing matrix and updates the column vectors of the solution mixing matrix using the following iterative formula: First, the product of the current column vector and the whitened observed signal matrix is ​​calculated to obtain the estimated source signal components. Then, the estimated source signal components are substituted into the derivative of a preset nonlinear function, and the result is multiplied by the whitened observed signal matrix to obtain the expectation. The product of the expectation of the second derivative of the nonlinear function and the current column vector is subtracted to obtain the updated column vectors. Finally, the updated column vectors are normalized. This iterative process continues until the solution mixing matrix converges or the preset maximum number of iterations is reached. The nonlinear function is chosen as the hyperbolic tangent function.

[0020] In one embodiment of the present invention, the kurtosis value is calculated by dividing the fourth central moment of the signal sequence by the square of its variance. Specifically, the signal physical characteristic criterion involves sorting the calculated kurtosis values. Signal components with kurtosis values ​​much greater than 3 are identified as super-Gaussian signals, i.e., the dynamic response source signal of the target bearing capacity, while signal components with kurtosis values ​​close to 0 or negative are identified as sub-Gaussian signals, i.e., vibration noise source signals.

[0021] Furthermore, the formula for calculating the dynamic deformation modulus is: ,in For dynamic deformation modulus, The Poisson's ratio of the material, Peak force of externally applied load, The radius of the rigid bearing plate. The maximum deformation or maximum stress response amplitude of the material is extracted from the pure time-series waveform of the target bearing capacity dynamic response source signal.

[0022] The present invention also provides a bearing capacity testing device for road construction materials, comprising:

[0023] The sensor array module includes a main measurement sensor and a reference noise sensor. The main measurement sensor is used to acquire a first mixed signal sequence after superimposing the dynamic response signal of the load-bearing capacity and the vibration noise signal. The reference noise sensor is used to synchronously acquire a second reference noise signal sequence.

[0024] The signal conditioning and synchronous acquisition module is electrically connected to the sensor array module and is used to amplify and filter the first mixed signal sequence and the second reference noise signal sequence, and perform high-precision synchronous analog-to-digital conversion to output a digital multidimensional observation signal matrix.

[0025] A data processing module, connected to the signal conditioning and synchronous acquisition module, integrates a preprocessing unit, a blind source separation unit, a target signal identification unit, and a bearing capacity calculation unit. The preprocessing unit performs centering and whitening operations on the multidimensional observation signal matrix. The blind source separation unit runs an independent component analysis algorithm based on maximizing non-Gaussianity to separate independent source signal components. The target signal identification unit identifies the target bearing capacity dynamic response source signal by calculating the kurtosis value of each source signal component and based on a preset criterion. The bearing capacity calculation unit calculates the final dynamic deformation modulus based on the target bearing capacity dynamic response source signal.

[0026] The result output and storage module is connected to the data processing module and is used to display the calculated dynamic deformation modulus in real time, and to persistently store the original acquired signal, the separated source signal and the calculation result.

[0027] In one embodiment of the present invention, the data processing module is specifically a field-programmable gate array (FPGA) chip. The functions of the preprocessing unit, blind source separation unit, target signal recognition unit, and load-bearing capacity calculation unit are programmed using a hardware description language and embedded in the logic resources of the FPGA chip to achieve parallel hardware acceleration of the signal processing flow.

[0028] Furthermore, the blind source separation unit is implemented in the field-programmable gate array chip as a hardware computing core based on pipeline structure and coordinate rotation digital computer algorithm, used to execute vector and matrix operations in the fast fixed-point iteration algorithm at high speed, thereby meeting the computing power requirements of real-time detection at the construction site. The result output and storage module includes an LCD screen, a large-capacity solid-state drive, and a universal serial bus interface for data exchange with external devices.

[0029] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0030] 1. By constructing a specific physical layout that includes a main measurement sensor and a reference noise sensor, this invention establishes a correlation observation model between mixed signals and pure noise signals from the source of data acquisition, providing the necessary data foundation for subsequent signal separation.

[0031] 2. This invention uses blind source separation technology based on independent component analysis to replace traditional filtering technology. Instead of eliminating signals in the frequency domain, it decouples and separates signals from different sources in the statistical domain. This can completely remove coherent vibration noise with an intensity several times higher without damaging the waveform details of the target signal, thus greatly improving the signal-to-noise ratio.

[0032] 3. This invention proposes an automatic target signal identification method based on the high-order statistical characteristic of kurtosis. It utilizes the inherent physical difference in probability density distribution between load-bearing capacity response signal and mechanical vibration noise to achieve accurate identification of each signal component after separation, thus ensuring the reliability of the final calculation results.

[0033] 4. This invention deploys complex signal processing algorithms on a dedicated hardware computing platform, realizing fully automated and real-time processing of the entire process from signal acquisition, separation, identification to bearing capacity calculation. It can provide immediate and accurate feedback for on-site construction quality control, effectively solving the fundamental technical problem that existing technologies are insufficient in detecting subtle defects such as insufficient early compaction under strong vibration environments. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of the overall technical solution architecture of the bearing capacity testing method and testing device for road construction materials proposed in this invention;

[0035] Figure 2 This is a schematic diagram of the core principle framework of the blind source separation algorithm based on higher-order statistics in this invention;

[0036] Figure 3 This is a diagram of the distributed sensing and synchronous sampling framework of the sensor array module and the signal synchronous acquisition logic in this invention.

[0037] Figure 4 This is a flowchart illustrating the logical process of preprocessing (centering and whitening) the multidimensional observation signal matrix in this invention.

[0038] Figure 5 This is a framework diagram of the criteria and index generation for identifying the dynamic response source signal of the target bearing capacity and calculating the dynamic deformation modulus in this invention;

[0039] Figure 6 This is a schematic diagram of the multi-level interaction relationship and data flow of various functional units within the terminal hardware system of this invention. Detailed Implementation

[0040] Please refer to Figures 1 to 6This invention provides a method and device for detecting the bearing capacity of road construction materials, aiming to solve the technical problem that traditional detection methods cannot effectively identify the weak bearing capacity response signals of materials due to strong vibration and noise interference during heavy machinery construction. Existing technologies generally rely on low-pass filters to reduce noise in sensor-acquired signals. However, while suppressing high-frequency vibration noise, such methods inevitably filter out high-frequency components containing key compaction state information in the dynamic response of bearing capacity, resulting in a high rate of missed detection for subtle defects such as insufficient early compaction. To overcome this deficiency, this invention constructs a detection system based on distributed sensing and blind source separation. Through a combination of physical layout design and statistical signal processing, it accurately extracts the pure dynamic response signal of material bearing capacity under strong noise background, and calculates the dynamic deformation modulus accordingly, achieving high-precision real-time assessment of construction quality.

[0041] The load-bearing capacity detection method includes the following steps: using a main measuring sensor set on the surface of the material to be tested, a first mixed signal sequence is formed by the linear superposition of the dynamic response signal of the load-bearing capacity and the vibration and noise signal of the construction machinery in real time;

[0042] A reference noise sensor, installed on the main structure of the construction machinery and having no direct mechanical contact with the material under test, is used to synchronously collect a second reference noise signal sequence generated by the vibration of the construction machinery. The acquisition process of the first mixed signal sequence and the second reference noise signal sequence is based on a unified high-precision clock source to achieve synchronous sampling.

[0043] The first mixed signal sequence and the second reference noise signal sequence are used to construct a multidimensional observation signal matrix. Preprocessing operations are performed on the multidimensional observation signal matrix, including centering and whitening. The centering process involves subtracting the time average value of each signal sequence in the observation signal matrix. The whitening process involves performing a linear transformation on the centered observation signal matrix using a whitening matrix, so that the transformed signal components are pairwise uncorrelated and have a variance of 1.

[0044] An independent component analysis algorithm based on maximizing non-Gaussianity is used to iteratively solve the preprocessed multidimensional observation signal matrix, and a source signal matrix containing multiple statistically independent source signal components is obtained. The source signal components include the target bearing capacity dynamic response source signal and the vibration noise source signal.

[0045] For each independent source signal component in the source signal matrix, calculate the higher-order statistical parameters of its signal waveform, where the higher-order statistical parameters are the kurtosis values ​​of the signal.

[0046] Based on the preset signal physical characteristic criteria, the kurtosis values ​​are compared and analyzed, and the independent source signal component with the largest positive kurtosis value is identified as the target bearing capacity dynamic response source signal. The signal physical characteristic criteria are that the dynamic response signal generated by the material when subjected to external load impact exhibits transient pulse characteristics, and its probability density distribution has a super Gaussian distribution feature with sharp peaks and heavy tails.

[0047] Based on the pure time series waveform of the identified target load-bearing capacity dynamic response source signal, the peak value of its signal amplitude is extracted, and the dynamic deformation modulus characterizing the load-bearing capacity of the material under test is calculated according to the peak value, the known externally applied load value, and the geometric dimensional parameters of the loaded structure.

[0048] In this step, the main measuring sensor is a high-sensitivity piezoresistive thin-film pressure sensor, which is encapsulated at the bottom of a rigid bearing plate and directly attached to the compacted surface of the road material to be tested. The sensor has a measurement range of 0 to 10 MPa, a sensitivity of not less than 10 mV / MPa, and a response frequency upper limit of not less than 5000 Hz, to ensure that it can fully capture the transient stress response of the material under impact load.

[0049] The rigid load-bearing plate is made of high-strength alloy steel, and its bottom surface is precision ground to ensure full contact with the material surface and uniform load transfer. The plate diameter is set at 300 mm and the thickness at 25 mm to meet the geometric requirements of standard load-bearing plate tests. The reference noise sensor is a triaxial microelectromechanical system accelerometer, which is mounted on the metal frame of the road roller or paver that generates vibration noise by magnetic adsorption or bolt fixing. Its measuring axis is consistent with the direction of the main mechanical vibration.

[0050] The accelerometer has a measurement range of ±50 times the gravitational acceleration, a bandwidth of no less than 100,000 Hz, and a noise density of less than 100 micrograms per square root of hertz. It can accurately record the vibration characteristics of the mechanical body under strong vibration environment without being disturbed by material reaction forces.

[0051] The synchronous sampling is achieved through a multi-channel synchronous data acquisition card. The time offset between each channel of the data acquisition card is less than 1 microsecond. Its built-in analog-to-digital converter has a resolution of no less than 24 bits, and the sampling frequency is set to no less than 20 kHz to ensure complete capture of the high-frequency components of the material's dynamic response signal and the broadband spectral characteristics of vibration noise. The data acquisition card integrates a high-stability temperature-compensated crystal oscillator as a clock source, with a frequency stability better than ten parts per million, ensuring phase consistency of each channel during long-term continuous sampling. The analog front end of the acquisition card includes a programmable gain amplifier and an anti-aliasing filter, with a gain range adjustable from 1 to 1000 times. The cutoff frequency is dynamically configured according to the actual signal bandwidth to prevent high-frequency noise from aliasing into the effective frequency band. All raw analog signals are isolated and amplified before entering the analog-to-digital converter to eliminate ground loop interference and common-mode noise.

[0052] The first mixed signal sequence and the second reference noise signal sequence are used to construct a multi-dimensional observation signal matrix. This matrix is ​​a two-dimensional array with the number of rows equal to the number of sampling points and the number of columns equal to the number of sensor channels, which is 2 in this embodiment. Preprocessing is performed on the multi-dimensional observation signal matrix. First, centering is performed, that is, each column of the signal sequence is subtracted from its corresponding time average value, so that the mean of the processed signal is 0, eliminating the influence of DC offset on subsequent statistical analysis. Then, whitening is performed, specifically implemented as follows:

[0053] The covariance matrix of the centered observed signal matrix is ​​calculated. This covariance matrix is ​​a second-order square matrix, and its elements represent the second-order statistical correlation between signals from different channels. Eigenvalue decomposition is performed on this covariance matrix to obtain an eigenvalue diagonal matrix and an eigenvector orthogonal matrix. The whitening matrix is ​​the product of the eigenvector orthogonal matrix and the eigenvalue diagonal matrix raised to the power of -1 / 2. The whitening matrix is ​​then left-multiplied by the centered observed signal matrix to obtain the whitened signal matrix. The covariance matrix of this whitened signal matrix is ​​an identity matrix, indicating that the components are uncorrelated and have a variance of 1, providing ideal input conditions for subsequent independent component analysis.

[0054] An independent component analysis algorithm based on maximizing non-Gaussianity is used to iteratively solve the whitened signal matrix. Specifically, this algorithm is a fast fixed-point iterative algorithm. The algorithm initializes a randomized solution mixing matrix with the same dimension as the number of observed signal channels. In each iteration, an update operation is performed on each column vector of the solution mixing matrix:

[0055] First, the product of the current column vector and the whitened observation signal matrix is ​​calculated to obtain the estimated source signal components. This estimated source signal component is then substituted into the derivative of a preset nonlinear function, chosen as the hyperbolic tangent function, whose derivative is the square of the hyperbolic tangent function minus the derivative. This derivative value is then multiplied element-wise by the whitened observation signal matrix, and the expected value is calculated to obtain a vector. Simultaneously, the expected value of the second derivative of this derivative is calculated, which is -2 times the hyperbolic tangent function multiplied by its derivative. The product of this expected value and the current column vector is subtracted from the aforementioned vector to obtain the updated column vector. Finally, this column vector is normalized to have a Euclidean norm of 1. This iterative process continues until the Frobenius norm between two adjacent iterations of the solution mixing matrix is ​​less than a preset convergence threshold, or the maximum number of iterations (500) is reached. Finally, the converged solution mixing matrix is ​​left-multiplied by the whitened observation signal matrix to obtain the source signal matrix, where each row corresponds to statistically independent source signal components.

[0056] For each independent source signal component in the source signal matrix, calculate the kurtosis value of its signal waveform. The kurtosis value is defined as the fourth central moment of the signal sequence divided by the square of its variance, and is used to measure the sharpness of the signal probability density distribution relative to a Gaussian distribution. For a discrete signal sequence of length N... Its peak The calculation formula is as follows:

[0057] ;

[0058] in This is the signal average. Since centralization processing was performed previously, If the value is 0, the formula can be simplified to the ratio of the square of the fourth moment to the square of the second moment. The theoretical kurtosis value of the Gaussian distribution is 3, and in actual calculations, excess kurtosis (kurtosis minus 3) is often used as the criterion. This invention uses the original kurtosis value as the criterion because the dynamic response signal of bearing capacity has obvious transient impulse characteristics, and its distribution exhibits a peak-heavy tail morphology, with a kurtosis value significantly greater than 3; while mechanical vibration noise is usually composed of multiple random processes superimposed, and according to the central limit theorem, its distribution approaches a Gaussian or sub-Gaussian morphology, with a kurtosis value close to or less than 3. Therefore, the calculated multiple kurtosis values ​​are numerically compared, and the source signal component with the largest kurtosis value is selected as the source signal of the target bearing capacity dynamic response. This criterion does not require prior knowledge, but only relies on the inherent statistical characteristics of the signal, and has strong robustness and adaptability.

[0059] Based on the pure time-series waveform of the identified target bearing capacity dynamic response source signal, the peak value of its signal amplitude is extracted. This peak value corresponds to the maximum stress response generated by the material under external impact load. The externally applied load value P is measured in real time by the hydraulic system pressure sensor or force sensor of the loading device, with an accuracy of not less than 0.5% of full scale. The radius r of the rigid bearing plate is a known geometric parameter and is set to 150 mm. The Poisson's ratio μ of the material is preset according to the material type, taking a value of 0.35 for asphalt mixtures and 0.25 for cement-stabilized crushed stone. Substituting the above parameters into the calculation formula of dynamic deformation modulus E:

[0060] ;

[0061] Where $d_{\text{max}}$ represents the maximum deformation extracted from the dynamic response source signal of the target bearing capacity. This deformation is obtained by converting the stress peak value through the stress-strain relationship, or directly measured by the integrated displacement sensor. The calculated dynamic deformation modulus E is the core indicator characterizing the material's bearing capacity; the larger its value, the better the material's compaction quality and the stronger its bearing capacity.

[0062] The load-bearing capacity detection device includes a sensor array module, a signal conditioning and synchronous acquisition module, a data processing module, and a result output and storage module. The sensor array module contains a main measurement sensor and a reference noise sensor, which are connected to the corresponding input channels of the signal conditioning and synchronous acquisition module via shielded twisted-pair cables.

[0063] The signal conditioning and synchronous acquisition module includes a preamplifier circuit, an anti-aliasing filter, a multi-channel synchronous analog-to-digital converter, and a high-precision clock distribution network, responsible for converting analog signals into high-fidelity digital signals. The data processing module is connected to the signal conditioning and synchronous acquisition module via a high-speed serial interface, and internally integrates a preprocessing unit, a blind source separation unit, a target signal identification unit, and a load-bearing capacity calculation unit.

[0064] The preprocessing unit performs centering and whitening operations, the blind source separation unit runs a fast fixed-point iterative algorithm, the target signal identification unit calculates the kurtosis value of each source signal component and executes the criterion logic, and the bearing capacity calculation unit performs numerical calculation of the dynamic deformation modulus. The result output and storage module includes an LCD screen, a large-capacity solid-state drive, and a universal serial bus interface for real-time display of detection results, storage of raw and processed data, and support for data interaction with a host computer or cloud platform.

[0065] In this embodiment, the data processing module is specifically a field-programmable gate array (FPGA) chip, specifically a Xilinx Kintex-7 series chip. The functions of the preprocessing unit, blind source separation unit, target signal recognition unit, and load-bearing capacity calculation unit are programmed using a hardware description language and embedded in the logic resources of the FPGA chip. The blind source separation unit is implemented in the chip as a hardware computing core based on a pipelined structure and a coordinate rotation digital computer algorithm, used for high-speed execution of operations such as vector inner product, matrix multiplication, nonlinear function lookup, and iterative control. The coordinate rotation digital computer algorithm is used to efficiently calculate the hyperbolic tangent function and its derivative, avoiding the use of time-consuming floating-point arithmetic units. The entire signal processing flow is fully parallelized at the hardware level, with a single detection cycle not exceeding 200 milliseconds, meeting the real-time feedback requirements of the construction site.

[0066] In actual deployment, the main measuring sensor is placed along with the bearing plate in the area of ​​the road surface to be tested, while the reference noise sensor is firmly installed at the location of the roller frame where vibration is most significant. After the detection process is initiated, the system automatically completes the entire process of signal acquisition, synchronization, preprocessing, blind source separation, target identification, and bearing capacity calculation, ultimately displaying the dynamic deformation modulus value and the pass / fail judgment result on the LCD screen. If the detected value is lower than a preset threshold, an alarm is triggered, guiding construction personnel to perform additional compaction in the area. All detection data is packaged and stored according to timestamp and geographic location information, supporting subsequent quality traceability and big data analysis.

[0067] This embodiment fundamentally solves the challenge of extracting weak load-bearing capacity signals under strong vibration and noise environments through the deep integration of physical sensing layout and advanced signal processing algorithms. The blind source separation technology does not rely on frequency domain assumptions and can preserve signal details in both the time and statistical domains, avoiding signal distortion caused by traditional filtering. The kurtosis-based target recognition mechanism fully utilizes the essential physical differences between load-bearing capacity response and mechanical noise, ensuring the accuracy of the recognition results. The hardware acceleration architecture guarantees the real-time performance of the algorithm, making this invention applicable to engineering scenarios with stringent compaction quality requirements, such as highways and airport runways, effectively improving the durability and safety of pavement engineering.

Claims

1. A method for testing the bearing capacity of road construction materials, characterized in that, include: The first mixed signal sequence is formed by the linear superposition of the dynamic response signal of bearing capacity and the vibration and noise signal of construction machinery by the main measurement sensor set on the surface of the material to be tested in real time. A reference noise sensor, installed on the main structure of the construction machinery and having no direct mechanical contact with the material under test, is used to synchronously collect a second reference noise signal sequence generated by the vibration of the construction machinery. The acquisition process of the first mixed signal sequence and the second reference noise signal sequence is based on a unified high-precision clock source to achieve synchronous sampling. The first mixed signal sequence and the second reference noise signal sequence are used to construct a multidimensional observation signal matrix. Preprocessing operations are performed on the multidimensional observation signal matrix, including centering and whitening. The centering process involves subtracting the time average value of each signal sequence in the observation signal matrix. The whitening process involves performing a linear transformation on the centered observation signal matrix using a whitening matrix, so that the transformed signal components are pairwise uncorrelated and have a variance of 1. An independent component analysis algorithm based on maximizing non-Gaussianity is used to iteratively solve the preprocessed multidimensional observation signal matrix, and a source signal matrix containing multiple statistically independent source signal components is obtained. The source signal components include the target bearing capacity dynamic response source signal and the vibration noise source signal. For each independent source signal component in the source signal matrix, calculate the higher-order statistical parameters of its signal waveform, where the higher-order statistical parameters are the kurtosis values ​​of the signal. Based on the preset signal physical characteristic criteria, the kurtosis values ​​are compared and analyzed, and the independent source signal component with the largest positive kurtosis value is identified as the target bearing capacity dynamic response source signal. The signal physical characteristic criteria are that the dynamic response signal generated by the material when subjected to external load impact exhibits transient pulse characteristics, and its probability density distribution has a super Gaussian distribution feature with sharp peaks and heavy tails. Based on the pure time series waveform of the identified target load-bearing capacity dynamic response source signal, the peak value of its signal amplitude is extracted, and the dynamic deformation modulus characterizing the load-bearing capacity of the material under test is calculated according to the peak value, the known externally applied load value, and the geometric dimensional parameters of the loaded structure.

2. The method for testing the bearing capacity of road construction materials according to claim 1, characterized in that, The main measuring sensor is a high-sensitivity piezoresistive thin-film pressure sensor, which is encapsulated in the bottom of a rigid bearing plate and directly attached to the compacted surface of the road material to be tested. The reference noise sensor is specifically a triaxial microelectromechanical system accelerometer, which is installed on the metal frame of the road roller or paver that generates vibration noise by means of magnetic adsorption or bolt fixation, and its measuring axis is consistent with the main vibration direction of the machine.

3. The method for testing the bearing capacity of road construction materials according to claim 2, characterized in that, The synchronous sampling is achieved through a multi-channel synchronous data acquisition card. The time offset between each channel of the data acquisition card is less than 1 microsecond, and its built-in analog-to-digital converter has a resolution of not less than 24 bits and a sampling frequency of not less than 20 kHz.

4. The method for testing the bearing capacity of road construction materials according to claim 3, characterized in that, The specific implementation method of the whitening process is as follows: First, the covariance matrix of the centered observed signal matrix is ​​calculated. Then, the covariance matrix is ​​decomposed into eigenvalues ​​to obtain the eigenvalue diagonal matrix and the eigenvector orthogonal matrix. The whitening matrix is ​​the product of the eigenvector orthogonal matrix and the eigenvalue diagonal matrix raised to the power of -1 / 2.

5. The method for testing the bearing capacity of road construction materials according to claim 4, characterized in that, The independent component analysis algorithm based on maximizing non-Gaussianity is specifically a fast fixed-point iterative algorithm. The algorithm initializes a random solution mixing matrix and updates the column vectors of the solution mixing matrix using the following iterative formula: First, the product of the current column vector and the whitened observation signal matrix is ​​calculated to obtain the estimated source signal component. Then, the estimated source signal component is substituted into the derivative of a preset nonlinear function, and the result is multiplied with the whitened observation signal matrix to obtain the expectation. The product of the expectation of the second derivative of the nonlinear function and the current column vector is subtracted to obtain the updated column vector. Finally, the updated column vector is normalized. The nonlinear function is selected as the hyperbolic tangent function.

6. The method for testing the bearing capacity of road construction materials according to claim 5, characterized in that, The formula for calculating the kurtosis value is the fourth central moment of the signal sequence divided by the square of its variance; The specific criteria for judging the physical characteristics of the signal are as follows: the calculated kurtosis values ​​are sorted, and the signal components with kurtosis values ​​much greater than 3 are identified as super-Gaussian signals, i.e., the dynamic response source signals of the target bearing capacity, while the signal components with kurtosis values ​​close to 0 or negative are identified as sub-Gaussian signals, i.e., vibration noise source signals.

7. The method for testing the bearing capacity of road construction materials according to claim 6, characterized in that, The formula for calculating the dynamic deformation modulus is: ,in For dynamic deformation modulus, The Poisson's ratio of the material, Peak force of externally applied load, The radius of the rigid bearing plate. The maximum deformation or maximum stress response amplitude of the material is extracted from the pure time-series waveform of the target bearing capacity dynamic response source signal.

8. The method for testing the bearing capacity of road construction materials according to claim 7, characterized in that, The rigid load-bearing plate is made of high-strength alloy steel, and its bottom surface is precision ground. The plate diameter is set at 300 mm and the thickness is 25 mm. The externally applied load value is measured in real time by the hydraulic system pressure sensor or force sensor of the loading device, and its accuracy is not less than 0.5% of the full scale.

9. A bearing capacity testing device for road construction materials, characterized in that, include: The sensor array module is used to acquire a first mixed signal sequence after superimposing the dynamic response signal of the load-bearing capacity and the vibration noise signal through the main measurement sensor, and to synchronously acquire a second reference noise signal sequence through the reference noise sensor. The signal conditioning and synchronous acquisition module is electrically connected to the sensor array module and is used to amplify and filter the first mixed signal sequence and the second reference noise signal sequence, and perform high-precision synchronous analog-to-digital conversion to output a digital multidimensional observation signal matrix. A data processing module, connected to the signal conditioning and synchronous acquisition module, integrates a preprocessing unit, a blind source separation unit, a target signal identification unit, and a bearing capacity calculation unit. The preprocessing unit performs centering and whitening operations on the multidimensional observation signal matrix. The blind source separation unit runs an independent component analysis algorithm based on maximizing non-Gaussianity to separate independent source signal components. The target signal identification unit identifies the target bearing capacity dynamic response source signal by calculating the kurtosis value of each source signal component and based on a preset criterion. The bearing capacity calculation unit calculates the final dynamic deformation modulus based on the target bearing capacity dynamic response source signal. The result output and storage module is connected to the data processing module and is used to display the calculated dynamic deformation modulus in real time, and to persistently store the original acquired signal, the separated source signal and the calculation result.

10. The bearing capacity testing device for road construction materials according to claim 9, characterized in that, The data processing module is specifically a field-programmable gate array (FPGA) chip. The functions of the preprocessing unit, blind source separation unit, target signal recognition unit, and load-bearing capacity calculation unit are programmed using a hardware description language and embedded in the logic resources of the FPGA chip.

Images