Noise reduction asphalt pavement maintenance decision method based on acoustic characteristic attenuation prediction

Abstract

This invention discloses a noise-reducing asphalt pavement maintenance decision-making method based on acoustic characteristic attenuation prediction, comprising: acquiring non-destructive testing signals of the porous asphalt pavement under test, including ultrasonic reflection signals; extracting acoustic feature indices for hierarchical screening, and performing inversion using a pre-constructed equivalent fluid physics model to obtain current microscopic acoustic parameters; calculating the acoustic safety margin of the pavement under test using an environmental acoustic model; acquiring historical microscopic acoustic parameter sequences, and performing feature direction decomposition with current parameters to deduce multi-dimensional independent evolution trends; and obtaining pavement maintenance decisions based on independent evolution trends and acoustic safety margins. This invention, through the combination of non-destructive testing and physical model inversion, can distinguish between the attenuation mechanisms of blockage and aging, supporting the shift from periodic maintenance to maintenance decisions oriented towards achieving acoustic environmental standards.

Description

Technical Field

[0001] This invention relates to the field of noise reduction pavement monitoring technology, and in particular to a noise reduction asphalt pavement maintenance decision-making method based on acoustic characteristic attenuation prediction. Background Technology

[0002] Porous noise-reducing asphalt pavement, through its interconnected pore network, effectively dissipates and absorbs the sound energy generated between tires and the road surface, thus significantly reducing traffic noise pollution. This pavement structure plays a crucial technical role in improving the acoustic environment quality around urban roads and highways. Maintaining the connectivity of pores and sufficient void volume is the physical basis for ensuring the long-term stability of the pavement's acoustic absorption performance. With increasing service life, accurately understanding the evolution of pavement pore conditions is of significant engineering and technical importance for extending the service life of noise-reducing pavements.

[0003] In monitoring the service performance of noise-reducing pavements, current technical practices mainly rely on two methods: one is the traditional in-situ coring method, which brings the pavement structure back to the laboratory for testing porosity and sound absorption coefficient; the other is using ground-penetrating radar or portable impedance tube equipment for on-site fixed-point scanning and acoustic measurement. Both methods have limitations in the spatial coverage of the data acquired, making it difficult to achieve large-scale, continuous, dynamic monitoring without disrupting normal traffic flow. Furthermore, current monitoring and assessment methods often simply attribute the decline in pavement sound absorption capacity to a single dust filling process, lacking a detailed distinction between different physical degradation mechanisms and a refined prediction of evolution trends when dealing with complex service attenuation phenomena.

[0004] Therefore, in the dynamic measurement environment of high-speed vehicle travel, existing pavement monitoring methods still have significant uncertainties in capturing the correspondence between changes in microscopic pore structure and the degradation of macroscopic acoustic performance, making it difficult to effectively transform transient sensor signals into reliable predictive indicators to support long-term maintenance planning. There is an urgent need to research a technical method that can overcome the limitations of detection resolution, accurately identify pavement damage mechanisms, and proactively guide pavement maintenance interventions. Summary of the Invention

[0005] The purpose of this invention is to provide a noise-reducing asphalt pavement maintenance decision-making method based on acoustic characteristic attenuation prediction, in order to solve the above-mentioned problems existing in the prior art.

[0006] Technical solution: A noise-reducing asphalt pavement maintenance decision-making method based on acoustic characteristic attenuation prediction, comprising:

[0007] Acquire non-destructive testing signals of the porous asphalt pavement to be tested, including ultrasonic reflection signals;

[0008] Based on the ultrasonic reflection signal, acoustic feature indexes are extracted for hierarchical screening, and combined with a pre-constructed equivalent fluid physics model for inversion, the current microscopic acoustic parameters of the porous asphalt pavement to be tested are obtained.

[0009] The acoustic safety margin of the porous asphalt pavement under test is calculated based on the current microscopic acoustic parameters and the pre-configured environmental acoustic model.

[0010] Obtain a pre-stored sequence of historical microacoustic parameters;

[0011] Based on the historical microacoustic parameter sequence and the current microacoustic parameters, the characteristic direction decomposition of the road surface deterioration mechanism is performed, and the multi-dimensional independent evolution trend is deduced.

[0012] Road maintenance decisions are derived based on independent evolution trends and acoustic safety margins.

[0013] Optionally, in the process of graded screening based on acoustic feature indices extracted from ultrasonic reflection signals and inverted using a pre-constructed equivalent fluid physics model to obtain the current microscopic acoustic parameters of the porous asphalt pavement to be tested, the graded screening includes:

[0014] The acoustic characteristic index is obtained by calculating the ratio of the reflected energy of the ultrasonic reflected signal in the high-frequency band to the reflected energy in the low-frequency band.

[0015] The acoustic feature index is compared with the pre-configured screening threshold;

[0016] If the acoustic characteristic index is higher than the pre-configured screening threshold, the step of inversion combined with the pre-built equivalent fluid physics model is triggered to obtain the current microscopic acoustic parameters.

[0017] If the acoustic feature index is not higher than the pre-configured screening threshold, the acoustic feature index is converted and estimated using the pre-stored feature mapping relationship to obtain the current microscopic acoustic parameters.

[0018] Optionally, the boundary frequency between the high-frequency band and the low-frequency band is pre-optimized and configured; the boundary frequency is pre-optimized and configured in the following way:

[0019] Multidimensional microscopic parameters are extracted from the equivalent fluid physics model and divided into clogging-related parameters and non-clogging-related parameters.

[0020] Calculate the analytical sensitivity of acoustic characteristic indices to each parameter in a multidimensional micro-parameter;

[0021] With the goal of maximizing the analytical sensitivity for clogging-related parameters and minimizing the analytical sensitivity for non-clogging-related parameters, a search is conducted within a preset available ultrasonic frequency band to determine the boundary frequency.

[0022] Optionally, the acoustic safety margin of the porous asphalt pavement under test is calculated based on the current microscopic acoustic parameters and a pre-configured environmental acoustic model, including:

[0023] By introducing pre-configured ground acoustic interaction coefficients, the current microacoustic parameters are converted into equivalent frequency band noise reduction for each analysis frequency band.

[0024] Substitute the equivalent frequency band noise reduction into the environmental acoustic model to perform spatial propagation simulation and obtain the expected equivalent sound level of the target sound receiving point for the predetermined roadside sound receiving point.

[0025] The pre-configured environmental functional zone benchmark limit is compared with the expected sound level at the receiving point, and the difference is used as the acoustic safety margin.

[0026] Optionally, the non-destructive testing signal also includes the measured near-field sound level acquired simultaneously in the same vehicle; the calculation of the acoustic safety margin of the porous asphalt pavement under test also includes closed-loop verification for prediction accuracy:

[0027] Based on the current microacoustic parameters, calculate the expected near-field equivalent sound level in the tire-road interaction area;

[0028] The expected near-field equivalent sound level is compared with the measured near-field sound level to obtain the verification deviation value;

[0029] When the calibration deviation exceeds the preset tolerance range, the preset safety margin is deducted from the calculated acoustic safety margin, or a recalibration trigger command for the equivalent fluid physics model is generated.

[0030] Optionally, feature orientation decomposition of pavement deterioration mechanisms is performed based on historical microacoustic parameter sequences and current microacoustic parameters, including:

[0031] Based on pre-constructed unblocked state parameters and extremely blocked state parameters, the blocking direction vector in the multidimensional physical space is determined;

[0032] Based on the record of parameter changes before and after cleaning in the historical microacoustic parameter sequence, the aging direction vector in multidimensional physical space is determined.

[0033] Calculate the micro-parameter increments of the current micro-acoustic parameters and the parameters of the previous period in the historical micro-acoustic parameter sequence;

[0034] By geometrically projecting the micro-parameter increments onto the blockage direction vector and the aging direction vector respectively, the blockage evolution component and the aging evolution component of the stripping degradation mechanism are obtained.

[0035] Optionally, the pre-built equivalent fluid physics model can be calibrated and corrected through the following offline build steps:

[0036] Based on the physical dimensional properties of the porous asphalt skeleton, the upper limit of the dispersion frequency for the equivalent fluid physical model is determined.

[0037] The highest frequency of the ultrasonic reflection signal used for offline calibration and online detection is limited to not exceeding the upper limit of the diffusion frequency;

[0038] By comparing the measured absorption coefficient of the physical specimen in the audible frequency band with the theoretical absorption coefficient extrapolated across the frequency band by the equivalent fluid physical model, a frequency correlation correction function is generated to compensate for the long-wavelength approximation failure, and this function is built into the pre-constructed equivalent fluid physical model.

[0039] Optionally, the non-destructive testing signal also includes an infrared temperature signal; in the step of inversion combined with a pre-built equivalent fluid physics model, the infrared temperature signal is used to obtain the road surface temperature field, and the temperature-sensitive physical parameters in the equivalent fluid physics model are compensated in real time based on the road surface temperature field.

[0040] Optionally, acquiring the non-destructive testing signal of the porous asphalt pavement to be tested also includes:

[0041] Simultaneously acquire vehicle driving status signals and spatial positioning signals;

[0042] Doppler frequency shift compensation preprocessing is performed on ultrasonic reflection signals based on vehicle driving status signals;

[0043] Spatial-temporal registration of nondestructive testing signals based on spatial positioning signals.

[0044] Optionally, pavement maintenance decisions may include differentiated cleaning instructions; generating differentiated cleaning instructions includes:

[0045] Compare the surface blockage state with the deep blockage state in the current microacoustic parameters that have information on the depth and shallow stratification.

[0046] If the severity of surface clogging is higher than that of deep clogging, a low-pressure flushing recommendation is generated in the differentiated cleaning instruction;

[0047] If the severity of deep blockage is higher than that of surface blockage, a high-pressure adsorption cleaning recommendation will be generated in the differentiated cleaning instruction.

[0048] A noise-reducing asphalt pavement maintenance decision system based on acoustic characteristic attenuation prediction includes:

[0049] At least one processor; and,

[0050] A memory communicatively connected to at least one of the processors; wherein,

[0051] The memory stores instructions that can be executed by the processor to implement the steps of the noise reduction asphalt pavement maintenance decision method based on acoustic characteristic attenuation prediction as described above.

[0052] Beneficial effects: By combining non-destructive testing with physical model inversion, this invention can distinguish between the decay mechanisms of blockage and aging, supporting the shift from periodic maintenance to maintenance decisions oriented towards achieving acoustic environmental standards. Attached Figure Description

[0053] Figure 1 This is a schematic diagram of a noise reduction asphalt pavement maintenance decision-making method based on acoustic characteristic attenuation prediction provided in an embodiment of this application.

[0054] Figure 2 This is a schematic diagram of the hierarchical screening process provided in the embodiments of this application.

[0055] Figure 3 This is a schematic diagram of the process for determining the boundary frequency provided in the embodiments of this application.

[0056] Figure 4 This is a schematic diagram of the process for calculating the acoustic safety margin of the porous asphalt pavement under test based on the current microscopic acoustic parameters and the pre-configured environmental acoustic model, provided in the embodiments of this application.

[0057] Figure 5 This is a schematic diagram of the closed-loop verification process for prediction accuracy provided in the embodiments of this application. Detailed Implementation

[0058] Example 1

[0059] Describe a noise reduction asphalt pavement maintenance decision-making method based on acoustic characteristic attenuation prediction, such as Figure 1 As shown, the method includes the following steps:

[0060] Step 101: Obtain the non-destructive testing signal of the porous asphalt pavement to be tested. The non-destructive testing signal includes ultrasonic reflection signal.

[0061] Specifically, the non-destructive testing signal serves as the raw input characterizing the physical state of the road surface, manifested as sequential data containing timestamps and spatial coordinates. The ultrasonic reflection signal is generated by an air-coupled ultrasonic transducer array mounted on the underside of the inspection vehicle, emitting broadband ultrasonic pulses to the road surface and receiving the reflected echoes. During data acquisition, the inspection vehicle's speed is constrained between 20 km / h and 40 km / h, the transducer installation height is set within the range of 200 mm to 300 mm from the road surface, and physical buffering is achieved through a spring-damped vibration isolation bracket. Simultaneously, a laser rangefinder sensor monitors distance changes in real time.

[0062] Non-contact vehicle-mounted dynamic detection methods can quickly and continuously acquire the sound field reflection characteristics of long-distance road sections without damaging the porous asphalt pavement structure.

[0063] Furthermore, under the multi-sensor fusion architecture, the non-destructive testing signal can also simultaneously include the electromagnetic reflection signal acquired by ground penetrating radar and the surface temperature field signal acquired by infrared camera, providing multi-dimensional physical information for subsequent parameter compensation and hierarchical inversion.

[0064] Step 102: Based on the ultrasonic reflection signal, the acoustic feature index is extracted for hierarchical screening, and combined with the pre-constructed equivalent fluid physics model for inversion to obtain the current microscopic acoustic parameters of the porous asphalt pavement to be tested.

[0065] Specifically, the acoustic characteristic index is a quantitative indicator used to characterize subtle changes in the initial stage of pore blockage. Its value reflects the relative change in reflected energy between high-frequency and low-frequency sound waves in porous media. The equivalent fluid physics model is an acoustic theoretical model used to describe the viscous and thermal dissipation mechanisms of fluids in porous frameworks. Currently, microscopic acoustic parameters are specifically represented as multidimensional physical quantity vectors, including porosity, static flow resistance, tortuosity, viscous characteristic length, and thermal characteristic length.

[0066] In this step, the acoustic characteristic index is first extracted by frequency domain transformation of the ultrasonic reflection signal. A graded screening mechanism is then used to determine the index. When the index indicates that the road surface is in a low-congestion state, the parameters are directly estimated using empirical mapping relationships; when the index indicates that the road surface is significantly congested, an inversion calculation is triggered, that is, the measured acoustic signal is input into a pre-established equivalent fluid physics model, and the current microscopic acoustic parameters that match the theoretical output with the measured input are obtained through inverse optimization.

[0067] By combining screening and inversion mechanisms, the system avoids overloading the onboard computing power caused by blindly calculating the full amount of data, and accurately transforms the macroscopic sound wave reflection phenomenon into specific physical parameters describing the internal structure of the pores through rigorous deduction of the microscopic physical model.

[0068] Step 103: Calculate the acoustic safety margin of the porous asphalt pavement to be tested based on the current microscopic acoustic parameters and the pre-configured environmental acoustic model.

[0069] The environmental acoustic model consists of a tire and road noise source intensity prediction module and a sound wave spatial propagation attenuation module. The acoustic safety margin is represented by the difference between the expected sound pressure level within a predetermined acoustic environment functional zone and the reference sound pressure level defined by national standards.

[0070] Since porosity and flow resistance at the microscopic level are difficult to directly use as assessment indicators for highway maintenance, this step utilizes the laws of acoustic propagation to forward extrapolate the microscopic parameters into an equivalent frequency band noise reduction, which is then substituted into the environmental acoustic model to calculate the expected received noise at the roadside sound receiving point. By comparing the difference between this expected received noise and the environmental protection requirements, the acoustic safety margin can be obtained.

[0071] This step enables the conversion of data from microscopic material parameters to macroscopic engineering regulatory compliance, providing intuitive numerical evidence for subsequent maintenance triggering.

[0072] Step 104: Obtain the pre-stored historical microacoustic parameter sequence.

[0073] Specifically, the historical microacoustic parameter sequence is time-series data accumulated by the system through repeated measurements and inversions of the same road surface section over multiple past detection cycles. This sequence data is stored in an onboard database or cloud server and includes not only the five-dimensional microacoustic parameters (multi-dimensional microparameters) for each inspection, but also the corresponding timestamps, cumulative traffic equivalent axle counts, and parameter jump records before and after road historical cleaning. This long-term observation data provides a temporal reference benchmark for exploring the dynamic process of road surface performance degradation.

[0074] Step 105: Based on the historical microacoustic parameter sequence and the current microacoustic parameters, the characteristic direction decomposition of the road surface deterioration mechanism is performed, and the multi-dimensional independent evolution trend is deduced.

[0075] In this step, the feature direction is decomposed within a multidimensional vector space composed of microscopic acoustic parameters, breaking down the changes in parameters over time into independent components with different physical meanings. The acoustic performance degradation of porous asphalt pavements is typically caused by two mechanisms: reversible pore blockage and irreversible asphalt aging and skeleton wear. By extracting abrupt changes in the historical microscopic acoustic parameter sequence before and after cleaning, the vector direction representing aging can be defined; by extracting gradual changes from no blockage to severe blockage, the vector direction representing blockage can be defined. By geometrically projecting the increments of current microscopic acoustic parameters and historical data in two directions, the mixed degradation phenomena can be separated, and independent evolutionary trends pointing to the future can be obtained through fitting and extrapolation. This solves the technical problem of traditional methods misjudging aging as blockage, leading to indiscriminate cleaning.

[0076] Step 106: Based on the independent evolution trend and acoustic safety margin, a pavement maintenance decision is obtained. That is, a comprehensive judgment on the timing and method of maintenance is made based on the independent evolution trend and acoustic safety margin to generate a pavement maintenance decision.

[0077] Specifically, road maintenance decisions include the triggering time for road maintenance, a priority list of cleaning sections, and differentiated physical operation methods. Combining the predictive data obtained from previous steps, the system will simulate the effects of various maintenance methods at a certain point in the future.

[0078] For example, the degree of parameter recovery after cleaning can be inferred based on the blockage evolution trend, and then the acoustic performance after recovery can be judged in combination with the aging evolution trend to determine whether it can meet the bottom line requirements of acoustic safety margin.

[0079] Based on comprehensive predictions, the system generates specific execution instructions and outputs them to the road maintenance department. This enables a shift from routine, indiscriminate maintenance to precise intervention focused on noise reduction performance.

[0080] According to another aspect of this application, a noise-reducing asphalt pavement maintenance decision system based on acoustic characteristic attenuation prediction is provided, comprising:

[0081] At least one processor; and a memory communicatively connected to at least one of the processors; wherein,

[0082] The memory stores instructions that can be executed by the processor to implement the noise reduction asphalt pavement maintenance decision method based on acoustic characteristic attenuation prediction as described in this invention.

[0083] Example 2

[0084] Based on the above embodiments, the online detection front-end preprocessing and environmental compensation mechanism will be further described in detail.

[0085] In one possible implementation, the step of acquiring the non-destructive testing signal of the porous asphalt pavement to be tested further includes the following steps:

[0086] Step 201: Simultaneously acquire vehicle driving status signals and spatial positioning signals.

[0087] Specifically, the vehicle driving status signal refers to the set of dynamic parameters of the inspection vehicle's movement in physical space, typically including instantaneous speed, acceleration, and direction of travel. This signal can be obtained by reading data from the vehicle's on-board diagnostic system interface or wheel speed sensor data. The spatial positioning signal is used to determine the vehicle's absolute position in the geographic coordinate system and can be obtained through the fusion of a Global Navigation Satellite System receiver and an inertial measurement unit. Acquiring both signals provides objective physical reference parameters for subsequent signal frequency domain correction and multi-source data alignment.

[0088] Furthermore, regarding the specific hardware deployment for acquiring ultrasonic reflection signals, a streamlined wind noise shield is configured around the air-coupled ultrasonic transducer to counteract airflow interference during vehicle movement. This shield alters the airflow boundary layer distribution on the transducer surface, reducing aerodynamic noise. Simultaneously, the transducer's transmission drive source outputs a linear frequency modulated pulse signal, rather than a traditional single-frequency short pulse.

[0089] Through pulse compression processing at the subsequent receiving end, additional processing gain can be obtained by utilizing the time-bandwidth product, and more than 20 pulse superposition averaging operations can be performed within a single spatial measurement point to ensure that the signal-to-noise ratio meets the effective calculation requirements under high vehicle speed and strong wind noise environments.

[0090] Step 202: Perform Doppler frequency shift compensation preprocessing on the ultrasonic reflection signal based on the vehicle driving state signal to obtain the compensated ultrasonic reflection signal.

[0091] Doppler shift refers to the physical phenomenon where the frequency of the received ultrasonic reflected wave deviates from the original transmission frequency due to the relative motion between the vehicle-mounted transducer and the relatively stationary road surface. Since the acoustic characteristic index used for graded screening in this method highly depends on accurate calculation of high-frequency and low-frequency energy integration, Doppler shift causes frequency band crossings in the spectral energy, resulting in systematic calculation errors. Therefore, it is necessary to resample the received signal or perform frequency domain coordinate scaling using the synchronously acquired instantaneous vehicle speed.

[0092] Specifically, the calculation logic for Doppler frequency shift compensation is expressed as follows:

[0093] f _comp =f _measure *c _0 / (c _0 -v _proj );

[0094] Among them, f _comp f is the compensated true reflection frequency. _measure c is the measured frequency acquired by the receiving end. _0 v is the speed of sound in air at the current ambient temperature. _proj This is the projection component of the velocity vector in the vehicle's driving status signal onto the ultrasonic wave propagation path.

[0095] By performing this compensation operation, the spectral distortion caused by relative motion can be eliminated, and the true acoustic reflection spectrum of the road surface in the static reference frame can be restored.

[0096] Step 203: Perform spatial-temporal registration on the non-destructive testing signal based on the spatial positioning signal to obtain the registered non-destructive testing signal, which is a non-destructive testing signal with a unified spatial coordinate index.

[0097] In this embodiment, the non-destructive testing signal is a typical multi-sensor asynchronous heterogeneous data stream. Since the sampling frequencies and hardware delay times of the ultrasonic transducer, ground-penetrating radar, and near-field microphone are different, direct fusion will result in spatial misalignment.

[0098] Based on this, the timestamps and mileage markers provided by the spatial positioning signals are used as unified index coordinate axes. Linear interpolation or nearest neighbor interpolation algorithms are employed to resample various types of sensor data onto the same spatial grid node. For example, if the system's unified output interval is set to 0.5m, then the data sequences of all sensors are mapped and aggregated into this 0.5m discrete road segment interval, ensuring that the data sources used for subsequent micro-parameter inversion are physically identical.

[0099] The non-destructive testing signal also includes an infrared temperature signal. In the step of inversion combined with a pre-built equivalent fluid physical model, the infrared temperature signal is used to obtain the surface temperature field of the road surface. Based on the surface temperature field of the road surface, the temperature-sensitive physical parameters in the equivalent fluid physical model are compensated in real time to obtain the temperature-compensated physical parameters. Inversion is then performed based on the temperature-compensated physical parameters.

[0100] Specifically, the infrared temperature signal is acquired by scanning the road surface radiation downwards using an onboard infrared thermal imager. In the actual service environment of porous asphalt pavements, the surface temperature difference between blocked and connected pores under normal temperature and dry conditions is relatively weak and insufficient to directly serve as a sensing basis for clogging determination. However, temperature changes have a decisive impact on air velocity and aerodynamic viscosity, and these two physical quantities are the basis for forward extrapolation of the equivalent fluid physics model. Ignoring temperature fluctuations will lead to severe drift in the microscopic acoustic parameters obtained through inversion.

[0101] Based on this, the real-time temperature compensation process for temperature-sensitive physical parameters is achieved through the following calculation process:

[0102] c _0_T =331.3*(T _surface / 273.15) 0.5 ;

[0103] Among them, c _0_T The dynamic air speed of sound is updated after compensation, 331.3 is the reference speed of sound at 0℃, and T _surface 273.15 is the absolute surface temperature of the road surface extracted from the infrared temperature signal, 273.15 is the conversion constant from Celsius to Kelvin temperature, and 0.5 indicates that the square root of this ratio is performed.

[0104] η _T =1.716*10 -5 *(T _surface / 273.15) 0.756 ;

[0105] Where, η _T The updated dynamic aerodynamic viscosity after compensation is 1.716*10. -5 The reference dynamic viscosity is 0℃, and 0.756 is the empirical fit index. This is achieved by calculating the T value corresponding to each spatial measurement point. _surface Substituting the above calculation process, the system can update the reference values ​​of sound velocity and viscosity in the inversion objective function in real time.

[0106] In some optional implementations, the road surface moisture content exhibits spatial heterogeneity during conditions such as the evaporation phase after rainfall or the application of de-icing agents in winter. In these situations, the thermal conductivity of the blocked area changes due to water retention, resulting in a local temperature difference of 1-2°C between the blocked area and the connected area. When the overall temperature field variance exceeds a preset threshold, the system can extract local temperature gradient features from the infrared temperature signal and input them as an auxiliary evidence matrix into the subsequent hierarchical inversion fusion module to improve the reliability of blockage location under certain climatic conditions. If the infrared sensor fails due to mud and water splashing, the system automatically switches to using data from the vehicle's ambient temperature sensor for global smoothing compensation.

[0107] Example 3

[0108] This paper further details the offline construction and calibration methods for the pre-built equivalent fluid physics model and its associated database. Through offline experiments and theoretical modeling, a high-precision parameter inversion basis and predictive compensation function are provided for online detection.

[0109] In some embodiments, the pre-built equivalent fluid physics model is calibrated and corrected through the following offline construction steps:

[0110] Step 301: Based on the physical size properties of the porous asphalt skeleton, determine the upper limit of the dispersion frequency for the equivalent fluid physical model.

[0111] Specifically, the equivalent fluid physics model can be the Johnson-Champoux-Allard model, the Biot model, or other equivalent medium acoustic models. In this embodiment, the Johnson-Champoux-Allard model, abbreviated as JCA model, is preferred. The physical applicability of this model is based on the long-wavelength approximation condition, meaning the sound wave wavelength must be significantly larger than the characteristic size of the porous medium framework. When the sound wave frequency is too high, significant Rayleigh scattering occurs in the porous framework, causing the equivalent medium assumption of this model to fail.

[0112] To define this physical boundary, this embodiment calculates the first elastic resonance frequency of the porous asphalt layer based on the gradation characteristics of the porous asphalt mixture, and determines the physical safety upper limit of the detection frequency accordingly.

[0113] The upper limit of the diffusion frequency is calculated as follows:

[0114] f _disp =c _agg / (4*D _max );

[0115] Among them, f _disp c is the upper limit of the diffusion frequency. _agg ρ is the longitudinal wave velocity of the aggregate, 4 is the proportionality constant determined based on the quarter-wavelength resonance condition, and D is the longitudinal wave velocity of the aggregate. _max This represents the nominal maximum particle size of porous asphalt mixtures. In practical applications, if the nominal maximum particle size D is selected... _max The OGFC-13 mixture is an open-graded, drainage-type asphalt wearing course with a thickness of 13.2 mm, and the longitudinal wave velocity of the aggregate is c. _agg If the value is 3500 m / s, then the calculated f _disp It is approximately 66,000 Hz.

[0116] Step 302: The highest frequency of the ultrasonic reflection signal used for offline calibration and online detection is limited to not exceeding the upper limit of the diffusion frequency.

[0117] Specifically, to ensure that the physical model used in the inversion calculation is in the linear steady-state region, the system strictly adheres to the physical constraints determined in step 301 when selecting the hardware transducer and setting the output bandwidth of the signal generator. For example, if the calculated upper limit of the dispersion frequency is 66 kHz, the center frequency of the selected air-coupled ultrasonic transducer is typically set to 40 kHz or 50 kHz to ensure that the effective spectral energy of the ultrasonic reflection signal is concentrated in the frequency domain below this upper limit. This method is used to eliminate systematic prediction biases in the model caused by high-frequency scattering and to ensure the physical legitimacy of the microscopic parameter inversion.

[0118] Step 303: Compare the measured sound absorption coefficient of the physical specimen in the audible frequency band with the theoretical sound absorption coefficient extrapolated across the frequency band by the equivalent fluid physical model, generate a frequency correlation correction function to compensate for the long-wavelength approximation failure, and embed the frequency correlation correction function into the pre-built equivalent fluid physical model.

[0119] In this embodiment, there is a scale difference between the inversion environment of the ultrasonic frequency band and the predicted environment of the audible frequency band, and direct extrapolation may produce residual bias. To address this, the system prepares physical specimens covering different degrees of blockage in the offline stage, uses the standing wave tube method to obtain the measured sound absorption coefficient in the 200Hz to 2000Hz frequency band, and compares it with the theoretical value calculated by the JCA model using ultrasonic inversion parameters.

[0120] The frequency-dependent correction function is calculated as follows:

[0121] β(f,Cr)=α _lab (f) / α_JCA (f,p);

[0122] Where β(f,Cr) is the frequency-related correction function, f is the sound wave frequency, Cr is the physical blockage rate of the specimen calibrated manually, and α _lab (f) represents the measured sound absorption coefficient of the physical specimen in the audible frequency range, α. _JCA (f,p) represents the theoretical sound absorption coefficient calculated by the equivalent fluid physics model based on the inversion parameter p. Correction factors at different frequencies and blockage rates are stored to form a coefficient matrix used for real-time compensation of model prediction errors during the online phase.

[0123] Furthermore, to achieve millisecond-level online inversion, this embodiment also involves an offline learning and proxy model construction of a congestion trajectory manifold. The congestion trajectory manifold refers to the characteristic curve that evolves from an uncongested state to a highly congested state in a five-dimensional microscopic acoustic parameter space.

[0124] The parameterized representation of the congestion trajectory manifold is as follows:

[0125] p(ξ)=∑ n=0 N_b c _n *B _n (ξ);

[0126] Where p(ξ) is the parameter estimation vector corresponding to the manifold, ∑ n=0 N_b This indicates that the basis function index n ranges from 0 to N. _b Summation, c _n B is a preset five-dimensional vector coefficient. _n (ξ) is the cubic spline basis function corresponding to the normalized congestion severity parameter ξ. N _b The number of cubic spline basis functions is reduced by 1. Its specific value is determined by the blockage rate sampling density during the offline calibration stage and can be adaptively selected according to the fitting accuracy requirements. The value range of ξ is 0~1, where 0 corresponds to the unblocked state and 1 corresponds to the preset maximum measured blockage rate.

[0127] Based on the aforementioned manifold representation, the system performs equally spaced grid sampling on the parameter ξ, pre-calculates the spectral characteristics of the reflection coefficient corresponding to each sampling point, and stores them as a two-dimensional lookup table (surrogate model lookup table). During online detection, spline interpolation matching is performed in the two-dimensional lookup table, which reduces the original high-dimensional inversion search problem to a one-dimensional localization problem for a single variable ξ, thereby improving the execution efficiency of the algorithm.

[0128] Example 4

[0129] The specific process of obtaining microscopic acoustic parameters through graded screening and inversion is further explained in detail.

[0130] In one possible implementation, the graded screening process, which involves extracting acoustic feature indices based on ultrasonic reflection signals for graded screening and then performing inversion using a pre-constructed equivalent fluid physics model to obtain the current microscopic acoustic parameters of the porous asphalt pavement under test, includes the following steps:

[0131] The acoustic characteristic index is obtained by calculating the ratio of the reflected energy of the ultrasonic reflected signal in the high-frequency band to the reflected energy in the low-frequency band. This acoustic characteristic index is then compared with a pre-configured screening threshold. If the acoustic characteristic index is higher than the pre-configured threshold, an inversion step using a pre-built equivalent fluid physics model is triggered to obtain the current microscopic acoustic parameters. If the acoustic characteristic index is not higher than the pre-configured threshold, the acoustic characteristic index is converted and estimated using a pre-stored feature mapping relationship to obtain the current microscopic acoustic parameters. Figure 2 As shown.

[0132] Specifically, pore blockage in porous asphalt pavements typically exhibits a physical process of gradual expansion from microscopic necks to larger pores. High-frequency ultrasound, with its shorter wavelength, is more sensitive to changes in small-scale pore structure; low-frequency ultrasound, with its longer wavelength, is more sensitive to changes in macroscopic porosity. Therefore, by calculating the ratio of high-frequency reflected energy to low-frequency reflected energy, an acoustic characteristic index that is more sensitive to the initial stage of blockage can be constructed.

[0133] In the specific calculation process, the high-frequency reflected energy is obtained by integrating or discretely summing the square of the reflected spectrum amplitude within a preset high-frequency integration interval; the low-frequency reflected energy is obtained by integrating or discretely summing the square of the reflected spectrum amplitude within a preset low-frequency integration interval. Dividing the high-frequency reflected energy by the low-frequency reflected energy yields the acoustic characteristic index.

[0134] Furthermore, to significantly reduce the computational load on real-time vehicle-mounted computing, this embodiment introduces a cascaded filtering logic. The pre-configured filtering threshold can be obtained through offline testing and is typically set to a value corresponding to the initial warning level of the road surface. Specifically, the filtering threshold is calibrated so that the miss rate—the probability that a significant blockage is actually misjudged as safe—does not exceed a preset acceptable upper limit. This is achieved by repeatedly collecting acoustic characteristic indices on standard specimens with different levels of blockage and statistically analyzing their distribution to determine the minimum threshold that satisfies the constraint.

[0135] When the inspection vehicle travels at 40 km / h, it needs to process a large number of spatial measurement points per second. Since the calculation of this acoustic characteristic index only involves basic algebraic operations on frequency domain energy ratios, the latency of a single calculation is extremely low. By comparing it with a screening threshold, most safe road sections in a low-congestion or non-congestion state can be screened out.

[0136] For this section of the road, the system does not initiate computationally complex physical model inversion. Instead, it directly calls pre-stored feature mapping relationships for one-dimensional mapping. These feature mapping relationships can be established through multinomial regression fitting, directly converting acoustic feature indices into corresponding current microscopic acoustic parameters. Only when this index exceeds a threshold does the system trigger a complete physical model inversion for the data frame corresponding to that specific spatial coordinate. This mechanism ensures congestion detection sensitivity while eliminating 70% to 90% of redundant inversion calculations.

[0137] The boundary frequency between the high-frequency and low-frequency bands is pre-optimized and configured. The boundary frequency is pre-optimized and configured in the following way:

[0138] Multidimensional microscopic parameters are extracted from the equivalent fluid physics model and divided into clogging-related and non-clogging-related parameters. The analytical sensitivity of acoustic characteristic indices to each parameter in the multidimensional microscopic model is calculated. With the goal of maximizing the analytical sensitivity for clogging-related parameters and minimizing the analytical sensitivity for non-clogging-related parameters, a search is conducted within a preset available ultrasonic frequency band to determine the boundary frequency, such as... Figure 3 As shown.

[0139] Among them, multidimensional microscopic parameters encompass all independent variables describing the acoustic behavior of porous media. Considering actual physical wear and contamination processes, the physical filling of pores primarily leads to a decrease in open porosity, an increase in static flow resistance, and a shortening of the viscous feature length; therefore, these three parameters are classified as clogging-related parameters. In contrast, the tortuosity and thermal feature length of the skeleton are less affected by dust filling and are more influenced by the natural aging and hardening of the skeleton itself; therefore, they are classified as non-clogging-related parameters.

[0140] Fixed frequency band allocation methods are often difficult to adapt to porous asphalt pavements with different gradations. This embodiment provides an adaptive frequency band optimization method. By substituting the forward analytical solution of the aforementioned equivalent fluid physics model into the definition of the acoustic characteristic index, the analytical sensitivity of the index to any parameter can be obtained. Its calculation logic can be expressed as follows:

[0141] S pj =(1 / E L )*D H -(E H / E 2 L )*D L ;

[0142] Among them, S pj E represents the analytical sensitivity of the acoustic characteristic index to the j-th microscopic parameter. H E represents the theoretical high-frequency reflected energy. L For the theoretical low-frequency reflected energy, D HD is the integral term of the partial derivative of the square of the theoretical high-frequency reflection spectrum amplitude with respect to the j-th microscopic parameter. L This is the integral term of the partial derivative of the square of the theoretical low-frequency reflection spectrum amplitude with respect to the j-th microscopic parameter.

[0143] Within a pre-defined available ultrasonic frequency band, the band is divided into high and low portions using a candidate boundary frequency. This candidate frequency is then substituted into the aforementioned sensitivity model. The sum of the absolute sensitivity values ​​of all blockage-related parameters is used as the numerator of the objective function, while the sum of the absolute sensitivity values ​​of all non-blockage-related parameters, superimposed with a small constant, is used as the denominator of the objective function. A one-dimensional bounded optimization algorithm is used to find the optimal boundary frequency within the available frequency band, maximizing this ratio. This process is performed offline, and the optimized boundary frequency is subsequently incorporated into the graded screening step to guide the online calculation of the energy ratio.

[0144] Specifically, by combining a pre-constructed equivalent fluid physics model with inversion, the current microscopic acoustic parameters of the porous asphalt pavement under test are obtained, including:

[0145] Obtain a pre-constructed univariate choking trajectory manifold characterizing the severity evolution, and a surrogate model lookup table pre-calculated based on an equivalent fluid physics model; extract the spectral features of the ultrasonic reflection signal and match the spectral features with the surrogate model lookup table; perform dimensionality reduction interpolation search within the one-dimensional variable space defined by the univariate choking trajectory manifold to locate the optimal severity variable value with the minimum matching error; inversely map the optimal severity variable value back to multidimensional physical parameters (multidimensional microscopic parameters) to obtain the current microscopic acoustic parameters.

[0146] Specifically, to locate the optimal severity variable value with the minimum matching error, one could also perform a dimensionality reduction interpolation search within a one-dimensional variable space defined by the univariate congestion trajectory manifold based on the spectral matching error between the spectral features and the surrogate model lookup table, thereby locating the optimal severity variable value with the minimum spectral matching error.

[0147] Specifically, traditional Bayesian inversion requires extensive matrix iterations in a five-dimensional parameter space, often taking more than one second on automotive embedded devices, which is insufficient for high-density spatial dynamic representation. This embodiment overcomes this computational barrier by introducing topological dimensionality reduction constraints in a multi-dimensional physical space.

[0148] The univariate clogging trajectory manifold is represented as a smooth curve traversing a five-dimensional parameter space, controlled by a univariate severity ξ. The value range of ξ is normalized to 0–1. As ξ increases from 0 to 1, the five-dimensional parameters on the manifold synchronously evolve along the physical trajectory from unclogging to extreme clogging. Before system deployment, a two-dimensional surrogate model lookup table containing the amplitude and phase of the reflection spectrum is pre-generated by inputting discrete ξ sampling points into an equivalent fluid physics model for forward extrapolation.

[0149] When online detection initiates inversion, the system performs Fourier transform on the measured ultrasonic reflection signal to extract the spectral features of the reflection coefficients. A cubic spline interpolation algorithm is then used to continuously read data from the surrogate model lookup table. The optimization process employs a one-dimensional numerical optimization algorithm to search for the minimum point of the target residual function. The design method of the target residual function is as follows:

[0150] Loss _ξ =∑ _i [(|R _field_i |-|R _table_i_ξ |) 2 ]+w _phase *∑ _i [(P _field_i -P _table_i_ξ ) 2 ];

[0151] Among them, Loss _ξ ∑ is the total residual value under the predetermined severity ξ. _i R represents the summation over all discrete sampling frequency points. _field_i R is the amplitude of the measured reflection spectrum at the i-th frequency point. _table_i_ξ To find the amplitude of the i-th frequency point derived from the severity ξ in the surrogate model lookup table, w _phase P is the preset phase residual weighting constant. _field_i For the measured phase angle, P _table_i_ξ The deduced phase angle is looked up in the table for the surrogate model.

[0152] Phase residual weighting constant w _phase The value used to balance the contributions of the amplitude residual term and the phase residual term to the objective function can be adaptively determined based on the actual signal-to-noise ratio of the sensor and the relative relationship between the amplitude and phase measurement accuracy. In one optional implementation, the value is 0.1.

[0153] Loss is calculated repeatedly in a one-dimensional variable space. _ξ By continuously narrowing the search interval, the optimal severity variable value can be quickly located. This process reduces the computational complexity from inverting a high-dimensional matrix in the complex field to scalar table lookup and interpolation, and the inversion time for a single measurement point can be reduced to less than 0.2ms.

[0154] By querying the pre-constructed manifold analytical equation, the single optimal severity variable value is directly mapped and restored to a specific five-dimensional value, thereby obtaining the current microscopic acoustic parameters.

[0155] As an optional implementation, if the system detects the optimal target residual value Loss obtained by the interpolation search... _ξ If the current road surface condition exceeds the preset anomaly tolerance threshold, it indicates that the current road surface condition deviates from the conventional congestion evolution trajectory calibrated in the laboratory. At this point, the system automatically discards the dimensionality reduction inversion results, marks the measurement section as a low-confidence anomaly area, and retains only the original reflection signal for subsequent offline manual verification by road administration personnel.

[0156] Example 5

[0157] Based on Embodiments 1 and 4 above, another method for implementing microscopic acoustic parameter inversion is provided. Specifically, this embodiment addresses the need to differentiate the physical states of road surfaces at different depths under complex working conditions, detailing the specific process of joint ultrasonic and ground-penetrating radar dual-layer inversion and differentiated decision-making for physical cleaning.

[0158] In one possible implementation, the non-destructive testing signal also includes ground-penetrating radar signals; inversion is performed using a pre-constructed equivalent fluid physics model to obtain the current microscopic acoustic parameters of the porous asphalt pavement under test, and the process also includes a branching step that triggers layered inversion, specifically including the following steps:

[0159] Step 501: Extract the dielectric characteristics of the ground-penetrating radar signal.

[0160] Ground-penetrating radar (GPR) signals typically contain a sequence of electromagnetic reflections from the interior and bottom of a porous road surface layer. The process of extracting dielectric characteristics includes removing DC offset from the sequence, performing bandpass filtering, and eliminating background direct wave components. The surface dielectric constant is calculated by comparing the local amplitude of the reflected wave from the porous layer surface with the amplitude of the reflection from the reference metal plate. Furthermore, the time difference between the surface reflection peak and the bottom reflection peak is extracted as the two-way travel time, and this time difference, combined with the surface dielectric constant, is used to estimate the physical thickness of the porous layer.

[0161] The calculation logic for estimating the thickness of the porous layer is as follows:

[0162] d=(c _em *Δ _t ) / (2*ε _surf 0.5 );

[0163] Where d is the estimated thickness of the porous layer, and c _em Let Δ be the speed constant of electromagnetic wave propagation in a vacuum environment. _t To extract the obtained round-trip travel, ε _surfThe calculated surface dielectric constant is represented by 0.5, which indicates the square root operation.

[0164] This thickness estimation process defines the spatial physical boundaries for subsequent layered inversion calculations. Furthermore, a short-time Fourier transform can be performed on the extracted waveform to extract the centroid frequency and spectral variance features from the frequency domain matrix, serving as supplementary dielectric attenuation parameters.

[0165] Step 502: Construct a joint inversion objective function consisting of multiple constraint components, wherein the surface micro parameters of the road surface are constrained by ultrasonic reflection signals, the average parameters of the entire road surface are constrained by dielectric characteristics, and parameter continuity constraints are applied between the surface and deep layers.

[0166] Considering that the penetration depth of high-frequency sound waves in porous asphalt is usually limited to a predetermined depth range on the surface, ultrasonic reflection signals mainly reflect the porosity of the surface layer. Ground penetrating radar signals, on the other hand, can penetrate and obtain the average electromagnetic reflection characteristics across the entire depth.

[0167] To establish the correlation between dielectric characteristics and internal microstructure, the system utilizes Bruggeman's effective dielectric theory to build a volume-proportional dielectric calculation model that includes a three-phase mixture of aggregate skeleton, air, and blockages. The difference in spatial sensitivity between the two sensing signals is used to construct the optimization objective for the inversion.

[0168] The calculation logic for the joint inversion objective function is as follows:

[0169] Loss _joint =w _1 *L _us +w _2 *L _GPR +w _3 *R _reg ;

[0170] Among them, Loss _joint To calculate the joint inversion objective function value, w _1 L represents the ultrasonic constraint weighting coefficient. _us Calculate the residual for the acoustic reflection spectrum corresponding to the surface micro-parameters, w _2 L represents the radar constraint weighting coefficient. _GPR To calculate the residual based on the dielectric characteristics corresponding to the full-layer average micro-parameters weighted by depth and thickness, w _3 R is the regularization weight coefficient. _reg This is a smoothing regularization constraint term used to measure the degree of dispersion between surface and deep microscopic parameters. The aforementioned weighting coefficients can be dynamically adjusted based on the signal-to-noise ratio (SNR) measurements of the signals received by each sensor at the current moment.

[0171] Step 503: Optimize the joint inversion objective function to obtain the current microscopic acoustic parameters with information on the depth and shallow stratification.

[0172] By executing a nonlinear optimization algorithm, a combination of variables is searched in the multidimensional parameter space to minimize the joint inversion objective function. By introducing electromagnetic property constraints including depth and regularization constraints representing physical continuity, the ambiguity blind zone of the single-layer inversion model on the vertical profile is eliminated. The final generated output values ​​of the current microacoustic parameters contain two independent data arrays, representing the porosity and flow resistance states in the first region from the road surface to the intermediate depth, and the second region from the intermediate depth to the base layer, respectively, providing a precise parameter matrix for determining whether pollution has penetrated deep into the base layer.

[0173] In some alternative implementations, an evidence-based multi-sensor fusion algorithm can be used to replace the optimization process of the joint inversion objective function. Specifically, this involves: pre-setting discrete state assumption intervals such as low congestion rate and high congestion rate; using the probability density distribution obtained from separate inversions, calculating the basic confidence value assigned to each assumption interval by the ultrasonic channel; simultaneously combining the dielectric constant extracted from ground-penetrating radar and its statistical distribution model, calculating the basic confidence value assigned to each assumption interval by the radar channel; and using an orthogonal synthesis operator to perform matrix multiplication and addition on the two sets of basic confidence values ​​to calculate the final comprehensive probability of the congestion state.

[0174] According to one aspect of this application, pavement maintenance decisions include differentiated cleaning instructions; the steps for generating differentiated cleaning instructions include:

[0175] Compare the surface blockage state with the deep blockage state in the current microscopic acoustic parameters that have information on the depth and shallow stratification.

[0176] After obtaining the layered microacoustic parameters, the volume proportion of blockage in the surface region and the volume proportion of blockage in the deep region are calculated through volume mapping relationships. The difference or quotient between these two proportions is calculated to assess the non-uniform distribution characteristics of solid deposits within the pores along the road surface depth.

[0177] Furthermore, if the severity of surface blockage is higher than that of deep blockage, a low-pressure flushing recommendation is generated in the differentiated cleaning instruction.

[0178] When the numerical comparison determines that the proportion of surface blockage volume is significantly greater than that of deep blockage volume, it is determined that the deposits are mainly deposited in the pore necks of the upper part of the asphalt mixture. Based on this, the system generates a low-pressure flushing suggestion that calls conventional water spraying equipment. This control logic can strip away the free deposits on the surface to restore the surface pore connectivity, while avoiding unnecessary high-pressure jet stripping of the asphalt binder wrapped in shallow aggregate.

[0179] If the severity of deep blockage is close to or greater than that of surface blockage, a high-pressure adsorption cleaning recommendation will be generated in the differentiated cleaning instruction.

[0180] When the numerical comparison determines that the conditions of the previous step are not met, it indicates that the fine particles have penetrated into the deep structure along the pores and undergone consolidation and hardening. The kinetic energy attenuation of the low-pressure jet makes it difficult for it to reach the lower deposition zone. Based on this, the system updates the maintenance strategy state machine and generates an operation instruction to call a high-pressure cleaning vehicle equipped with a combined jet cleaning and vacuum suction module. By directly mapping the deep and shallow stratification data to the predetermined engineering machinery call strategy, a complete data transfer closed loop from physical parameter analysis to terminal action implementation is achieved.

[0181] According to one aspect of this application, pavement maintenance decisions include differentiated cleaning instructions; the steps for generating differentiated cleaning instructions include:

[0182] Based on the current microacoustic parameters that have information on the depth and shallow stratification, the surface blockage state and the deep blockage state are respectively mapped.

[0183] Compare the surface blockage state with the deep blockage state;

[0184] If the severity of surface clogging is higher than that of deep clogging, a low-pressure flushing recommendation is generated in the differentiated cleaning instruction;

[0185] If the severity of deep blockage is no less than that of surface blockage, a high-pressure adsorption cleaning recommendation will be generated in the differentiated cleaning instruction.

[0186] Example 6

[0187] Based on the above embodiments, another implementation method is provided, which further details the specific process of calculating the acoustic safety margin and the measured closed-loop verification of the sound pressure level.

[0188] In one possible implementation, the acoustic safety margin of the porous asphalt pavement under test is calculated based on the current microscopic acoustic parameters and a pre-configured environmental acoustic model, such as... Figure 4 As shown, the specific steps include the following:

[0189] Step 601: Introduce pre-configured ground acoustic interaction coefficients to convert the current microacoustic parameters into equivalent frequency band noise reduction for each analysis frequency band.

[0190] Specifically, the sound absorption coefficient at the microscopic level is not directly equivalent to the noise reduction decibel value assessed in macroscopic engineering. The radiation of tire and road noise involves a complex interaction mechanism of ground reflection and horn effect. To achieve a cross-physical scale, this embodiment constructs an equivalent conversion model and introduces a ground acoustic interaction coefficient to characterize the proportion of sound energy that interacts acoustically with the road surface in different frequency bands.

[0191] The specific calculation logic for the equivalent frequency band noise reduction is as follows:

[0192] Δ _L_k =-10*log 10 ((1-γ _k *α _pred ) / (1-γ _k *α _ref ));

[0193] Where, Δ _L_k For the kth third octave band analysis frequency band, logk 10 γ represents the logarithmic operation to the base 10. _k Let α be the ground acoustic interaction coefficient corresponding to the k-th analysis frequency band. _pred α is the sound absorption coefficient of the porous asphalt pavement to be tested, predicted based on the current microscopic acoustic parameters. _ref The reference sound absorption coefficient for dense asphalt pavement is typically taken as 0.03 to 0.05.

[0194] The system provides two calculation pathways for determining the specific values ​​of the ground acoustic interaction coefficient.

[0195] The first method is a preferred path derived from near-field measurements, suitable for detection environments equipped with vehicle-mounted near-field microphones. In this case, γ _k The system performs reverse analysis by measuring the near-field sound level difference between unblocked porous pavement and dense pavement obtained during the initial construction phase. To avoid the computational singularity of the denominator approaching zero, the system automatically adjusts the γ value when the absorption coefficient of the predetermined frequency band is greater than 0.95. _k A correction strategy of linear extrapolation from adjacent frequency bands is adopted.

[0196] The second approach is an alternative calculation method, which is suitable for far-field sound receiving point deduction scenarios. It is determined by substituting into the spherical wave reflection interference approximation physical model and directly generating frequency-related coefficient vectors using the surface impedance of the porous road surface and the geometric relationship between the source and receiving ends.

[0197] Step 602: Substitute the equivalent frequency band noise reduction amount into the environmental acoustic model to perform spatial propagation simulation, and obtain the expected equivalent sound level of the pre-configured roadside sound receiving point.

[0198] In this embodiment, the environmental acoustic model includes a source strength superposition operator and a propagation attenuation operator. The system acquires the traffic flow, vehicle speed distribution, and vehicle type composition of the current road segment, and inputs the source strength superposition operator to calculate the total source strength of tire-road noise under the baseline state. The equivalent frequency band noise reduction of each frequency band output in step 601 is subtracted from the total source strength, and spatial propagation loss parameters such as geometric diffusion attenuation, atmospheric absorption attenuation, and sound barrier attenuation are further superimposed.

[0199] The above comprehensive calculation projects the acoustic attenuation characteristics originally represented by the road surface to a predetermined spatial coordinate position, and calculates and outputs the equivalent sound level at the expected receiving point. This equivalent sound level is expressed as the broadband total sound pressure level after A-weighting.

[0200] Step 603: Compare the pre-configured environmental functional zone reference limit with the expected sound level at the receiving point and calculate the difference, using the obtained difference as the acoustic safety margin.

[0201] The environmental functional zone baseline limits are obtained based on national or local environmental protection standards and are represented by the legally mandated daytime or nighttime noise limits for the area where the tested road section is located. The acoustic safety margin is obtained by calculating the difference between the baseline limit and the expected equivalent sound level at the receiving point. This margin directly characterizes the redundancy of the current road surface noise reduction capacity while meeting regulatory constraints, and forms the basis for subsequent decisions to initiate cleaning or maintenance.

[0202] The non-destructive testing signal also includes the measured near-field sound level acquired synchronously by the same vehicle, i.e., the measured near-field sound level acquired synchronously by the same vehicle equipped with the non-destructive testing signal acquisition equipment; the step of calculating the acoustic safety margin of the porous asphalt pavement under test also includes a closed-loop verification step for the prediction accuracy, such as... Figure 5 As shown, specifically:

[0203] 1) Calculate the expected near-field equivalent sound level in the tire-road interaction area based on the current microscopic acoustic parameters.

[0204] Furthermore, to verify the reliability of the cross-scale physical parameter conversion model, the system initiates a fault-tolerant verification mechanism. Unlike the roadside sound receiving point, the near-field interaction region is extremely close to the tire contact surface. The system calls the near-field radiation transfer operator to directly map the current microscopic acoustic parameters to the expected acoustic performance at the expected monitoring location next to the tire, generating the expected near-field equivalent sound level. This step provides an intermediate state verification benchmark for the theoretical prediction model.

[0205] 2) Compare the expected near-field equivalent sound level with the measured near-field sound level to obtain the verification deviation value.

[0206] Since the inspection vehicle simultaneously collects near-field microphone signals containing actual tire and road surface radiated noise, the system extracts the measured near-field sound level through signal filtering and A-weighted integration. The absolute value of the difference between the expected near-field equivalent sound level and the measured near-field sound level is calculated, and the output is the verification deviation value. This deviation value objectively quantifies the prediction error of the physical inversion model under the current operating conditions.

[0207] 3) When the verification deviation value does not exceed the preset tolerance range, the acoustic safety margin remains unchanged; when the verification deviation value exceeds the preset tolerance range, the preset safety margin is additionally deducted from the calculated acoustic safety margin to obtain the adjusted acoustic safety margin, or a recalibration trigger command for the equivalent fluid physics model is generated. The safety margin is the deviation compensation deduction or the model confidence deduction.

[0208] Specifically, the tolerance interval is pre-set based on the microphone hardware measurement uncertainty and the variance of environmental wind noise fluctuations. The system performs segmented judgment logic on the verification deviation value. In a simplified normalization example, for instance, the first threshold of the tolerance interval is set to 2, and the second threshold is set to 4. When the calculated verification deviation value δ≤2, the system determines that the theoretical model is highly consistent with the actual sound field and adopts the current acoustic safety margin. When 2<δ≤4, the system determines that there is a certain degree of unmodeled environmental interference, executes a downgraded conservative strategy, and deducts 1 from the original acoustic safety margin value as a safety margin. When δ>4, the system determines that the theoretical model is inaccurate in the current abnormal road segment, marks the current data segment as a high-risk state, and outputs a recalibration trigger command, prompting the system to check and update the long-wavelength approximation correction function in the equivalent fluid physics model during offline cycles. By introducing a closed-loop feedback mechanism based on the measured sound pressure level signal, the risk of overfitting and divergence of the prediction results generated by a single microscopic inversion model is avoided.

[0209] Example 7

[0210] The process of decomposing the characteristic directions of road surface deterioration mechanism and inferring and deciding on independent evolution trends is further explained in detail.

[0211] In one possible implementation, the characteristic direction decomposition of the road surface deterioration mechanism is performed based on the historical microacoustic parameter sequence and the current microacoustic parameters, specifically including the following steps:

[0212] Step 701: Based on the pre-constructed unblocked state parameters and extremely blocked state parameters, determine the blockage direction vector in the multidimensional physical space.

[0213] In a vector space composed of multidimensional physical parameters, different types of pavement distress can cause parameters to migrate along different directions. This embodiment defines a characteristic direction vector to separate the contribution of a single physical process from mixed parameter changes. The clogging direction vector characterizes the acoustic parameter changes caused by a pure pore-filling process. During system initialization, the unclogging state parameters of the same gradation material without any attachments and the extremely clogging state parameters at the maximum capacity limit are read from an offline database. The unclogging state parameters are subtracted from the extremely clogging state parameters, and the difference is normalized using the L2 norm of the difference to obtain the clogging direction vector per unit length. This calculation ensures that the vector retains only the directional attribute while stripping away the modulus attribute.

[0214] Step 702: Based on the record of parameter changes before and after cleaning in the historical microacoustic parameter sequence, determine the aging direction vector in the multidimensional physical space.

[0215] The aging direction vector characterizes the irreversible structural changes caused by the oxidative hardening of asphalt binder and the wear of the aggregate skeleton. Since cleaning operations can remove most of the attached material, two adjacent measurements before and after cleaning can be approximated as a state where the blockage has been cleared. The system extracts the parameter sequence measured immediately after cleaning operations for the same road segment from the historical microacoustic parameter sequence. The latest parameter of this post-cleaning time series is subtracted from the parameter after the first cleaning in the early stages of road construction, and normalized using its L2 norm to calculate the aging direction vector per unit length. When the sample size of historical cleaning data is insufficient, a pre-configured drift matrix of standard aging test parameters for this type of material can be used as the initial substitute input.

[0216] Furthermore, before performing subsequent decomposition, a conditional judgment mechanism is introduced to ensure the validity of the physical meaning of the projection. The cosine similarity between the blockage direction vector and the aging direction vector is calculated. If the absolute value of the cosine of the angle between the two is less than a preset orthogonality judgment threshold, the two mechanisms are deemed to be approximately independent in the parameter space, and decomposition is allowed; if the angle is less than the threshold, it indicates that the aging of the current road segment has led to abnormal pore shrinkage, causing the aging features and blockage features to overlap. In this case, the system will skip the vector decomposition step and revert to using a single evolutionary model to fit and predict the overall parameters.

[0217] Step 703: Calculate the micro-parameter increments of the current micro-acoustic parameters and the parameters of the previous period in the historical micro-acoustic parameter sequence.

[0218] In each detection cycle, the current microscopic acoustic parameters obtained from the inversion at the current time point are acquired, and the parameter records retained from the previous fixed detection cycle (such as the previous calendar month) are subtracted to obtain the microscopic parameter increment characterizing the extent of pavement performance degradation within the current cycle. To prevent external atypical factors from contaminating the data sequence, the system uses the Shapiro-Wilk normality test method to perform a residual normality test on the time series data before calculating the increment. If the skewness and kurtosis of the fitting residual caused by a certain data point exceed the threshold range, the system determines that it is an occasional outlier caused by construction dust or rainstorm erosion, performs a removal operation, and replaces it with a smoothed value generated by spline interpolation.

[0219] Step 704: Geometrically project the micro-parameter increments onto the blockage direction vector and the aging direction vector respectively to obtain the blockage evolution component and the aging evolution component that strip away the degradation mechanism.

[0220] By performing a dot product operation between the micro-parameter increment and the congestion direction vector, a scalar projection on the congestion dimension is obtained, which is the congestion evolution component. Similarly, by performing a dot product operation between the micro-parameter increment and the aging direction vector, a scalar projection on the aging dimension is obtained, which is the aging evolution component. Through this geometric projection process, the originally coupled composite degradation phenomenon is accurately decomposed into two independent feature sequences that evolve over time.

[0221] Furthermore, to obtain a complete time series for trend fitting, the system repeatedly performs the above incremental calculation and geometric projection operations on parameter pairs for every two adjacent detection cycles in the historical microacoustic parameter sequence. By arranging the clogging evolution components output in each cycle in chronological order, a cumulative time series of the clogging evolution components is constructed; similarly, by arranging the aging evolution components output in each cycle in chronological order, a cumulative time series of the aging evolution components is constructed.

[0222] For example, the following illustrates the operation process of geometric projection decomposition through a typical application scenario. Assume that the OGFC-13 section of a certain urban expressway has been in operation for two years. During the inspection in the 24th month, the increment of the current microscopic acoustic parameters obtained by the system inversion relative to the microscopic parameters of the previous month (the 23rd month) is a five-dimensional vector Δ. _p The system had previously determined the blockage direction vector e based on laboratory data. _clog (The direction dominated by decreased porosity, increased flow resistance, and shortened viscous characteristic length), and the aging direction vector e _age (The direction dominated by a slight increase in tortuosity and a shortening of thermal characteristic length). Δ _p respectively with e _clog and e _agePerforming a dot product operation, the congestion evolution component for this month is positive, indicating the presence of new congestion, while the aging evolution component is a smaller positive value, indicating slow aging. If the road section had just undergone cleaning operations last month, the congestion evolution component should be close to zero, while the aging evolution component remains positive, verifying the effective separation of the two mechanisms.

[0223] According to one aspect of this application, multi-dimensional independent evolution trends are derived, and pavement maintenance decisions are obtained based on these independent evolution trends and acoustic safety margins, specifically including:

[0224] A growth model with a saturation upper limit is used to fit the trend of the time series of the clogging evolution component, and a monotonic evolution model is used to fit the trend of the time series of the aging evolution component. The two models are then combined to obtain independent evolution trends.

[0225] Specifically, another approach is to perform time series aggregation on the blockage evolution component and aging evolution component obtained by performing geometric projection on multiple detection cycles to obtain time series of the blockage evolution component and time series of the aging evolution component, respectively; use a growth model with a saturation upper limit to perform trend fitting on the time series of the blockage evolution component, and use a monotonic evolution model to perform trend fitting on the time series of the aging evolution component, and merge them to obtain independent evolution trends.

[0226] The pore blockage process of porous asphalt is physically constrained by the available pore volume, making it difficult to grow indefinitely. Therefore, the system abandons the traditional unbounded logarithmic model. In this embodiment, a Logistic growth model with a saturation upper limit is used to fit the blockage evolution component.

[0227] The Logistic growth model is as follows:

[0228] Cr(N _cum )=Cr _max / (1+K*exp(-κ*N _cum ));

[0229] Among them, Cr(N) _cum ) represents the blockage rate corresponding to the current cumulative number of shafts, Cr _max The upper limit of saturation is pre-calibrated based on the physical porosity characteristics, K is the integration constant determined by the initial state, exp represents the exponential function with the natural constant as the base, κ is the evolution parameter characterizing the clogging rate, and N... _cum This refers to the total number of traffic equivalent axles that have passed through this section of road.

[0230] By using the least squares method to solve for K and κ based on historical time series data, the congestion state of any future node can be extrapolated. For the aging evolution component, since material aging is a long-term continuous process, the system uses a monotonic evolution model to fit its trend.

[0231] In one alternative implementation, a linear model is used for fitting when the time series of the aging evolution component exhibits an approximately constant growth trend; a power function model is used for fitting when the time series shows a decelerating convergence trend. The model selection can be determined by comparing the sum of squared residuals after fitting the two models. The two prediction results are combined to output the independent evolution trend.

[0232] Furthermore, based on the fitted blockage evolution component and aging evolution component, the expected acoustic recovery performance after cleaning and maintenance is deduced using an equivalent fluid physics model.

[0233] Before generating pavement maintenance decisions, the system simulates the maintenance effect using a virtual intervention algorithm. At the predicted cleaning time point, the congestion evolution component is forcibly reset to a near-zero residual lower limit value, while the aging evolution component maintains its predicted node value. Specifically, the system uses the initial state parameters of the road segment at the beginning of its construction as a base point, superimposes the product of the reset congestion evolution component and the congestion direction vector, and the product of the predicted aging evolution component and the aging direction vector, to reconstruct the expected five-dimensional microacoustic parameter vector after cleaning.

[0234] The two reset components are recombined and mapped back to the sound absorption coefficient prediction module to calculate the optimal sound absorption coefficient achievable after simulated cleaning, which is then converted into the expected acoustic recovery performance. This performance characterizes the maximum acoustic benefit that physical cleaning can recover.

[0235] Based on this, if the expected acoustic recovery performance, after adding the acoustic safety margin, still fails to meet the preset acoustic compliance conditions, it is determined that the road acoustic attenuation caused by aging is irreversible, and a surface renovation warning is generated in the road maintenance decision; if the expected acoustic recovery performance, after adding the acoustic safety margin, meets the preset acoustic compliance conditions, a cleaning and maintenance schedule recommendation is generated in the road maintenance decision.

[0236] The system extracts preset standard values ​​for acoustic functional zones. It then adds the expected acoustic recovery performance to the current acoustic safety margin. If this sum is less than a preset safety buffer threshold (e.g., 3dB), it indicates that even after completely removing all obstructions, the irreversible aging of the road surface has caused it to lose its legally mandated noise reduction capability. Based on this, the system determines that conventional cleaning has failed and directly outputs a warning instruction for surface renovation involving the resurfacing of the underlying material. Conversely, if the sum is greater than the buffer threshold, it generates a suggested schedule for conventional high and low pressure cleaning based on the remaining time to reach the warning line and the ranking of other road sections.

[0237] The above embodiments demonstrate that the method of the present invention, by constructing a hierarchical screening and dimensionality reduction inversion process, can achieve real-time acquisition of microscopic acoustic parameters of porous asphalt pavement in a vehicle-mounted dynamic detection environment; by introducing a characteristic direction decomposition mechanism, it can effectively distinguish the independent contributions of two deterioration mechanisms: reversible blockage and irreversible aging. It should be understood that specific detection accuracy, inversion speed, and prediction accuracy may vary depending on pavement type, sensor configuration, and environmental conditions.

[0238] This application effectively eliminates the physical distortion of the ultrasonic reflection spectrum caused by relative motion and environmental fluctuations by introducing Doppler frequency shift compensation, dynamic correction mechanism for road surface infrared temperature, and supplementing it with system error correction based on the upper limit of skeleton dispersion frequency. Addressing the computational bottleneck of real-time solution for complex acoustic models, an adaptive frequency band energy ratio is used for pre-screening. Based on the pre-calculated univariate congestion trajectory manifold, the high-dimensional parameter inversion is reduced to table lookup interpolation optimization, compressing the single-point inversion time to the millisecond level, which can meet the data throughput requirements of high-speed continuous vehicle inspection.

[0239] In terms of breaking through the physical resolution limit of a single sensor, the scheme constructs a joint objective function that integrates acoustic wave and electromagnetic dielectric characteristics, and separates the surface and deep blockage states of porous asphalt pavement, providing an objective basis for the subsequent generation of differentiated high and low pressure cleaning commands.

[0240] In order to break down the barriers between micro and macro physical quantities, the sound absorption rate of micropores is equivalently converted into the noise reduction of macro frequency bands by using the ground acoustic interaction coefficient, and the measured sound pressure level is superimposed to form a closed-loop verification, thus constructing a high-confidence acoustic safety margin judgment system.

[0241] By introducing a geometric feature orientation decomposition algorithm within a multidimensional parameter space, reversible pore dust filling events are separated from irreversible skeleton wear and aging events. This makes the prediction of cleaning effects and early warning of surface overhaul more scientific, helping to achieve a leap from passive periodic maintenance to accurate and forward-looking decision-making guided by acoustic environment compliance. It solves the pain point of predictive failure caused by confusion in disease mechanisms.

[0242] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the protection scope of the present invention.

Claims

1. A noise-reducing asphalt pavement maintenance decision-making method based on acoustic characteristic attenuation prediction, characterized in that, include: Acquire non-destructive testing signals of the porous asphalt pavement to be tested, including ultrasonic reflection signals; Based on the ultrasonic reflection signal, acoustic feature indexes are extracted for hierarchical screening, and combined with a pre-constructed equivalent fluid physics model for inversion, the current microscopic acoustic parameters of the porous asphalt pavement to be tested are obtained. The acoustic safety margin of the porous asphalt pavement under test is calculated based on the current microscopic acoustic parameters and the pre-configured environmental acoustic model. Obtain a pre-stored sequence of historical microacoustic parameters; Based on the historical microacoustic parameter sequence and the current microacoustic parameters, the characteristic direction decomposition of the road surface deterioration mechanism is performed, and the multi-dimensional independent evolution trend is deduced. Road maintenance decisions are derived based on independent evolution trends and acoustic safety margins.

2. The method according to claim 1, characterized in that, In the process of graded screening based on acoustic feature indices extracted from ultrasonic reflection signals, and inversion using a pre-constructed equivalent fluid physics model to obtain the current microscopic acoustic parameters of the porous asphalt pavement under test, the graded screening includes: The acoustic characteristic index is obtained by calculating the ratio of the reflected energy of the ultrasonic reflected signal in the high-frequency band to the reflected energy in the low-frequency band. The acoustic feature index is compared with the pre-configured screening threshold; If the acoustic characteristic index is higher than the pre-configured screening threshold, the step of inversion combined with the pre-built equivalent fluid physics model is triggered to obtain the current microscopic acoustic parameters. If the acoustic feature index is not higher than the pre-configured screening threshold, the acoustic feature index is converted and estimated using the pre-stored feature mapping relationship to obtain the current microscopic acoustic parameters.

3. The method according to claim 2, characterized in that, The boundary frequency between the high-frequency and low-frequency bands is pre-optimized and configured; the boundary frequency is pre-optimized and configured in the following way: Multidimensional microscopic parameters are extracted from the equivalent fluid physics model and divided into clogging-related parameters and non-clogging-related parameters. Calculate the analytical sensitivity of acoustic characteristic indices to each parameter in a multidimensional micro-parameter; With the goal of maximizing the analytical sensitivity for clogging-related parameters and minimizing the analytical sensitivity for non-clogging-related parameters, a search is conducted within a preset available ultrasonic frequency band to determine the boundary frequency.

4. The method according to claim 1, characterized in that, The acoustic safety margin of the porous asphalt pavement under test is calculated based on the current microscopic acoustic parameters and the pre-configured environmental acoustic model, including: By introducing pre-configured ground acoustic interaction coefficients, the current microacoustic parameters are converted into equivalent frequency band noise reduction for each analysis frequency band. Substitute the equivalent frequency band noise reduction into the environmental acoustic model to perform spatial propagation simulation and obtain the expected equivalent sound level of the target sound receiving point for the predetermined roadside sound receiving point. The pre-configured environmental functional zone benchmark limit is compared with the expected sound level at the receiving point, and the difference is used as the acoustic safety margin.

5. The method according to claim 4, characterized in that, The non-destructive testing signal also includes the measured near-field sound level acquired simultaneously in the same vehicle; the calculation of the acoustic safety margin of the porous asphalt pavement under test also includes closed-loop verification for prediction accuracy: Based on the current microacoustic parameters, calculate the expected near-field equivalent sound level in the tire-road interaction area; The expected near-field equivalent sound level is compared with the measured near-field sound level to obtain the verification deviation value; When the calibration deviation exceeds the preset tolerance range, the preset safety margin is deducted from the calculated acoustic safety margin, or a recalibration trigger command for the equivalent fluid physics model is generated.

6. The method according to claim 1, characterized in that, The pre-built equivalent fluid physics model is calibrated and corrected through the following offline construction steps: Based on the physical dimensional properties of the porous asphalt skeleton, the upper limit of the dispersion frequency for the equivalent fluid physical model is determined. The highest frequency of the ultrasonic reflection signal used for offline calibration and online detection is limited to not exceeding the upper limit of the diffusion frequency; By comparing the measured absorption coefficient of the physical specimen in the audible frequency band with the theoretical absorption coefficient extrapolated across the frequency band by the equivalent fluid physical model, a frequency correlation correction function is generated to compensate for the long-wavelength approximation failure, and this function is built into the pre-constructed equivalent fluid physical model.

7. The method according to claim 1, characterized in that, The non-destructive testing signal also includes an infrared temperature signal; in the step of inversion combined with a pre-built equivalent fluid physics model, the infrared temperature signal is used to obtain the road surface temperature field, and the temperature-sensitive physical parameters in the equivalent fluid physics model are compensated in real time based on the road surface temperature field.

8. The method according to claim 1, characterized in that, Acquiring the non-destructive testing signal of the porous asphalt pavement to be tested also includes: Simultaneously acquire vehicle driving status signals and spatial positioning signals; Doppler frequency shift compensation preprocessing is performed on ultrasonic reflection signals based on vehicle driving status signals; Spatial-temporal registration of nondestructive testing signals based on spatial positioning signals.

9. The method according to claim 5, characterized in that, Road maintenance decisions include differentiated cleaning instructions; generating differentiated cleaning instructions includes: Compare the surface blockage state with the deep blockage state in the current microacoustic parameters that have information on the depth and shallow stratification. If the severity of surface clogging is higher than that of deep clogging, a low-pressure flushing recommendation is generated in the differentiated cleaning instruction; If the severity of deep blockage is higher than that of surface blockage, a high-pressure adsorption cleaning recommendation will be generated in the differentiated cleaning instruction.

10. A noise-reducing asphalt pavement maintenance decision system based on acoustic characteristic attenuation prediction, characterized in that, include: At least one processor; as well as, A memory communicatively connected to at least one of the processors; wherein, The memory stores instructions that can be executed by the processor to implement the method according to any one of claims 1 to 9.

Images