Method for rapid non-destructive detection of subgrade compactness based on polyphase wave dispersion characteristics

Abstract

The application provides a kind of based on the method for quickly and non-destructively detecting subgrade compactness of polyphase wave dispersion characteristics, comprising: the target subgrade surface is laid out with intelligent sensing system, control multi-point excitation array according to multiple detection frequencies and array element phase combination input polyphase mechanical wave, collect multi-channel response and construct dispersion response matrix, to identify surface layer dominant propagation mode;According to the channel amplitude distribution and channel phase delay of the surface layer dominant propagation mode, adjust the output amplitude, initial excitation phase and emission time of each excitation unit, form a coherent quiet zone in the predetermined receiving area, and obtain a first residual wave response data set;Extract abnormal early arrival wave front from it, identify rigid propagation shortcut and generate compensation parameters;The quiet control parameters and rigid propagation shortcut compensation parameters are complexly synthesized, double blanking is implemented, and a second residual wave response data set is obtained.

Description

Technical Field

[0001] This invention relates to the field of roadbed engineering testing technology, and in particular to a rapid non-destructive testing method for roadbed compaction based on multiphase wave dispersion characteristics. Background Technology

[0002] Subgrade compaction is an important indicator for evaluating filling quality, bearing capacity, and settlement stability. Existing testing methods are mainly based on non-destructive testing methods such as vibration, acoustic waves, or surface waves. Traditional sampling and testing are inefficient and cannot meet the needs of large-scale rapid testing. Although existing non-destructive testing methods improve testing efficiency, the test results are easily affected by the subgrade layering structure, particle size distribution, water content, and local rigid inclusions.

[0003] In actual roadbeds, the surface soil, after compaction, typically forms a relatively dense area. Mechanical waves tend to propagate preferentially along the surface, forming a surface-dominant propagation mode with high energy. This mode masks the slow-propagating waves, scattered wake waves, and mode-transformation waves generated in the deeper loose layers, localized voids, and areas with insufficient particle contact. Consequently, the test results tend to overemphasize the surface condition and fail to accurately reflect the true compaction status within the roadbed.

[0004] Meanwhile, isolated boulders, large particle agglomerates, localized high-rigidity regions, or interconnected particle force chains in coarse-grained fillers may form rigid propagation shortcuts and generate abnormally early arrival wavefronts. Existing methods typically determine compaction based on propagation velocity, attenuation characteristics, or response energy, making it difficult to distinguish between the effective propagation response formed by the overall particle skeleton and the abnormal response formed by localized high-speed paths. This may lead to the misjudgment of subgrades with internal looseness or porosity as properly compacted.

[0005] Therefore, this invention proposes a rapid non-destructive testing method for roadbed compaction based on multiphase wave dispersion characteristics. The information disclosed in the background section is only for enhancing understanding of the background of this disclosure and may therefore contain prior art information that is not common knowledge to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of existing technologies by providing a rapid and non-destructive testing method for roadbed compaction based on multiphase wave dispersion characteristics, thereby solving the technical problems mentioned in the background section.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A rapid non-destructive testing method for subgrade compaction based on multiphase wave dispersion characteristics includes the following steps: S1. Deploy an intelligent sensing system on the roadbed surface, control a multi-point excitation array to input multi-phase mechanical waves according to the combination of detection frequency and array element phase, collect multi-channel responses, construct a dispersion response matrix and identify the dominant propagation mode of the surface layer. S2. Based on the channel amplitude distribution and channel phase delay of the surface dominant propagation mode, adjust the output amplitude, initial phase and transmission time of each excitation unit so that they cancel each other out in the predetermined receiving area, forming a coherent silent zone and obtaining the first residual wave response dataset. S3. Extract candidate early arrival wavefronts from the first residual wave response dataset, identify abnormal early arrival wavefronts based on propagation speed, propagation direction, channel coherence and spatial continuity, determine rigid propagation shortcuts and generate rigid propagation shortcut compensation parameter sets. S4. The final silent control parameter set and the rigid propagation shortcut compensation parameter set are combined in complex form to output the second set of coherent mechanical waves. The composite blanking control parameters are corrected according to the residual energy ratio of the surface mode and the residual energy ratio of the abnormal early arrival wavefront to obtain the second residual wave response dataset. S5. Extract the skeleton response feature set from the second residual fluctuation response dataset, determine the effective compaction degree, particle contact network integrity and particle contact network stability according to the calibration relationship, and obtain the hidden undercompaction risk.

[0008] S1 specifically includes: deploying an intelligent sensing system on the target roadbed surface to collect and calibrate the mechanical wave response, determine the channel time deviation, amplitude correction coefficient, and phase deviation, and form an array calibration parameter set; determining the detection frequency and array element phase combination based on the fill layer thickness, the initial value of the equivalent propagation velocity, and the equipment operating frequency band, outputting multiphase mechanical waves and correcting the original wave response to form a calibration wave response dataset; performing time-frequency decomposition and wavenumber domain superposition on the calibration wave response dataset to construct an initial dispersion response matrix, connecting the local maximum points of dispersion energy to form dispersion branches, and identifying the dominant propagation modes of the surface layer based on the surface sensitive frequency band, modal energy ratio, and phase continuity to form a surface dominant mode parameter set.

[0009] S2 specifically includes: determining a predetermined receiving area based on the surface dominant mode parameter set and the array calibration parameter set; obtaining complex transfer coefficients; and determining the output amplitude, initial excitation phase, and transmission time with the goal of minimizing the synthetic residual energy, thus forming an initial quiescent control parameter set; outputting a first set of coherent mechanical waves according to the initial quiescent control parameter set; iteratively correcting the output amplitude and initial excitation phase based on the surface mode residual energy ratio, thus forming a coherent quiescent region and a final quiescent control parameter set; acquiring the response while keeping the final quiescent control parameters unchanged; removing residual components consistent with the surface dominant propagation mode; and labeling the early arrival response, main scattering response, and late tail wave response to form a first residual wave response dataset.

[0010] S3 specifically includes: extracting candidate early arrival components that meet the requirements of arrival time, energy ratio, cross-channel correlation, and phase continuity from the first residual wave response dataset, removing electrical crosstalk, and connecting them to form a candidate early arrival wavefront set; performing multi-channel propagation fitting on the candidate early arrival wavefronts, identifying abnormal early arrival wavefronts based on equivalent propagation speed, propagation direction, channel coherence, and cross-channel duration, determining rigid propagation shortcuts, and forming a rigid propagation shortcut parameter set; establishing a shortcut compensation transfer matrix when the final silence control parameters are continuously effective, and obtaining the compensation output amplitude, compensation excitation phase, and compensation emission time based on the abnormal response vector to form a rigid propagation shortcut compensation parameter set.

[0011] S4 specifically includes: aligning, synthesizing, and verifying the energy of the two types of complex control quantities by matching the final silent control parameter set with the rigid propagation shortcut compensation parameter set according to the detection frequency, forming a composite blanking control parameter set and outputting a second set of coherent mechanical waves; acquiring the dual blanking residual response, calculating the residual energy ratio of the surface mode and the residual energy ratio of the anomalous early arrival wavefront, and alternately correcting the two types of control components according to the corresponding thresholds to form a dual blanking control parameter set; repeatedly detecting by calling the dual blanking control parameter set, removing projections consistent with the surface dominant propagation mode and the anomalous early arrival wavefront, and marking slow propagation candidate responses, scattering wake wave candidate responses, and mode transition candidate responses to form a second residual wave response dataset.

[0012] S5 specifically includes: extracting slow-propagating waves, scattered wake waves, and mode-converted waves from the second residual wave response dataset; calculating the number of residual propagation paths, coherent wake wave duration, dispersion branch continuity, and multi-channel spatial response uniformity to form a skeleton response feature set; establishing a calibration relationship based on the measured compaction degree of the calibrated subgrade and skeleton response characteristics to obtain an initial value of effective compaction degree; and determining the particle contact network integrity, particle contact network stability, and data confidence level to form an effective compaction state parameter set; correcting the initial value of effective compaction degree based on the particle contact network integrity and particle contact network stability; and determining the effective compaction degree, under-compaction risk type, and risk level in conjunction with the compaction degree qualification threshold.

[0013] The beneficial effects of this invention are as follows: This invention utilizes a multi-point excitation array to output multiphase mechanical waves with controllable frequency, amplitude, and phase. Combined with a multi-channel response to construct a dispersion response matrix, it can accurately identify the dominant propagation mode of the surface layer in the roadbed, reducing the insufficient detection information caused by a single excitation method. By adjusting the output amplitude, initial phase, and transmission time of each excitation unit, a coherent silent zone is formed in a predetermined receiving area, effectively suppressing the strong response generated by the dense surface layer and reducing the shading effect of the surface compaction state on the detection results of the internal loose area.

[0014] This invention, by extracting anomalous early-arriving wavefronts and analyzing their propagation speed, direction, channel coherence, and spatial continuity, can identify rigid propagation shortcuts formed by isolated boulders, large particle agglomerates, or penetrating particle force chains, thus avoiding misjudging local high-speed responses as good overall compaction. Through complex synthesis and closed-loop correction of silent control parameters and rigid propagation shortcut compensation parameters, it can simultaneously weaken the dominant surface propagation mode and anomalous early-arriving wavefronts while retaining slow-propagating waves, scattering tails, and mode-transition waves, thereby improving the signal-to-noise ratio and reliability of the effective skeleton response.

[0015] This invention comprehensively calculates the number of remaining propagation paths, coherent wake duration, dispersion branch continuity, and multi-channel spatial response uniformity to evaluate the internal state of the roadbed from both the integrity and stability of the particle contact network, overcoming the limitations of single propagation velocity or response energy indicators. By determining the effective compaction degree through calibration relationships and correcting it in conjunction with the integrity and stability of the particle contact network, it can distinguish risks such as overall undercompaction, localized overburden, single bearing path, and contact network instability, achieving rapid non-destructive testing of roadbed compaction degree and identification of hidden undercompaction. Attached Figure Description

[0016] Figure 1 This is a flowchart of the rapid non-destructive testing method for roadbed compaction based on the multiphase wave dispersion characteristics of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example: Figure 1 As shown, this embodiment provides a rapid non-destructive testing method for roadbed compaction based on multiphase wave dispersion characteristics, including the following steps: S1. Deploy an intelligent sensing system on the roadbed surface, control a multi-point excitation array to input multi-phase mechanical waves according to the combination of detection frequency and array element phase, collect multi-channel responses, construct a dispersion response matrix and identify the dominant propagation mode of the surface layer. S2. Based on the channel amplitude distribution and channel phase delay of the surface dominant propagation mode, adjust the output amplitude, initial phase and transmission time of each excitation unit so that they cancel each other out in the predetermined receiving area, forming a coherent silent zone and obtaining the first residual wave response dataset. S3. Extract candidate early arrival wavefronts from the first residual wave response dataset, identify abnormal early arrival wavefronts based on propagation speed, propagation direction, channel coherence and spatial continuity, determine rigid propagation shortcuts and generate rigid propagation shortcut compensation parameter sets. S4. The final silent control parameter set and the rigid propagation shortcut compensation parameter set are combined in complex form to output the second set of coherent mechanical waves. The composite blanking control parameters are corrected according to the residual energy ratio of the surface mode and the residual energy ratio of the abnormal early arrival wavefront to obtain the second residual wave response dataset. S5. Extract the skeleton response feature set from the second residual fluctuation response dataset, determine the effective compaction degree, particle contact network integrity and particle contact network stability according to the calibration relationship, and obtain the hidden undercompaction risk.

[0019] S1 specifically includes the following sub-steps: S110. Before performing target subgrade testing, a calibration parameter library shall be established in advance. The calibration parameter library consists of calibration subgrade data using the same type of filler as the target subgrade, and includes at least channel repetition error, regional energy threshold, modal energy ratio threshold, velocity change threshold, correlation threshold, silent threshold, upper limit of normal propagation velocity, wake coherence threshold, path merging threshold, particle contact network integrity qualification threshold, particle contact network stability qualification threshold, and calibration model parameters.

[0020] The term "same type of filler" as used in this application refers to roadbed fillers that have the same source, the same particle size distribution range, the maximum particle size difference within the allowable deviation of the project, the same design filling thickness, and whose moisture content is within the allowable range for construction during testing.

[0021] Subsequently, an intelligent sensing system is deployed on the target roadbed surface. The intelligent sensing system includes a multi-point excitation array, a multi-channel receiving array, an array controller, and a signal processor. The multi-point excitation array includes at least four excitation units that can independently adjust the output frequency, output amplitude, initial phase, and transmission time. The multi-channel receiving array includes at least six triaxial vibration sensors. Each triaxial vibration sensor constitutes a receiving unit and corresponds to a receiving channel. The receiving unit represents a physical sensor, and the receiving channel represents the signal acquisition path corresponding to that receiving unit.

[0022] The excitation unit uses an electromagnetic exciter or a piezoelectric exciter, whose effective output frequency band covers at least 20Hz to 2000Hz; the triaxial vibration sensor uses a detector or a MEMS accelerometer, whose frequency response range matches the effective output frequency band of the excitation unit.

[0023] The array controller provides a unified clock to each excitation unit and each receiving channel. The sampling frequency of the signal processor is no less than five times the highest candidate detection frequency, and the quantization accuracy is no less than 16 bits. The current fill layer thickness is obtained from the roadbed construction design documents, the detection depth is determined according to the detection task, and the area covering the current fill layer is defined as the preset detection range.

[0024] Controlling an excitation unit to output a linear sweep calibration mechanical wave covering the effective operating frequency band of the equipment, obtaining the initial value of the equivalent propagation velocity of the target roadbed, and dividing the initial value of the equivalent propagation velocity by the corresponding candidate detection frequency to obtain the target wavelength; the distance between adjacent receiving units is not greater than 1 / 2 of the shortest wavelength of the target, and the total length of the receiving array is not less than the longest wavelength of the target.

[0025] Each excitation unit is sequentially controlled to transmit calibration mechanical waves individually with the same preset calibration amplitude, and each excitation unit transmits at least 3 times. The preset calibration amplitude is 10% of the rated output amplitude of the excitation unit, and the total energy of the calibration mechanical waves is kept below the upper limit of safe excitation energy. The receiving channel with the highest signal-to-noise ratio is used as the reference channel, and data with a repetition response correlation coefficient lower than 0.90 are discarded.

[0026] The channel time deviation of the q-th receiving channel relative to the reference channel is determined by the following formula: In the formula, The channel time offset of the q-th receiving channel is expressed in seconds. is the time shift to be searched, in seconds; T is the effective response time of the calibration mechanical wave, in seconds; The calibration response of the reference channel at time variable t; This is the calibration response of the q-th receiving channel at time variable t.

[0027] Based on the amplitude ratio and phase difference between the q-th receiving channel and the reference channel at each candidate detection frequency, the channel amplitude correction coefficient and channel phase deviation are determined respectively. The spatial coordinates of each excitation unit and each receiving unit, the channel time deviation, the channel amplitude correction coefficient, and the channel phase deviation are combined to form an array calibration parameter set.

[0028] For example, when the initial equivalent propagation speed is 240m / s, the highest candidate detection frequency is 1200Hz, the shortest wavelength of the target is 0.20m, and the spacing between adjacent receiving units can be set to 0.10m; when the fifth receiving channel lags behind the reference channel by 0.13ms, the response amplitude at 400Hz is 0.92 times the response amplitude of the reference channel, and the phase lag is 8°, then subsequent receiving channels will be processed with 0.13ms time advance processing, amplitude correction, and 8° phase compensation.

[0029] S120. Based on the current filling layer thickness, the initial value of the equivalent propagation velocity obtained in S110, and the effective operating frequency band of the intelligent sensing system, determine the candidate detection frequencies. The candidate detection frequencies are arranged in order from low to high, and the ratio of adjacent candidate detection frequencies is set to 1.25 to 1.60, with a number of no less than 6. After eliminating candidate detection frequencies that do not meet the requirements for signal-to-noise ratio or repeatability, the remaining frequencies are determined as the detection frequencies.

[0030] The multi-point excitation array is controlled to output multiphase mechanical waves according to the array calibration parameter set. The multiphase mechanical wave referred to in this application refers to the mechanical wave field formed by the superposition of multiple excitation units in the target roadbed after synchronous excitation at the same detection frequency with different initial phases and controlled output amplitudes. The term "multiphase" indicates that the initial phases of the multiple excitation units are different, but does not indicate that the target roadbed contains multiple material phases.

[0031] At least one in-phase combination, forward progressive combination, reverse progressive combination, and alternating anti-phase combination are set at each detection frequency; when the multi-point excitation array includes 4 excitation units, the above array element phase combinations can be set to 0°, 0°, 0°, 0°, 0°, 0°, 60°, 120°, 180°, 180°, 120°, 60°, 0°, and 0°, 180°, 0°, 180°, 0°, 180°, respectively.

[0032] The original fluctuation response corresponding to each detection frequency and each array element phase combination is synchronously acquired by a multi-channel receiving array, and time, amplitude, and phase corrections are performed according to the following formula: In the formula, This is the corrected frequency domain response of the q-th receiving channel at the detection frequency f; This corresponds to the original frequency domain response; This is the channel amplitude correction factor; This refers to channel time deviation; This refers to channel phase deviation; It is the imaginary unit.

[0033] The corrected frequency domain response is converted back to the time domain and arranged according to the detection frequency, array element phase combination, and receiving channel number to form a calibration fluctuation response dataset. The signal-to-noise ratio (SNR) is calculated using the logarithm of 20 times the ratio of the effective response time window root mean square amplitude to the background noise time window root mean square amplitude. When the SNR is below 10 dB, or the correlation coefficient between three repeated acquisitions is below 0.85, the contact status of the corresponding excitation unit and receiving unit is checked, and the data is reacquired. Data that does not meet the requirements is not written into the calibration fluctuation response dataset.

[0034] S130. Perform time-frequency decomposition on the calibration fluctuation response dataset. Discrete frequency excitation uses a stable response time window covering at least 5 complete vibration cycles for discrete Fourier transform; linear sweep frequency excitation uses short-time Fourier transform, with the overlap ratio of adjacent analysis windows set to 50% to 75%. Determine the arrival time difference based on the cross-correlation peak positions of the responses of adjacent receiving channels, obtain the propagation phase by performing phase unfolding along the receiving array arrangement direction based on the phase angle of the corrected frequency domain response, and obtain the equivalent propagation speed based on the ratio of the actual distance between receiving units to the arrival time difference.

[0035] The corrected frequency domain responses of each receiving channel are superimposed in an array according to the propagation wavenumber to obtain the dispersion energy corresponding to the detection frequency f and the propagation wavenumber k: In the formula, For dispersive energy; N is the number of receiving channels; This is the corrected frequency domain response of the q-th receiving channel; is the distance of the q-th receiving unit relative to the origin of the receiving array coordinates; k is the propagation wavenumber. The initial dispersion response matrix is ​​composed of the dispersion energy corresponding to each detection frequency and propagation wavenumber.

[0036] The dispersion branch referred to in this application is a curve formed by connecting multiple local maximum points of dispersion energy at multiple adjacent detection frequencies according to the continuous change relationship of propagation wavenumber; the corresponding local maximum points are connected only when the rate of change of propagation velocity at adjacent frequency points does not exceed the velocity change threshold in the calibration parameter library, the correlation coefficient of channel phase distribution is not lower than the correlation threshold, and the vibration direction is consistent; the velocity change threshold and the correlation threshold are respectively taken as the 95th percentile of the rate of change of propagation velocity at adjacent frequency points of the calibrated roadbed and the 5th percentile of the correlation coefficient of channel phase distribution.

[0037] The sensitive frequency bands for the roadbed surface are determined based on the current fill layer thickness and the corresponding wavelengths of each detection frequency. Detection frequencies whose main impact depth does not exceed the current fill layer thickness are included in this frequency band. The ratio of the dispersion energy within the coverage area of ​​each dispersion branch to the total effective dispersion energy is determined as the modal energy percentage.

[0038] The term "surface dominant propagation mode" as used in this application refers to a propagation mode located in the sensitive frequency band of the roadbed surface, forming continuous dispersion branches at at least three adjacent detection frequencies, having a mode energy ratio not lower than the mode energy ratio threshold in the calibration parameter library, and maintaining phase continuity between adjacent receiving channels; the mode energy ratio threshold is taken as the 5th percentile value of the corresponding mode energy ratio of the calibrated roadbed.

[0039] The dominant propagation modes on the surface are numbered according to their modal energy proportion from high to low. The dominant frequency range, channel amplitude distribution, channel phase delay and propagation direction of each dominant propagation mode are extracted to form a set of dominant mode parameters on the surface, which are then output to S210.

[0040] S2 specifically includes the following sub-steps: S210: Based on the surface dominant mode parameter set formed in S130 and the array calibration parameter set formed in S110, determine the predetermined receiving area and the initial silence control parameter set corresponding to each detection frequency. The predetermined receiving area, as referred to in this application, means the receiving space jointly covered by at least three spatially adjacent receiving channels in a multi-channel receiving array, whose surface dominant propagation mode response energy is not lower than a preset area energy threshold.

[0041] The preset regional energy threshold is read from the calibration parameter library. When establishing the calibration parameter library, no less than 20 sets of data are collected from the calibration subgrade with the same type of fill and qualified compaction quality. The response energy of each receiving channel in the dominant frequency range is calculated, and the larger value between 60% of the response energy of the maximum channel and 50% of the response energy of all effective channels is taken as the preset regional energy threshold.

[0042] For each detection frequency, the nth excitation unit is individually controlled to output a mechanical wave with a preset calibration amplitude. Time, amplitude, and phase corrections are performed using the array calibration parameter set to obtain the complex transfer coefficient from the nth excitation unit to the qth receiving channel. Its magnitude represents the magnitude contribution, and its phase angle represents the phase contribution.

[0043] The objective function for silent control is established with the goal of minimizing the synthetic residual energy of the dominant propagation mode on the surface within the predetermined receiving region: In the formula, To detect the synthetic residual energy at frequency f; M represents the set of receiving channels corresponding to the predetermined receiving area; M is the number of excitation units. For the complex transfer coefficient from the nth excitation unit to the qth receiving channel; This represents the output amplitude of the nth excitation unit. Its initial excitation phase; It is the imaginary unit.

[0044] Using sequential quadratic programming (SQP), interior-point method, or genetic algorithm, find the optimal output within the rated output range of the equipment. The algorithm aims to minimize the optimal output amplitude and the optimal initial excitation phase, and within its constraints, limits the total output energy of all excitation units to below the safe excitation energy upper limit. This safe excitation energy upper limit is obtained by progressively increasing the excitation energy through calibrating the roadbed. When the rate of change of the low-frequency equivalent propagation velocity or the change in roadbed surface displacement after two adjacent excitation stages first exceeds the upper limit of the repeated detection error, the excitation energy of the preceding stage is determined as the safe excitation energy upper limit.

[0045] The initial silent control parameter set is formed by combining the detection frequency, surface dominant propagation mode number, excitation unit number, output amplitude, initial excitation phase, transmission time, predetermined receiving area, and parameter validity identifier; the parameter validity identifier is recorded as valid only when the solution results meet the equipment output range and safe excitation energy constraints.

[0046] S220: Only the initial silent control parameters with valid parameter identifiers are called to control the multi-point excitation array to output the first set of coherent mechanical waves. The first set of coherent mechanical waves referred to in this application refers to the mechanical wave combination in which multiple excitation units are excited in concert according to the same detection frequency, output amplitude, initial excitation phase and transmission time, and form destructive superposition of the surface dominant propagation mode within the predetermined receiving area.

[0047] The surface modal residual response is acquired by a multi-channel receiving array, and the surface modal residual energy ratio is calculated: In the formula, The ratio of residual energy in the surface modes; The surface dominant propagation mode response in the q-th receiving channel when using the current silent control parameters; This is the corresponding baseline response without coherent cancellation. This refers to the set of receiving channels corresponding to the predetermined receiving area.

[0048] The preset silence threshold is read from the calibration parameter library and is the larger of the 95th percentile of the residual energy ratio of the surface mode obtained from no less than 20 repeated tests on the calibrated roadbed and 0.20. The preset upper limit of residual energy per channel is the larger of the 95th percentile of the residual energy ratio of each effective receiving channel on the calibrated roadbed and 0.30.

[0049] when When the value is not lower than the preset silence threshold, perform amplitude and phase trial corrections sequentially on each excitation unit: first increase the current output amplitude by 1% to 5%, if... If the amplitude decreases, continue correcting in the direction of increasing amplitude; otherwise, output the phase of the amplitude decreasing. The phase is probed by 1° to 5° in both the increasing and decreasing directions, retaining the desired value. The direction of the reduction. Recalculate after each round of correction. .when Iteration stops when the threshold is lower than the preset silence threshold; iteration stops when two adjacent rounds are in progress. If the decrease is less than 0.01, or if the requirement is not met after 30 iterations, check the contact status of the excitation unit and remeasure the complex transfer coefficient.

[0050] The single-channel residual energy ratio of at least three spatially adjacent receiving channels is lower than the preset single-channel residual energy upper limit, and the region The continuous receiving space below a preset silence threshold is defined as the coherent silence zone. The final output amplitude, final excitation phase, final transmission time, coherent silence zone, and final... The final set of silent control parameters consists of the number of iterations and the valid identifier of silent control.

[0051] For example, the preset silence threshold is 0.20, initially... When the value was 0.29, it was reduced to 0.14 after 7 rounds of correction, and the single-channel residual energy ratio of the four adjacent receiving channels was all below 0.30, the coherent silence zone at this detection frequency was confirmed to be effective.

[0052] S230. For each detection frequency, only the final silence control parameter with the valid silence control flag is called, and the parameter corresponding to that detection frequency remains unchanged. The first set of coherent mechanical waves is output at least 3 times. Each detection... When all responses are below the preset silence threshold and the correlation coefficient between repeated responses is not less than 0.85, the coherent silence state is confirmed to be valid.

[0053] The surface dominant mode spatial response vector is constructed based on the surface dominant mode parameter set, and spatial coherent projection is used to remove residual components consistent with its channel amplitude distribution and channel phase delay. In the formula, This is the response vector composed of each receiving channel within the coherent silence zone; The surface-level dominant propagation modal space response vector; The remaining response vector after removing this modal spatial component; superscript This indicates the conjugate transpose operation.

[0054] This processing only removes components with the same spatial structure as the dominant propagation mode on the surface, retaining anomalous responses with the same frequency but different arrival times, propagation directions, or spatial phase structures. The unexcited period before excitation is used as the background noise time window, and the start time of the first analysis window when the response energy is lower than three times the background noise energy for three consecutive analysis windows or the number of effective receiving channels is less than three for three consecutive analysis windows is used as the wakeout time.

[0055] The signal-to-noise ratio of the q-th receiving channel is calculated using the following formula: In the formula, The signal-to-noise ratio (SNR) of the q-th receiving channel is expressed in dB. The root mean square amplitude within the effective response time window; This represents the root mean square amplitude of the background noise within the time window.

[0056] when When the corresponding receiving channel is invalid, the response from the earliest theoretical arrival time to the peak value of the main response envelope is marked as an early arrival response. The response from near the peak value of the main response envelope to the point where the response energy decreases to 10% of the peak energy is marked as the main scattering response. The continuous effective response from then to the tailwave cutoff time is marked as the late tailwave response.

[0057] The detection frequency, analysis bandwidth, receiving channel number, spatial coordinates, sampling time, residual response amplitude, response phase, response time period label, channel signal-to-noise ratio, coherent silent zone, and surface mode residual energy ratio are combined to form the first residual fluctuation response dataset, which is then output to S310. The analysis bandwidth is the larger value between the time-frequency analysis resolution of the corresponding detection frequency and the effective bandwidth of the excitation signal.

[0058] S3 specifically includes the following sub-steps: S310. Read the first residual fluctuation response dataset formed in S230, and refine the early arrival response time window of each receiving channel according to the early arrival response label. Calculate the background envelope mean and background envelope standard deviation of the response envelope using the background noise time window before excitation triggering; when the response envelope of a certain receiving channel has at least 3 consecutive sampling points higher than the sum of the background envelope mean and 6 times the background envelope standard deviation, determine the first sampling time that meets the condition as the candidate early arrival time, and extract the corresponding continuous response as the candidate early arrival component.

[0059] Background noise statistics are obtained from no fewer than 30 sets of background responses collected by the intelligent sensing system under unexcited conditions; background responses are re-collected when the intelligent sensing system is replaced, the receiving unit is reinstalled, or the on-site vibration environment changes.

[0060] The cumulative response energy is calculated based on the subject scattering response time window marked by S230. The moment when the cumulative response energy reaches 10% of the total energy of the time window is determined as the start time of the subject scattering response. The candidate early arrival time should be earlier than the start time of the subject scattering response, and the time difference between the two should be no less than the maximum value of the following three: 2 sampling periods, 1 / 4 of the period corresponding to the highest detection frequency, and 3 times the standard deviation of the arrival time of repeated detection.

[0061] The energy of the candidate early arrival component should also be no less than the threshold of the proportion of early arrival energy of all effective response energy of the corresponding receiving channel. The threshold is read from the calibration parameter library and is taken as the 95th percentile value of the proportion of incoherent early disturbance energy of the calibrated roadbed.

[0062] Calculate the normalized cross-correlation coefficient for candidate early arrival components in spatially adjacent receiving channels: In the formula, Let be the normalized cross-correlation coefficient between the a-th receiving channel and the b-th receiving channel; and For the corresponding candidate early arrival component; The effective duration of the candidate early arrival component; t represents the time shift; t is the time variable.

[0063] Only candidate early arrival components whose arrival time changes continuously with spatial distance, whose normalized cross-correlation coefficient is not lower than the connection correlation threshold, whose center frequency difference does not exceed the analysis bandwidth recorded by S230, and whose deviation between the phase increment of adjacent receiving channels after phase expansion and the median value of all phase increments does not exceed the phase increment deviation threshold in the calibration parameter library are connected as candidate early arrival wavefronts.

[0064] The candidate early arrival wavefront referred to in this application is a set of candidate early arrival components that appear consecutively in at least three spatially adjacent receiving channels and conform to a unified propagation relationship; the connection correlation threshold is taken as the 5th percentile of the normalized cross-correlation coefficient of the early mechanical wave in the calibrated roadbed stability stage. If the same pulse appears simultaneously in multiple receiving channels within one sampling period, and the arrival time does not change with spatial distance, the pulse is determined to be electrical crosstalk and discarded.

[0065] For example, if all eight receiving channels produce in-phase pulses 0.12 ms after excitation, and there is no distinguishable propagation time difference between the channels, this pulse will not be included in the candidate early arrival wavefront. The candidate early arrival wavefront set is composed of the participating receiving channels, arrival time, response waveform, center frequency, phase sequence, and spatial coverage, and output to S320.

[0066] S320. Perform multi-channel propagation fitting on each candidate early-arriving wavefront in the candidate early-arriving wavefront set: In the formula, Let q be the candidate early arrival time for the q-th receiving channel; and These are the planar coordinates of the corresponding receiving unit; For fitting the reference time; and These are the slow propagation components in two coordinate directions.

[0067] The equivalent propagation speed is taken as the reciprocal of the magnitude of the propagation slowness vector, and the propagation direction is taken as the propagation direction. Only candidate early-arriving wavefronts with a fitting determination coefficient of not less than 0.85 and a maximum arrival time residual not exceeding the minimum resolvable time interval are retained.

[0068] The median value of the equivalent propagation velocity of the main body scattering response in the first residual wave response dataset is taken as the current main body propagation velocity, and the upper limit of the normal propagation velocity in the calibration parameter library is taken as the velocity determination benchmark. When the equivalent propagation velocity of the candidate early arrival wavefront is higher than the product of the upper limit of the normal propagation velocity and the current main body propagation velocity and the velocity separation coefficient, the channel coherence is not lower than the preset coherence threshold, and the cross-channel duration is not less than the larger value of the coverage length of 3 adjacent receiving channels and 1 / 2 of the target shortest wavelength, it is determined as an abnormal early arrival wavefront.

[0069] The speed separation coefficient is determined based on the speed error of repeated detection of the calibrated roadbed and is taken as 1.20 to 1.50; the channel coherence is taken as the average of the normalized cross-correlation coefficients of adjacent channels, and the normalized cross-correlation coefficients of any two adjacent effective channels are required to be no less than 0.75.

[0070] The rigid propagation shortcut referred to in this application refers to the advantageous propagation path reflected by the abnormally early arrival wavefront, which has an equivalent propagation velocity higher than that of the surrounding main scattering medium and has a continuous spatial propagation range. It can be formed by boulders, large particle agglomerates, local high-rigidity regions or penetrating particle force chains, and does not require a unique determination of the specific physical cause.

[0071] The coverage area of ​​the receiving channel where the abnormally early arrival wavefront stably appears is used as the surface projection influence range. The propagation direction under different combinations of detection frequencies and array element phases is extended in reverse, and the candidate crossing region is determined by the overlapping region. The positioning confidence is determined based on the fitting determination coefficient, the maximum arrival time residual, and the channel coherence. When the fitting determination coefficient is lower than 0.85 or the maximum arrival time residual exceeds the minimum resolvable time interval, the valid identifier of the corresponding rigid propagation shortcut parameter is marked as invalid.

[0072] For example, if a wavefront covers 5 adjacent receiving channels, the equivalent propagation speed is 760m / s, the current main propagation speed is 260m / s, the upper limit of normal propagation speed is 340m / s, the channel coherence is 0.93, and the fitting determination coefficient is 0.96; when the velocity separation coefficient is 1.30 and the preset coherence threshold is 0.85, the wavefront is identified as an abnormally early arrival wavefront.

[0073] The rigid propagation shortcut parameter set is composed of the detection frequency, participating receiving channel, arrival time, equivalent propagation speed, propagation direction, amplitude distribution, phase distribution, surface projection influence range, candidate crossing region, fitting residual, channel coherence, location confidence, and parameter validity identifier, and is output to S330.

[0074] S330: Only call the rigid propagation shortcut parameters that are validly identified by the parameter flag, and establish the shortcut compensation transfer matrix while keeping the final silent control parameter set formed in S220 in effect. Based on the candidate crossing area and propagation direction, prioritize the excitation unit or auxiliary excitation unit that is closest to the surface projection of the corresponding advantageous propagation path and whose output direction intersects with the propagation direction; sequentially control the selected excitation units to input the probe mechanical wave at the detection frequency corresponding to the abnormal early arrival wavefront, so that the total energy of all probe mechanical waves does not exceed 10% of the safe excitation energy limit, and the complex response is collected by each receiving channel in the coherent silent zone.

[0075] Each excitation unit is repeatedly probed at least 3 times, and the average value is taken when the correlation coefficient of the repeated response is not less than 0.90. The complex responses of each excitation unit to each target receiving channel are arranged to form a shortcut compensation transfer matrix. .

[0076] Anomaly response vectors for early arrival waves are constructed based on the rigid propagation shortcut parameter set. And solve: In the formula, This is the vector of complex control coefficients for optimal compensation. Let the vector of complex control coefficients to be compensated be denoted. For shortcut compensation transfer matrix; This is the abnormal response vector; The 2-norm of a vector; The regularization coefficients are taken from the matrix. 0.01 to 0.10 times the largest eigenvalue; superscript This indicates the conjugate transpose operation.

[0077] In the specific solution, the optimal compensation complex control coefficient vector is directly calculated by multiplying the conjugate transpose of the shortcut compensation transfer matrix with itself, adding the product of the regularization coefficient and the identity matrix, inverting the resulting matrix, and then multiplying it by the conjugate transpose of the shortcut compensation transfer matrix and the abnormal response vector in turn.

[0078] The modulus of the optimal compensation complex control coefficient is converted into the compensation output amplitude, and the phase angle is converted into the compensation excitation phase. The receiving channel with the largest abnormal early arrival wavefront energy is selected as the reference channel. The compensation transmission time is obtained by subtracting the propagation delay from each compensation excitation unit to the reference channel from the time when the abnormal early arrival wavefront arrives at the reference channel.

[0079] When multiple separable early arrival anomalies exist, they are solved separately; when they cannot be separated, the anomaly response vectors are combined and solved jointly. The rigid propagation shortcut compensation parameter set is composed of the rigid propagation shortcut number, detection frequency, compensation excitation unit number, compensation output amplitude, compensation excitation phase, compensation transmission time, target receiving channel set, expected residual energy ratio of early arrival anomaly, total compensation energy, and valid parameter identifier, and is output to S410.

[0080] S4 specifically includes the following sub-steps: S410: Read the final silence control parameter set formed by S220 and the rigid propagation shortcut compensation parameter set formed by S330. Only write the parameters in the joint control process where both the silence control valid identifier and the compensation parameter valid identifier are valid, and match them according to the detection frequency, excitation unit number, coherent silence zone and target receiving channel set.

[0081] The composite blanking control parameter set referred to in this application is a set of joint excitation parameters formed after time alignment, complex synthesis and safety energy verification of the silent control parameters used to suppress the dominant propagation mode on the surface and the rigid propagation shortcut compensation parameters used to offset the abnormal early arrival wavefront.

[0082] For the nth excitation unit that simultaneously undertakes two types of control tasks, the composite output control quantity is obtained according to the following formula: In the formula, This is the composite output control quantity of the nth excitation unit at the detection frequency f; It is a silent complex control quantity, which is constructed in polar coordinate complex form from the final output amplitude and the final excitation phase of S220; To compensate for the complex control quantity, it is constructed into a polar coordinate complex form by the compensation output amplitude of S330 and the compensation excitation phase; To compensate for the time offset of the launch time relative to the unified trigger time; It is the imaginary unit.

[0083] The magnitude of the composite output control quantity is determined as the composite output amplitude, and its phase angle is determined as the composite excitation phase. When the multi-point excitation array and the auxiliary excitation array are independent of each other, the array controller outputs synchronously according to the unified clock established by S110.

[0084] The composite output amplitude of a single excitation unit shall not exceed 90% of the rated output amplitude, and the composite output energy of all excitation units shall not exceed the safe excitation energy limit determined by S210. If the limit is exceeded, the composite output amplitude shall be reduced proportionally while keeping the relative phase and relative emission time unchanged. The complex transfer coefficient obtained by S210 and the shortcut compensation transfer matrix obtained by S330 shall be used to re-predict the residual energy ratio of the surface mode and the residual energy ratio of the abnormal early arrival wavefront.

[0085] If any predicted residual energy ratio is still higher than the corresponding threshold after reduction, the two types of control tasks will be assigned to different excitation units, or an auxiliary excitation unit will be added and the solution will be recalculated.

[0086] The detection frequency, excitation unit number, composite output amplitude, composite excitation phase, composite emission time, coherent silence zone, target surface dominant propagation mode number, target rigid propagation shortcut number, total output energy, expected surface mode residual energy ratio, expected abnormal early arrival wavefront residual energy ratio, and parameter validity identifier are written into the composite blanking control parameter set, and the intelligent sensing system is controlled to output a second set of coherent mechanical waves. The second set of coherent mechanical waves, as referred to in this application, refers to the combination of mechanical waves output by the intelligent sensing system according to the composite blanking control parameter set, used to simultaneously suppress the surface dominant propagation mode and the abnormal early arrival wavefront.

[0087] For example, at 600Hz, the residual energy ratios predicted by the two sets of parameters individually are 0.12 and 0.10, respectively. However, after synthesis, the residual energy ratio of the surface mode is expected to rise to 0.26, indicating that the compensating mechanical wave has re-excited the dominant propagation mode of the surface. Therefore, the solution should be recombined and not directly performed on-site detection.

[0088] S420: The dual blanking residual response corresponding to the second group of coherent mechanical waves is acquired by a multi-channel receiving array, and time, amplitude, and phase corrections are completed using the array calibration parameter set. The dual blanking residual response referred to in this application refers to the multi-channel wave response retained after the surface dominant propagation mode and the anomalous early arrival wavefront have been respectively controlled. "Dual blanking" does not mean that the two types of wave components are physically completely zero.

[0089] Based on the spatial response vector of the dominant propagation mode formed in S130, the dominant propagation mode response is extracted by projection within the coherent silent region; based on the arrival time, participating receiving channel, amplitude distribution, and phase distribution recorded in S320, the abnormal early arrival wavefront response is extracted by matching within the abnormal early arrival time window, and the following calculations are performed respectively: In the formula, The ratio of residual energy in the surface modes after double blanking; The ratio of the abnormally early arrival wavefront residual energy after double blanking; This is the set of receiving channels corresponding to the coherent silence zone; For abnormally early arrival wavefronts participating in the receiving channel set; and These are the surface dominant propagation mode response and the anomalous early arrival wavefront response after double blanking, respectively. The baseline response of the surface dominance propagation mode when silent control is not implemented; This refers to the abnormally early wavefront baseline response when a coherent silent zone has been formed but rigid propagation shortcut compensation has not yet been implemented.

[0090] The preset silent threshold adopts the value of S220; the preset shortcut blanking threshold is read from the calibration parameter library and is the larger of the 95th percentile of the residual energy ratio of abnormal early arrival wavefront obtained from no less than 20 compensation tests of the calibration subgrade and 0.20.

[0091] When the residual energy ratio of the surface mode exceeds the limit, a 1% to 3% amplitude correction and a 1° to 3° phase correction are applied to the silent control component; when the residual energy ratio of the abnormally early arrival wavefront exceeds the limit, a 1% to 5% amplitude correction, a 1° to 5° phase correction, and a transmission time correction of no more than one sampling period are applied to the compensation component; when both exceed the limit simultaneously, the one with the larger normalized exceedance ratio is corrected first; the normalized exceedance ratio is the ratio of the current residual energy ratio to the corresponding threshold minus 1.

[0092] After each correction, the dual-hidden-empty residual response is reacquired. The correction is only accepted if the target residual energy ratio decreases and the other residual energy ratio does not exceed 110% of the value before correction. Iteration stops when the two types of residual energy ratios are lower than their corresponding thresholds. When the number of iterations reaches 30, or when the absolute change of the sum of the two types of residual energy ratios in two adjacent rounds is less than 0.01 and still does not meet the corresponding threshold, the complex transfer coefficient and shortcut compensation transfer matrix are remeasured.

[0093] The final parameters, the ratio of the two types of residual energy, the number of iterations, the total output energy, and the effective control identifier are combined to form the final dual blanking control parameter set.

[0094] S430. For each detection frequency, only the final double blanking control parameters with valid control flags are called, and the detection is repeated at least 3 times. The double blanking state is confirmed to be valid only when the residual energy ratio of each type is lower than the corresponding threshold, the correlation coefficient between repeated responses is not lower than 0.85, the coefficient of variation of the root mean square amplitude of the valid response of the same receiving channel is not higher than 10%, and there are not less than 3 valid receiving channels in the coherent silence zone.

[0095] By leveraging the surface advantage to propagate the modal spatial response vector and the spatiotemporal template of the early arrival wavefront, residual projections consistent with the wave components of the two types of targets are removed respectively. The signals are not filtered out as a whole according to the detection frequency, so as to retain effective responses with the same frequency but different arrival times, propagation directions or spatial phase structures.

[0096] The responses that appear consecutively after the end of the abnormal early arrival time window and before the peak value of the main scattering response envelope are marked as slow propagation candidate responses; the responses that appear after the peak value of the main scattering response envelope, with energy higher than 3 times the background noise energy and lasting for no less than 3 analysis windows are marked as scattering wake wave candidate responses; a covariance matrix is ​​constructed based on the 3-axis vibration components, and the main vibration direction is determined by the eigenvector corresponding to the largest eigenvalue. When the deflection of the main vibration direction relative to the main vibration direction of the surface dominant propagation mode determined by S130 exceeds the preset deflection angle and appears consecutively in at least 2 adjacent receiving channels, the corresponding response is marked as a mode conversion candidate response.

[0097] The preset deflection angle is read from the calibration parameter library, and the 95th percentile value of the change in the principal vibration direction of the same propagation mode of the calibrated roadbed is taken. The detection frequency, composite blanking control parameter number, receiving channel number, receiving unit spatial coordinates, sampling time, correction response amplitude, response phase, 3-axis vibration components, 3 types of candidate response labels, channel signal-to-noise ratio, coherent silent zone, two types of final residual energy ratio, repeated detection correlation coefficient, and data validity identifier are combined to form the second residual wave response dataset, and output to S510.

[0098] If any detection frequency cannot achieve double blanking after the maximum number of iterations, it is marked as an invalid frequency; if the effective detection frequency is less than 2 / 3 of the number of detection frequencies set in S120, the intelligent sensing system is redeployed and detection starts from S110.

[0099] S5 specifically includes the following sub-steps: S510: Read the second residual wave response dataset formed in S430, and select only the data whose valid identifier is valid, whose valid detection frequency is not less than 2 / 3 of the set number, and whose valid receiving channels are not less than 3 within the coherent silence zone. Extract the slow propagation candidate response, the scattering wake wave candidate response, and the mode transition candidate response. The slow propagation wave referred to in this application refers to the residual wave component that, after completing the double blanking of the surface dominant propagation mode and the anomalous early arrival wavefront, has an equivalent propagation velocity not higher than the 95th percentile of the scattering response velocity of the same type of filler in the calibration parameter library, and can be continuously tracked in at least 3 adjacent receiving channels.

[0100] Multi-channel fitting was performed on the arrival times of each slow-propagating wave to obtain propagation speed, propagation direction, and spatial coverage. Wave components with propagation direction differences not exceeding a direction merging threshold, relative propagation speed differences not exceeding a speed merging threshold, and spatial coverage overlap ratios not lower than a range overlap threshold were merged into a single remaining propagation path. Three thresholds were read from the calibration parameter library, taking the 95th percentile of the repeated detection direction and speed errors of the calibration roadbed, and the 5th percentile of the spatial coverage overlap ratio, respectively.

[0101] The total number of independent paths after deduplication across detection frequencies and composite blanking control parameter numbers is determined as the number of remaining propagation paths. For candidate responses of the scattering wake wave after the main scattering response ends, the normalized cross-correlation coefficients of adjacent receiving channels are calculated; the duration for which the response energy is higher than 3 times the background noise energy, the normalized cross-correlation coefficients of at least 3 adjacent receiving channels are not lower than the wake coherence threshold, and the coherence wake wave duration is determined as the duration of the coherent wake wave.

[0102] The ratio of the number of successfully connected dispersion intervals between adjacent effective detection frequencies to the total number of connectable frequency intervals is used to determine the continuity of the dispersion branch. The dispersion energy peaks at adjacent frequencies are considered to be successfully connected only when the relative difference in propagation speed, the difference in propagation direction, and the correlation coefficient of the spatial response vector meet the connection thresholds in the calibration parameter library.

[0103] The sum of the square integrals of the three-axis vibration responses within the corresponding time window for the slow propagation wave, scattered wake wave response, and mode transition response is divided by the total input energy of all excitation units at the detection frequency to obtain the normalized skeleton response energy of each effective receiving channel, and the uniformity of the multi-channel spatial response is calculated: In the formula, U represents the uniformity of the multi-channel spatial response; This represents the average value of the normalized skeleton response energy of each effective receiving channel within the coherent silence region. The corresponding standard deviation; To prevent positive numbers with a denominator of 0, we take 0.001 times the average skeleton response energy of the calibrated roadbed; when the calculation result is less than 0, we take 0, and when it is greater than 1, we take 1.

[0104] Calculate the coefficients of variation for the number of remaining propagation paths, coherent wake duration, dispersion branch continuity, and multi-channel spatial response uniformity in at least three repeated detections, and take the maximum value of the four coefficients of variation as the repeated detection coefficient of variation. Combine the above four features, the number of effective detection frequencies, and the repeated detection coefficient of variation to form a skeleton response feature set.

[0105] For example, if the propagation directions of the two sets of responses at 600Hz and 900Hz are 31° and 35° respectively, the propagation speeds are 215m / s and 228m / s respectively, and the spatial coverage overlap ratio is 82%, then if the corresponding thresholds are 15°, 15% and 60% respectively, the two sets of responses will be merged into the same remaining propagation path to avoid double counting.

[0106] S520. On a calibration subgrade using the same type of filler as the target subgrade, according to the same intelligent sensing system layout, detection frequency, composite blanking control rules and feature calculation rules as S110 to S510, obtain a calibration skeleton response feature set of no less than 30 independent detection areas.

[0107] The on-site dry density of each calibration area was measured using the sand cone method, ring cutter method, or calibrated nuclear density method. The maximum dry density of the same batch of filler was obtained from the corresponding indoor compaction test. The calibrated compaction degree was calculated as the ratio of the on-site dry density to the maximum dry density multiplied by 100%. The test information was recorded in the calibration record. The calibrated compaction degree should cover at least 3 percentage points above and below the design compaction degree.

[0108] The number of remaining propagation paths and the duration of coherent wake waves were standardized using the mean and standard deviation of the calibrated samples, and the calibration relationship was established or updated using multiple linear regression. In the formula, This is the initial value for effective compaction. This represents the number of remaining propagation paths after standardization. The standardized coherent wake duration; U represents the continuity of the dispersion branch; U represents the uniformity of the multi-channel spatial response. to These are the regression coefficients obtained from the calibration data using the least squares method.

[0109] A 5-fold cross-validation method is used to evaluate the calibration relationship, requiring the root mean square error of compaction degree prediction to be no greater than 1.5 percentage points and the coefficient of determination to be no less than 0.80. If these requirements are not met, additional calibration samples are added, or calibration relationships are established separately according to filler type and moisture content range. When any skeleton response characteristic of the target subgrade is lower than the minimum value of the calibration sample, or higher than the maximum value of the calibration sample, and the excess exceeds 10% of the span of the calibration range of that characteristic, it is marked as an extrapolation detection, and the final effective compaction degree is not directly output.

[0110] Each feature is linearly mapped to 0 to 1 according to the lower limit of non-compliance and the median of compliance in the calibrated samples. Then, the normalized value of the number of remaining propagation paths is combined with U in a weighted manner to obtain the particle contact network integrity. The normalized value of the coherent wake duration is compared with... Weighted combination for particle contact network stability The weight is the proportion of the absolute correlation coefficient between the corresponding feature and the measured compaction degree to the sum of the two absolute correlation coefficients, and the correlation direction is determined based on the calibration data.

[0111] The data confidence level is calculated by taking the arithmetic mean of the normalized effective detection frequency ratio, effective receiving channel ratio, repeat detection consistency, and calibration model error term. Among them, the repeat detection consistency is 1 minus the repeat detection variation coefficient, and the calibration model error term is 1 minus the ratio of the root mean square error of cross-validation to the allowable error. All terms are limited to 0 to 1.

[0112] Initial value of effective compaction, , The effective compaction state parameter set consists of calibration model number, cross-validation error, applicable scope judgment, data confidence level and parameter validity identifier, and is output to S530.

[0113] S530. Only call the set of effective compaction state parameters whose parameter validity is marked as valid, and correct the initial value of effective compaction degree according to the following formula: In the formula, For effective compaction; This is the initial value for effective compaction. and These are, respectively, the integrity of the particle contact network and the stability of the particle contact network; and These are the 5th percentile values ​​for particle contact network integrity and particle contact network stability in the compaction-qualified calibration samples, respectively. and These are the integrity correction coefficient and the stability correction coefficient, respectively, which are obtained by using calibration data with the goal of minimizing the mean square error between the corrected compaction degree and the measured compaction degree. This indicates taking the larger value between 0 and the value within the parentheses.

[0114] The effective compaction degree must not exceed the initial value of the effective compaction degree. The acceptable compaction degree threshold is derived from engineering design documents, construction acceptance standards, or project quality control documents.

[0115] When the effective compaction degree is lower than the qualified compaction degree threshold, and and When all reach the corresponding qualified threshold, it is judged as an overall under-compaction risk; when and Simultaneously, if the compaction degree is below the corresponding qualified threshold, it is judged as a composite under-compaction risk. Specifically, if the effective compaction degree is not lower than the qualified compaction threshold, it is marked as a composite latent under-compaction risk; when only... Below When the risk is determined to be localized undercompaction or a single bearing path type, and the effective compaction degree is not lower than the qualified compaction degree threshold, it is marked as the corresponding latent undercompaction risk; when only Below When the contact network is unstable, it is identified as an undercompaction risk. When the effective compaction degree is not lower than the qualified compaction degree threshold, it is marked as the corresponding latent undercompaction risk.

[0116] If the maximum deviation of any indicator from its corresponding threshold does not exceed 5%, is greater than 5% but not more than 15%, or is greater than 15%, it is classified as a Level 1, Level 2, or Level 3 risk, respectively. If the effective detection frequency is insufficient, the skeleton response characteristics exceed the calibration range, the cross-validation error of the calibration model exceeds the limit, the repeated detection coefficient of variation exceeds 10%, or the data confidence level is lower than the project's set lower limit, the detection results will be output pending review.

[0117] For example, the acceptable compaction threshold is 95%, and the initial effective compaction value is 96.3%. It is 0.56. It is 0.82. It is 0.70. It is 0.68. When the effective compaction degree is 10.71 percentage points, the integrity reduction is 10.71 × (0.70 - 0.56) = 1.50 percentage points, and the corrected effective compaction degree is 94.8%. Since the effective compaction degree is less than 95% and the particle contact network integrity is insufficient, it is judged as a risk of under-compaction due to local bridging or single bearing path. The effective compaction degree, risk type, risk level and data confidence level are output.

[0118] All the above formulas are performed using dimensionless numerical calculations; the relevant formulas are based on empirical models that approximate the real situation, obtained through extensive data collection and software simulation fitting. The preset parameters and thresholds involved in the formulas can be conventionally set and adjusted by those skilled in the art according to the physical constraints of the actual application scenario.

[0119] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0120] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0121] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A rapid non-destructive testing method for roadbed compaction degree based on multiphase wave dispersion characteristics, characterized in that, Includes the following steps: S1. Deploy an intelligent sensing system on the roadbed surface, control a multi-point excitation array to input multi-phase mechanical waves according to the combination of detection frequency and array element phase, collect multi-channel responses, construct a dispersion response matrix and identify the dominant propagation mode of the surface layer. S2. Based on the channel amplitude distribution and channel phase delay of the surface dominant propagation mode, adjust the output amplitude, initial phase and transmission time of each excitation unit so that they cancel each other out in the predetermined receiving area, forming a coherent silent zone and obtaining the first residual wave response dataset. S3. Extract candidate early arrival wavefronts from the first residual wave response dataset, identify abnormal early arrival wavefronts based on propagation speed, propagation direction, channel coherence and spatial continuity, determine rigid propagation shortcuts and generate rigid propagation shortcut compensation parameter sets. S4. The final silent control parameter set and the rigid propagation shortcut compensation parameter set are combined in complex form to output the second set of coherent mechanical waves. The composite blanking control parameters are corrected according to the residual energy ratio of the surface mode and the residual energy ratio of the anomalous early arrival wavefront to obtain the second residual wave response dataset.

2. The rapid non-destructive testing method for roadbed compaction based on multiphase wave dispersion characteristics according to claim 1, characterized in that, Also includes: S5. Extract the skeleton response feature set from the second residual fluctuation response dataset, determine the effective compaction degree, particle contact network integrity and particle contact network stability according to the calibration relationship, and obtain the hidden undercompaction risk.

3. The rapid non-destructive testing method for roadbed compaction based on multiphase wave dispersion characteristics according to claim 1, characterized in that, S1 specifically includes: A smart sensing system is deployed on the surface of the target roadbed to collect and calibrate the mechanical wave response, determine the channel time deviation, amplitude correction coefficient and phase deviation, and form an array calibration parameter set. The detection frequency and array element phase combination are determined based on the fill layer thickness, the initial value of the equivalent propagation velocity, and the operating frequency band of the equipment. Multiphase mechanical waves are output and the original wave response is corrected to form a calibration wave response dataset.

4. The rapid non-destructive testing method for roadbed compaction based on multiphase wave dispersion characteristics according to claim 3, characterized in that, It also includes: performing time-frequency decomposition and wavenumber domain superposition on the calibration fluctuation response dataset to construct an initial dispersion response matrix, connecting the local maximum points of dispersion energy to form dispersion branches, and identifying the dominant propagation modes of the surface layer based on the surface sensitive frequency band, modal energy ratio and phase continuity to form a parameter set of the dominant mode of the surface layer.

5. The rapid non-destructive testing method for roadbed compaction based on multiphase wave dispersion characteristics according to claim 1, characterized in that, S2 specifically includes: Based on the surface dominant mode parameter set and array calibration parameter set, the predetermined receiving area is determined, the complex transfer coefficient is obtained, and the output amplitude, initial excitation phase and transmission time are obtained with the goal of minimizing the synthetic residual energy, thus forming the initial quiescent control parameter set. The first set of coherent mechanical waves is output according to the initial silence control parameter set. The output amplitude and initial excitation phase are iteratively corrected according to the residual energy ratio of the surface mode to form the coherent silence zone and the final silence control parameter set. The responses were collected while keeping the final silent control parameters unchanged. Residual components consistent with the surface dominant propagation mode were removed, and the early arrival response, main scattering response, and late tail wave response were labeled to form the first residual wave response dataset.

6. The rapid non-destructive testing method for roadbed compaction based on multiphase wave dispersion characteristics according to claim 1, characterized in that, S3 specifically includes: Candidate early arrival components that meet the requirements of arrival time, energy ratio, cross-channel correlation and phase continuity are extracted from the first residual wave response dataset. Electrical crosstalk is removed and they are connected to form a candidate early arrival wavefront set. Multi-channel propagation fitting is performed on candidate early arrival wavefronts. Abnormal early arrival wavefronts are identified based on equivalent propagation speed, propagation direction, channel coherence, and cross-channel duration. Rigid propagation shortcuts are determined and a rigid propagation shortcut parameter set is formed.

7. The rapid non-destructive testing method for roadbed compaction based on multiphase wave dispersion characteristics according to claim 6, characterized in that, Also includes: When the final silent control parameters remain in effect, a shortcut compensation transfer matrix is ​​established. The compensation output amplitude, compensation excitation phase, and compensation launch time are obtained based on the abnormal response vector, forming a rigid propagation shortcut compensation parameter set.

8. The rapid non-destructive testing method for roadbed compaction degree based on multiphase wave dispersion characteristics according to claim 1, characterized in that, S4 specifically includes: The final silent control parameter set and the rigid propagation shortcut compensation parameter set are matched according to the detection frequency. The two types of complex control quantities are aligned, synthesized and energy verified to form a composite blanking control parameter set and output the second set of coherent mechanical waves. Collect the dual blanking residual response, calculate the residual energy ratio of the surface mode and the residual energy ratio of the anomalous early arrival wavefront, and alternately correct the two types of control components according to the corresponding thresholds to form a dual blanking control parameter set; Repeated detection is performed using the dual blanking control parameter set to remove projections consistent with the surface dominant propagation mode and anomalous early arrival wavefronts. Slow propagation candidate responses, scattering tail wave candidate responses, and mode transition candidate responses are then labeled to form a second residual wave response dataset.

9. The rapid non-destructive testing method for roadbed compaction based on multiphase wave dispersion characteristics according to claim 2, characterized in that, S5 specifically includes: Slow propagation waves, scattered tail waves, and mode-converted waves are extracted from the second residual wave response dataset. The number of residual propagation paths, coherent tail wave duration, dispersion branch continuity, and multi-channel spatial response uniformity are calculated to form a skeleton response feature set. Based on the measured compaction degree and skeleton response characteristics of the calibrated subgrade, a calibration relationship is established to obtain the initial value of effective compaction degree, and the particle contact network integrity, particle contact network stability and data confidence are determined to form an effective compaction state parameter set.

10. The rapid non-destructive testing method for roadbed compaction based on multiphase wave dispersion characteristics according to claim 9, characterized in that, Also includes: The initial value of effective compaction is corrected based on the integrity and stability of the particle contact network, and the effective compaction, under-compaction risk type, and risk level are determined in conjunction with the qualified compaction threshold.

Images