A wharf surface layer concrete crack self-adapting monitoring and cooperative control system

Abstract

The application relates to the technical field of structural health monitoring and nondestructive testing, and discloses a wharf surface layer concrete crack self-adaptive monitoring and collaborative control system, which comprises an edge scheduling module, a common mode stripping module, a constraint inversion module and a collaborative reconstruction module.The edge scheduling module is used for generating node pair reliability; the common mode stripping module is used for constructing a prior probability field matrix by combining a current relative tensile factor, a virtual reference tensile factor and the node pair reliability; the constraint inversion module is used for extracting an abnormal grid area, injecting a continuous high-frequency wave to obtain a difference frequency envelope time sequence, generating a forward observation operator based on an influence line kernel function matrix, solving a sparse regularization equation to obtain a crack interface contact stiffness distribution; the thermal force hysteresis checking module is used for calculating a thermal force hysteresis loop envelope area and outputting a structural crack risk label; and the collaborative reconstruction module is used for calculating inference information entropy and switching a transmitting and receiving mode of a piezoelectric ceramic node array.The application weakens the interference of temperature change environment and working conditions, and improves the adaptive evaluation and dynamic tracking capability of the crack state.

Description

Technical Field

[0001] This invention relates to the field of structural health monitoring and non-destructive testing technology, specifically to an adaptive monitoring and collaborative control system for cracks in the concrete surface layer of a wharf. Background Technology

[0002] In the early stages of pouring, the concrete surface layer of high-pile wharves is prone to initial microcracks due to a combination of factors, including heat of hydration and shrinkage deformation. After commissioning, due to long-term exposure to the open-air service environment of the port, these microcracks induced by early factors will continue to evolve and expand under the influence of temperature fluctuations, wet-dry cycles, and repeated dynamic loads from gantry cranes. Currently, structural health monitoring of wharf surface concrete mainly relies on ultrasonic or piezoelectric sensor networks to acquire acoustic signals to assess the cracking status of the structure.

[0003] However, in actual dock operations and the natural environment, drastic temperature fluctuations can cause significant drift in the baseline of sound wave propagation characteristics. Existing monitoring technologies struggle to effectively isolate such common-mode interference from monitoring signals, easily misinterpreting fluctuations in routine environmental parameters as structural physical damage. Furthermore, microcracks induced by the pouring process and early natural environment are often in a physically closed state without external loads, making it difficult for conventional static detection to capture their subtle acoustic response characteristics.

[0004] Furthermore, the sensor network topology of existing monitoring systems is usually fixed, lacking dynamic focusing and active detection capabilities when faced with data logic conflicts caused by environmental noise. It also fails to effectively combine the unique working conditions such as the dynamic concentrated load of gantry crane operations at the dock to perform stress activation and multi-dimensional cross-verification of hidden closed cracks, resulting in the accuracy of the monitoring system being limited under complex interference conditions.

[0005] Therefore, this invention proposes an adaptive monitoring and collaborative control system for cracks in the concrete surface layer of a wharf to address the shortcomings of existing technologies. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides an adaptive monitoring and collaborative control system for cracks in the concrete surface layer of a wharf. This system solves the problems of existing monitoring systems being unable to eliminate misjudgments caused by complex environmental interference, and conventional static detection being unable to effectively identify closed microcracks and lacking dynamic adaptive focusing detection capabilities.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an adaptive monitoring and collaborative control system for cracks in wharf surface concrete, comprising: The edge scheduling module is used to issue wake-up commands to the piezoelectric ceramic node array, and perform frequency domain scanning decoupling and bidirectional acoustic reciprocity testing to generate piezoelectric ceramic node pair reliability. The common-mode stripping module is used to obtain the current relative stretching factor and the virtual reference stretching factor, and combine the current relative stretching factor, the virtual reference stretching factor and the piezoelectric ceramic node to perform weighted spatial difference calculation on the reliability to construct a prior probability field matrix; The constrained inversion module is used to extract abnormal grid regions based on the prior probability field matrix, inject continuous high-frequency waves into the abnormal grid regions to extract the difference frequency envelope time series, construct a sparse regularized equation by combining the influence line kernel function matrix constructed based on the instantaneous load, solve the sparse regularized equation to obtain the contact stiffness distribution of the crack interface, and output the stiffness anomaly coordinates. The thermal hysteresis verification module is used to extract response features under different thermal phases based on the stiffness anomaly coordinates, generate heating trajectory lines and cooling trajectory lines, calculate the envelope area of ​​the thermal hysteresis loop enclosed by the heating trajectory lines and the cooling trajectory lines, and output structural crack risk labels. The collaborative reconstruction module is used to aggregate multi-dimensional observation indicators, including the envelope area of ​​the thermal hysteresis loop, into a joint evidence vector, perform attribution arbitration, and calculate the inferred information entropy. Based on the inferred information entropy, it issues a topology scheduling command to switch the transmit and receive modes of the piezoelectric ceramic node array.

[0008] Preferably, the edge scheduling module calculates the complex electromechanical admittance spectrum by applying voltage excitation to the piezoelectric ceramic node array and acquiring the current response, and extracts the quality factor of the main mechanical resonance peak from the conductance spectrum of the complex electromechanical admittance spectrum; the edge scheduling module schedules adjacent piezoelectric ceramic node pairs to interactively transmit and receive low-frequency ultrasonic waves, and extracts the bidirectional reciprocity residuals of the positive and negative wake signals; the edge scheduling module calculates the reliability of the piezoelectric ceramic node pair based on the quality factor of the main mechanical resonance peak, the initial state calibration value of the quality factor of the main mechanical resonance peak, and the bidirectional reciprocity residuals.

[0009] Preferably, the common-mode stripping module constructs a current state vector, which includes the gantry section, travel direction, surface temperature, heating / cooling phase, surface temperature change rate, susceptance baseline slope drift, and a binary vector of effective piezoelectric ceramic node availability. The common-mode stripping module calculates the Mahalanobis distance between the current state vector and historical state vectors in the historical baseline database. The common-mode stripping module selects the stretching factor corresponding to the historical state vector that satisfies the matching threshold and has the smallest Mahalanobis distance as the virtual reference stretching factor.

[0010] Preferably, the common-mode stripping module calculates the cross-correlation coefficient between the current wake signal and the local baseline waveform signal under different time axis stretching ratios, and uses the stretching ratio corresponding to the maximum value of the cross-correlation coefficient as the current relative stretching factor; the common-mode stripping module calculates the difference between the current relative stretching factor and the virtual reference stretching factor, uses the piezoelectric ceramic nodes to weight the difference based on reliability, and maps the weighted difference onto a two-dimensional grid to construct the prior probability field matrix through an exponential decay kernel function based on the first Fresnel zone.

[0011] Preferably, the constraint inversion module selects grids with probability values ​​greater than the probability threshold from the prior probability field matrix as the anomalous grid regions by setting a probability threshold; the constraint inversion module schedules piezoelectric ceramic nodes at the edge of the anomalous grid regions to synchronously transmit two continuous high-frequency waves with different center frequencies, and collects aliased signals that penetrate the anomalous grid regions; the constraint inversion module performs low-pass filtering on the aliased signals to filter out high-frequency carrier components, and performs Hilbert transform on the filtered signals to extract the envelope modulus to generate the difference frequency envelope time series, and uses the difference frequency envelope time series as the observation vector of the sparse regularization equation.

[0012] Preferably, the constraint inversion module calculates the influence coefficient of the instantaneous load of the gantry crane wheel assembly on the normal stress of each spatial grid point in the discretized grid of the target area based on the Buschnesk spatial force attenuation equation, combines them to generate the influence line kernel function matrix, and generates the forward modeling observation operator of the sparse regularization equation based on the influence line kernel function matrix; the constraint inversion module converts the prior damage probability value of each grid point in the prior probability field matrix into penalty weights to construct a diagonal penalty matrix, uses the diagonal penalty matrix as a regularization penalty term to solve the sparse regularization equation to obtain the target state vector, and uses the target state vector as the contact stiffness distribution of the crack interface.

[0013] Preferably, after obtaining the contact stiffness distribution at the crack interface, the constraint inversion module extracts mesh elements with stiffness values ​​lower than a preset health baseline threshold and performs spatial clustering analysis on the stiffness connectivity of adjacent meshes; when the cumulative area of ​​continuous low-stiffness meshes exceeds a preset physical damage lower limit scale, the constraint inversion module outputs the two-dimensional mesh coordinates corresponding to the continuous low-stiffness meshes as the stiffness anomaly coordinates.

[0014] Preferably, the thermal hysteresis verification module filters historical monitoring events where the distance deviation between the gantry crane's passage coordinates and the stiffness anomaly coordinates, as well as the difference in the gantry crane's travel speed, are within a preset range. The filtered historical monitoring events are then divided into temperature bins according to the surface temperature of the surface layer. Heating phase events and cooling phase events are extracted separately within the same temperature bin. The thermal hysteresis verification module divides the extracted difference frequency envelope peak value by the instantaneous total load at the corresponding moment to obtain the normalized difference frequency response as the response feature. The thermal hysteresis verification module uses a least-squares polynomial to fit the normalized difference frequency response under the heating and cooling phases to generate the heating trajectory line and the cooling trajectory line, respectively. When the calculated thermal hysteresis loop envelope area is greater than the structural crack thermal hysteresis area threshold and the maximum longitudinal difference between the heating trajectory line and the cooling trajectory line is greater than the lower limit of the extreme value, the thermal hysteresis verification module generates a structural crack risk label for the stiffness anomaly coordinates.

[0015] Preferably, the multi-dimensional observation indicators collected by the collaborative reconstruction module include the quality factor drop of the main mechanical resonance peak, bidirectional reciprocity residual, prior damage probability value, inversion residual, and the envelope area of ​​the thermodynamic hysteresis loop; the collaborative reconstruction module establishes a Bayesian network containing the hypothetical states and the evidence nodes corresponding to the multi-dimensional observation indicators, and calculates the posterior probability distribution of each hypothetical state using the conditional probabilities statistically obtained from the attribution sample database; the collaborative reconstruction module calculates the inferred information entropy of the logical conflict grid based on the posterior probability distribution of each hypothetical state according to the principle of information entropy.

[0016] Preferably, the collaborative reconstruction module statistically analyzes the fluctuation state of the inferred information entropy over multiple consecutive analysis cycles, and extracts spatial grids whose inferred information entropy is greater than the judgment threshold over multiple consecutive analysis cycles as high-conflict grids; the collaborative reconstruction module issues the topology scheduling command to the piezoelectric ceramic node array drive control unit, the topology scheduling command instructs the piezoelectric ceramic nodes at the edge of the high-conflict grid to adjust the transmission delay, switch to directional acoustic beam focusing mode using the phased array interference principle, perform supplementary calculations in the high-conflict grid region, and refresh the joint evidence vector.

[0017] This invention provides an adaptive monitoring and collaborative control system for cracks in the concrete surface layer of a wharf. It has the following beneficial effects: 1. This invention constructs a closed-loop collaborative monitoring mechanism encompassing state assessment, environmental compensation, mechanism inversion, and dynamic scheduling. The system generates the reliability of piezoelectric ceramic node pairs through an edge scheduling module. This, combined with the virtual reference tensile factor obtained by the common-mode stripping module and the current relative tensile factor, constructs a priori probability field matrix. This matrix guides the constraint inversion module to obtain the contact stiffness distribution at the crack interface by combining the influence line kernel function matrix. The thermal hysteresis verification module then extracts the envelope area of ​​the thermal hysteresis loop for risk verification. Finally, the collaborative reconstruction module integrates multi-dimensional observation indicators and switches the transmit / receive modes of the piezoelectric ceramic node array based on the inferred information entropy feedback. This multi-module collaboration weakens the interference of external environment and equipment operating conditions on acoustic signals, improving the system's dynamic tracking and adaptive assessment capabilities for the state of cracks in concrete surfaces.

[0018] 2. This invention utilizes Mahalanobis distance to match historical states to obtain a virtual reference stretching factor, and then subtracts it from the current relative stretching factor obtained by cross-correlation with the wake wave. Based on this, piezoelectric ceramic nodes are introduced to weight the reliability, and an exponential decay kernel function based on the first Fresnel zone is used to map the difference features to a two-dimensional mesh. This design physically separates common-mode trend interferences such as temperature baseline drift, providing prior space constraints for damage feature extraction.

[0019] 3. This invention calculates the normal stress influence coefficient corresponding to the instantaneous load of the gantry crane based on the Buschnesk spatial stress attenuation equation, generates an influence line kernel function matrix, and generates a forward modeling observation operator based on the influence line kernel function matrix. The system injects continuous high-frequency waves into the abnormal grid region to extract the difference frequency envelope time series, and solves the sparse regularization equation by combining it with the diagonal penalty matrix obtained from the prior probability transformation. This process utilizes the dynamic stress field of the heavy-load equipment during operation to achieve spatial quantification of the contact stiffness of the crack interface, reducing the probability of missing closed microcracks under static testing.

[0020] 4. This invention combines thermal hysteresis verification with an information entropy-driven mechanism for secondary confirmation and calculation scheduling. By calculating the normalized difference frequency response of the heating and cooling phases under the same temperature compartment, the envelope area of ​​the thermal hysteresis loop enclosed by the heating and cooling trajectories is obtained. The system distinguishes between structural cracks and shallow surface cracks from a mechanical response perspective. Simultaneously, the inferred information entropy based on Bayesian network calculations can locate logically high-conflict regions and drive the piezoelectric ceramic node array to locally switch to directional acoustic beam focusing mode to perform supplementary calculations. This improves the arbitration confidence of local suspicious regions while balancing the overall system energy consumption. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the system architecture according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the edge scheduling and reliability generation process according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the common mode stripping and prior probability field matrix construction process according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the constraint inversion and localization process for crack interface contact stiffness anomalies according to an embodiment of the present invention. Figure 5 This is a schematic diagram of the thermal hysteresis verification process according to an embodiment of the present invention; Figure 6 This is a schematic diagram of the collaborative reconfiguration and topology scheduling control process according to an embodiment of the present invention; Figure 7 This is a schematic diagram comparing the prior probability field matrix and the crack interface contact stiffness distribution in an embodiment of the present invention; wherein, (a) is a contour map of the prior probability field matrix, and (b) is a crack interface contact stiffness distribution map after constraint inversion. Figure 8 This is a schematic diagram of the hysteresis evolution trajectory of the crack thermal response according to an embodiment of the present invention; Figure 9 This is a schematic diagram illustrating the changes in the posterior probability of competing hypotheses and the inference information entropy in an embodiment of the present invention. Figure 10 This is a comparison curve of the long-term operating performance of the system according to an embodiment of the present invention.

[0022] Among them, 100 is the edge scheduling module; 200 is the common mode stripping module; 300 is the constraint inversion module; 400 is the thermal hysteresis verification module; and 500 is the collaborative reconstruction module. Detailed Implementation

[0023] The technical solutions in 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.

[0024] Reference Figure 1 The present invention provides an adaptive monitoring and collaborative control system for cracks in the concrete surface layer of a wharf, which may include: The edge scheduling module 100 is used to perform front-end hardware resource allocation preset and underlying physical state self-check, maintain hardware underlying node time synchronization and remove degraded nodes.

[0025] The common-mode stripping module 200 is used to eliminate common-mode interference of environmental changes on acoustic measurements based on similar working condition matching logic, complete spatial differential mapping, and construct an initial damage probability field.

[0026] The constrained inversion module 300 is used to perform constrained inversion calculations by using the gantry motor load as a trigger source and combining it with the influence line forward model to locate the contact stiffness anomaly coordinates of the crack interface.

[0027] The thermal hysteresis verification module 400 is used to perform load pairing and alignment on monitoring data under different temperature rise and fall conditions, and to determine the thermal response hysteresis attribute and structural crack risk of the crack corresponding to the stiffness anomaly coordinate based on the temperature response trajectory envelope characteristics.

[0028] The collaborative reconstruction module 500 is used to process multidimensional observation data, perform attribution arbitration, quantify the uncertainty of spatial inference, and issue reconstruction instructions to change the underlying topology.

[0029] This invention provides an adaptive monitoring and collaborative control method for cracks in the concrete surface layer of a wharf, comprising the following steps: S100 sends a wake-up command to the piezoelectric ceramic node array before the gantry crane enters the target area to maintain sub-microsecond clock synchronization, and performs frequency domain scanning decoupling and bidirectional acoustic reciprocity test within the no-load time window to generate piezoelectric ceramic node pair reliability. S200 encodes kinematic parameters, thermal parameters, and nodal state parameters into state vectors and retrieves and extracts virtual reference stretching factors. It then performs weighted spatial difference calculations in conjunction with the current relative stretching factors to construct a prior probability field matrix that reflects the damage distribution. S300, inject coherent continuous high-frequency waves into the abnormal grid region indicated by the prior probability field matrix to extract the difference frequency envelope time series, combine the forward modeling observation operator generated by the influence line kernel function matrix with the prior probability field matrix to construct a sparse regularized equation with a penalty term, and solve it to obtain the contact stiffness distribution of the crack interface. S400 retrieves gantry crane passage events with stiffness anomaly coordinates at different thermodynamic phases, extracts the difference frequency envelope peak value for load normalization, calculates the area of ​​the thermodynamic hysteresis loop envelope formed by the heating trajectory line and the cooling trajectory line, and determines the thermal response hysteresis attribute of the crack and the structural crack risk. The S500 aggregates multi-dimensional observation indicators into a joint evidence vector to perform attribution arbitration, calculates the inferred information entropy of the logical conflict grid, and performs targeted supplementary calculations on the array transceiver topology for the extreme value region of information entropy.

[0030] To further clarify the implementation of each technical aspect of the present invention, the following will provide a detailed description of the implementation of each functional module involved above and its internal processing flow.

[0031] It should be noted that the piezoelectric ceramic node array used in this embodiment can physically utilize commercially available surface-mount piezoelectric ceramic patches (PZT patches) or pre-embedded smart aggregates. The nodes can communicate with the aforementioned edge scheduling module via wired cables or wireless sensor networks; the arrangement of the nodes can be grid-like or honeycomb-like topological distribution depending on the geometry of the target area. This invention does not limit the internal wafer structure or material formulation of the piezoelectric ceramic sensor itself.

[0032] Reference Figure 2 S100, based on the edge scheduling module 100, sends a wake-up command to the piezoelectric ceramic node array to maintain clock synchronization before the gantry crane enters the target area, and performs frequency domain scanning decoupling and bidirectional acoustic reciprocity testing within the no-load time window to generate the reliability of the piezoelectric ceramic node pairs. Specifically, it includes the following sub-steps: S101, Array pre-activation and timing synchronization network construction based on gantry crane trajectory prediction. In this embodiment, the edge scheduling module 100 receives gantry crane trajectory data from the terminal's programmable logic controller. Based on this gantry crane trajectory data, the edge scheduling module 100 parses the current position and running speed of the gantry crane, and calculates the estimated time for the gantry crane wheel assembly to reach the target area by combining the physical coordinates of the target area. When the estimated time is less than a time threshold, the edge scheduling module 100 sends a wake-up command to the piezoelectric ceramic nodes in the target area via industrial Ethernet. As a preferred implementation, the time threshold can be set to a range of 30s to 60s, and its specific value is determined by the sum of the system-preset hardware wake-up time of the piezoelectric ceramic node and the underlying network addressing time, to ensure that the nodes have sufficient warm-up and self-test cycles.

[0033] After the piezoelectric ceramic node is woken up, the edge scheduling module 100 initiates the clock synchronization process. Because high-frequency acoustic coherent measurements require extremely high time alignment accuracy, the edge scheduling module 100 adopts a master-slave clock architecture similar to a precise time protocol. This process includes master source time synchronization, local time calibration, phase holding within the working window, and a resynchronization mechanism after communication jitter.

[0034] The edge scheduling module 100, acting as the master control source, periodically sends synchronization messages carrying precise timestamps to the piezoelectric ceramic nodes. The piezoelectric ceramic nodes receive the synchronization messages and perform local time synchronization calculations. Specifically, the nodes calculate the one-way transmission delay of the network link by subtracting the transmission time of the bidirectional messages from the local reception time, and then compensate for and align their local clock registers.

[0035] Within the no-load time window, the piezoelectric ceramic node maintains clock phase using its internal crystal oscillator. The no-load time window refers to the physical period during which the gantry crane wheels have not yet entered the target area and the concrete surface of the target area is unaffected by the gantry crane wheel pressure. To prevent clock drift caused by the complex electromagnetic environment, when communication jitter in the industrial Ethernet link causes the piezoelectric ceramic node to lose 3 to 5 synchronization messages consecutively, the piezoelectric ceramic node actively sends an alignment request to the edge scheduling module 100 to trigger the resynchronization mechanism. The above clock synchronization process aims to provide the necessary clock synchronization accuracy for subsequent coherent excitation and bidirectional reciprocal comparison.

[0036] S102, edge-end no-load electromechanical admittance frequency sweep and characteristic decoupling calculation. Within the no-load time window, the piezoelectric ceramic node drives the local impedance analysis circuit to perform a wideband sweep operation. The piezoelectric ceramic node applies multi-band voltage excitation to the contacting concrete surface and simultaneously acquires the returned current response to calculate the complex electromechanical admittance spectrum. For the specific circuit implementation of impedance analysis and equivalent circuit decoupling, those skilled in the art can use conventional AC bridge or self-balancing impedance measurement techniques, which are well-known in the field and will not be elaborated here.

[0037] Based on the equivalent circuit model principle of piezoelectric oscillators, the complex electromechanical admittance spectrum is used as the input parameter. The digital signal processor inside the piezoelectric ceramic node performs frequency domain decoupling processing on the complex electromechanical admittance spectrum, separating it into the conductance spectrum reflecting dielectric dissipation and the susceptance spectrum reflecting static capacitance properties.

[0038] Due to changes in ambient temperature, the dielectric constant of piezoelectric materials drifts. To address this, the piezoelectric ceramic node extracts the baseline slope drift of the susceptance in the susceptance spectrum. This drift is used as a temperature compensation component input to the current state vector of S201 and is used for time-axis pre-compensation of the local baseline waveform signal before calculating the current relative stretching factor in S202. Simultaneously, the piezoelectric ceramic node locates the main mechanical resonance peak in the conductivity spectrum and calculates its quality factor. The formula for the quality factor of the main mechanical resonance peak is as follows: ; In the formula, For the first Quality factor of the main mechanical resonance peak of each piezoelectric ceramic node; For the first The center resonant frequency of the main mechanical resonant peak of the piezoelectric ceramic node; For the first The half-power bandwidth of a piezoelectric ceramic node is the width of the frequency range where the amplitude drops to 1 / √2 of the peak amplitude or the power drops to 1 / 2 of the peak power.

[0039] To ensure the robustness of the calculation, if no obvious resonance peak is detected within the scanning frequency band, then... If the value approaches 0 or cannot be obtained, the edge scheduling module 100 directly marks the node as a hardware failure. The quality factor of the main mechanical resonance peak quantifies the electromechanical energy conversion efficiency between the piezoelectric ceramic node and the concrete surface layer, and its output is provided to the piezoelectric ceramic node for reliability calculation.

[0040] S103, Integration rule for bidirectional ultrasonic reciprocity testing and reliability of topological piezoelectric ceramic node pairs. Based on the reciprocity principle in linear acoustics, in a lossless isotropic elastic medium, the acoustic transmission response obtained by interchanged positions of the sound source and receiver should exhibit high consistency. This symmetry is disrupted when nodes undergo debonding degradation or changes in surface microstructure. Based on this physical causal relationship, the edge scheduling module 100 schedules adjacent piezoelectric ceramic node pairs within the target area to perform bidirectional ultrasonic reciprocity testing.

[0041] For the adjacent first The piezoelectric ceramic node and the first The piezoelectric ceramic node, the first The piezoelectric ceramic node emits low-frequency ultrasonic waves, the first... The first piezoelectric ceramic node receives and records the positive wake signal; after this, the transmitting and receiving objects interact, with the first... The piezoelectric ceramic node emits low-frequency ultrasonic waves, the first... Each piezoelectric ceramic node receives and records the reverse wake signal.

[0042] The edge scheduling module 100 calculates the bidirectional reciprocal residuals of the forward and reverse tailwave signals within the same time domain interval. The formula for the bidirectional reciprocal residuals is as follows: ; In the formula, The residuals are bidirectional reciprocal. It is a positive wake signal; It is a reverse tailwave signal; This serves as the starting point for integration within the wake time window; This is the endpoint of the integration within the wake time window; This is a very small positive number used to prevent a logic dead zone from occurring when environmental noise is too high and the effective signal approaches 0. In this embodiment, unless otherwise stated, the formulas use very small positive numbers to prevent division by zero or undefined logarithms. The value range is 10 -6 Up to 10 -5 .

[0043] The edge scheduling module 100 calculates the reliability of piezoelectric ceramic node pairs based on the piezoelectric ceramic node pair reliability fusion rule. The piezoelectric ceramic node pair reliability fusion rule completes the calculation by assigning mathematical weights to the quality factor drop rate of the main mechanical resonance peak and the bidirectional reciprocity residual. The formula for calculating the reliability of piezoelectric ceramic node pairs is as follows: ; In the formula, For the reliability of piezoelectric ceramic node pairs; For the first The quality factor of the initial master mechanical resonance peak of each piezoelectric ceramic node during the initial installation phase; For the first The current quality factor of the master mechanical resonance peak of each piezoelectric ceramic node; For the first The quality factor of the initial master mechanical resonance peak of each piezoelectric ceramic node during the initial installation phase; For the first The current quality factor of the master mechanical resonance peak of each piezoelectric ceramic node; The residuals are bidirectional reciprocal. , , Each of them has its own fixed weight coefficient, satisfying .

[0044] In practical applications, the specific values ​​of the aforementioned weighting coefficients can be determined based on regression analysis using historical calibration data to balance the weight of electromechanical and acoustic characteristics in reliability assessment. As a preferred approach, and The values ​​range from 0.25 to 0.35. The value range is from 0.30 to 0.50, and When independent calibration data is lacking, , , The values ​​can be 0.30, 0.30, and 0.40 respectively.

[0045] The edge scheduling module 100 determines whether the calculated reliability of the piezoelectric ceramic node pair is less than the reliability threshold. The reliability threshold is a quantitative boundary value used to determine whether the physical coupling state of the corresponding propagation path of the piezoelectric ceramic node pair has failed. It is usually set between 0.6 and 0.8, and its value is determined based on the lower limit of reliability fluctuation obtained by long-term monitoring of the system under a non-damaging baseline state.

[0046] When the reliability of a piezoelectric ceramic node pair is less than the reliability threshold, the edge scheduling module 100 performs a logical masking operation on that piezoelectric ceramic node pair. The logical masking operation is manifested in that the propagation path weight formed by the piezoelectric ceramic node pair is forcibly assigned to 0 in the prior probability field matrix generated by the subsequent module, in order to prevent abnormal data caused by node debonding or hardware degradation from participating in spatial difference calculation.

[0047] Reference Figure 3 S200, based on the common-mode stripping module 200, encodes kinematic parameters, thermal parameters, and nodal state parameters into state vectors and retrieves and extracts the virtual reference stretching factor. It then performs weighted spatial difference calculations based on the current relative stretching factor to construct a prior probability field matrix reflecting the damage distribution. Specifically, it includes the following sub-steps: S201, Multi-source spatiotemporal feature state vector encoding and historical event matching rules. In this embodiment, the common-mode stripping module 200 acquires and concatenates feature parameters from multiple environmental and operational condition dimensions to construct the current state vector. Addressing the objective physical limitation of inconsistent sampling rates in multi-source heterogeneous data, before encoding, the common-mode stripping module 200 performs linear interpolation and time alignment processing on the kinematic parameters of the gantry crane issued by the terminal programmable logic controller and the thermodynamic parameters obtained by drift inversion of the susceptance baseline of the surface temperature sensing unit or piezoelectric ceramic node, based on a unified time reference. This alignment operation aims to ensure the operational condition synchronization of various features within the same analysis time window.

[0048] After time alignment, the specific components of the current state vector include the gantry section, travel direction, surface temperature, heating / cooling phase, surface temperature change rate, susceptance baseline slope drift, and a binary vector of available piezoelectric ceramic nodes encoded using a fixed piezoelectric ceramic node index sequence. Specifically, the surface temperature change rate is obtained by dividing the surface temperature difference between two adjacent sampling times by the corresponding time interval; the heating / cooling phase is determined based on the surface temperature change rate and a temperature change rate dead zone threshold. The value range is from 0.1℃ / h to 0.5℃ / h.

[0049] Among them, the gantry crane section and direction of travel reflect the location and dynamic trend of the macroscopic mechanical load applied to the wharf surface; the surface temperature, heating and cooling phases, and surface temperature change rate characterize the periodic changes in the thermal environment caused by sunlight, nighttime heat dissipation, and the lag in heat conduction inside the concrete; the binary vector of effective piezoelectric ceramic node availability records the distribution of piezoelectric ceramic nodes in normal working condition in the current array. By normalizing and encoding the above multi-source spatiotemporal features, the common mode stripping module 200 maps the complex field conditions into a multi-dimensional vector space.

[0050] To eliminate common-mode interference from environmental fluctuations on acoustic characteristics, the common-mode stripping module 200 uses Mahalanobis distance to retrieve similar events from the historical baseline database. The historical baseline database pre-stores numerous historical state vectors and their corresponding baseline acoustic features collected during the initial installation phase and subsequent events determined by the collaborative reconstruction module 500 to be in a non-damaged state; events in a state of local material damage or high information entropy conflict are not written to the historical baseline database. Compared to conventional Euclidean distance, Mahalanobis distance effectively eliminates the influence of inconsistent dimensions of various features and the interrelationships between features, thus more accurately measuring the similarity of operating conditions. The formula for calculating Mahalanobis distance is as follows: ; In the formula, The current state vector and the first Mahalanobis distance between historical state vectors; This is the current state vector; For the first A vector of historical states; It is the inverse of the covariance matrix of the set of historical state vectors; This is the symbol for the matrix transpose operation.

[0051] In the actual calculation of the inverse of the covariance matrix, to avoid the covariance matrix from becoming rank-reduced and singular due to the long-term invariance of some working condition characteristics (such as the speed characteristic remaining at 0 due to the long-term stagnation of the gantry crane), the common mode stripping module 200 uniformly adds a regularization coefficient to the main diagonal elements of the covariance matrix. As a preferred method, The value range is 10 -6 Up to 10 -5 This approach ensures matrix invertibility while avoiding disruption of the statistical distribution characteristics of the original data.

[0052] The common-mode stripping module 200 retrieves the historical state vector that meets the matching threshold and has the smallest Mahalanobis distance as the best matching event. The matching threshold is preferably set to 2 to 5. After the state vector is normalized and encoded, the matching threshold is preferably set to 3. The common-mode stripping module 200 extracts the stretching factor pre-calculated under the same piezoelectric ceramic node pair, the same wake time window, and the same local baseline waveform signal definition as the current one, and uses it as a virtual reference stretching factor. When there is no historical state vector that meets the matching threshold, the stretching factor corresponding to the most recent historical event marked as a non-damaged state for the same piezoelectric ceramic node pair is used as the backtracking virtual reference stretching factor.

[0053] S202, Local optimization of the current relative stretching factor based on the cross-correlation function.

[0054] For each pair of valid piezoelectric ceramic nodes, the common-mode stripping module performs a time-axis stretching and compression optimization operation on the current wake signal acquired by the 200 pairs of common-mode stripping modules and the local baseline waveform signal acquired and stored under the healthy state of the corresponding piezoelectric ceramic node pair during the initial installation phase. Before performing the time-axis stretching and compression optimization operation, the local baseline waveform signal is pre-compensated based on the slope drift of the susceptance baseline extracted by S102. According to the propagation mechanism of sound waves in solid media, the generation of tiny cracks or changes in the stress field can cause slight changes in the propagation speed of sound waves in concrete media. This slight change in wave velocity, in a long-duration wake signal that has undergone multiple scatterings, macroscopically manifests as an overall stretching or compression of the time axis.

[0055] To accurately quantify the degree of deformation along this time axis, the common-mode stripping module 200 performs local optimization by calculating the cross-correlation function under different stretching ratios. The formula for calculating the cross-correlation coefficient is as follows: ; In the formula, The cross-correlation coefficient; The relative stretch factor of the time axis is used to characterize the proportion by which the waveform is stretched or compressed. This is the current wake signal; This is the local baseline waveform signal; and These are the integration start and end points of the wake time window, respectively; It is a very small positive number. (Set) The purpose is to prevent the algorithm dead zone from overflowing when the energy of the acquired signal is too low, causing the denominator of the integral to approach 0.

[0056] In practical engineering, the relative change in wave velocity caused by environmental temperature differences or early microscopic material degradation is extremely small. Therefore, the relative stretching factor... The optimization interval is generally set at [ Within the range of 0.01, 0.01, the optimization step size can be set to 10. -4 The common-mode stripping module 200 continuously changes within the aforementioned preset stretching range. The value of that makes the cross-correlation coefficient reach its maximum value. This is determined to be the current relative stretching factor. This parameter will be passed as a direct input to the subsequent spatial difference calculation stage.

[0057] S203, a rule for constructing the prior probability field matrix based on weighted spatial difference. As a preferred implementation, the common-mode stripping module 200 extracts the reliability of the piezoelectric ceramic node pairs generated in S103 and uses it as the core weight parameter for spatial difference mapping. The technical basis for this is that signal propagation paths with low reliability of piezoelectric ceramic node pairs often incorporate hardware degradation noise from the sensor itself. If directly involved in damage localization calculations, this would severely contaminate the accuracy of the inversion results.

[0058] Based on this, the common-mode stripping module 200 calculates the absolute difference between the current relative stretching factor and the corresponding virtual reference stretching factor, and uses piezoelectric ceramic nodes to physically weight this difference based on reliability. Subsequently, the one-dimensional path-weighted difference is smoothly diffused and mapped onto a two-dimensional mesh using a spatial distribution function. The formula for assigning values ​​to the two-dimensional mesh points of the prior probability field matrix is ​​as follows: ; In the formula, Let be the prior damage probability value of the grid point with planar coordinates (x, y) in the prior probability field matrix; The first generated for S103 The piezoelectric ceramic node and the first Reliability of piezoelectric ceramic node pairs between individual piezoelectric ceramic nodes; For the first The piezoelectric ceramic node and the first Current relative stretching factor between piezoelectric ceramic nodes; This corresponds to the virtual reference stretching factor; For the first The piezoelectric ceramic node and the first The propagation path, composed of piezoelectric ceramic nodes, has a spatial influence weight on the grid point with coordinates (x, y). The physical purpose of this formula is to transform the relative acoustic variability along the path into an initial probability estimate of damage at each location on the monitoring surface.

[0059] To clarify the quantization boundary of grid influence, spatial influence weights are used. The kernel is defined using an exponentially decaying kernel based on the first Fresnel zone. Its calculation formula is as follows: ; In the formula, The coordinates of a point (x, y) on a two-dimensional plane are given by the distance from the first point to the second point. The piezoelectric ceramic node and the first The vertical projection distance of the line segment connecting each piezoelectric ceramic node; Effective distribution scale parameters for controlling the attenuation range. The value of is usually set according to the center wavelength of the detected ultrasound, preferably 0.5 to 2 times the center wavelength of the detected ultrasound, to truly characterize the physical properties of sound wave energy radiating to both sides in space.

[0060] By introducing the aforementioned exponential decay kernel function, the grid cells farther from the signal propagation path receive smaller weight assignment values. The common-mode stripping module 200 performs overall normalization on the grid point values ​​calculated across all valid paths, ultimately outputting a priori probability field matrix covering the target region. This matrix effectively suppresses common-mode interference caused by changes in the thermal environment and load, while reducing the interference proportion of degraded nodes, providing a spatial penalty constraint basis for the sparse regularization equations of the subsequent constraint inversion module 300.

[0061] Reference Figure 4 S300, based on the constraint inversion module 300, injects coherent continuous high-frequency waves into the anomalous grid region indicated by the prior probability field matrix to extract the difference-frequency envelope time series. A sparse regularized equation is constructed by combining the influence line kernel function matrix and the prior probability field matrix, and solved to obtain the contact stiffness distribution at the crack interface. Specifically, it includes the following sub-steps: S301, Dual-frequency coherent injection and difference-frequency envelope extraction of local anomalous meshes. In this embodiment, the constraint inversion module 300 reads the prior probability field matrix output in S203. To accurately locate potential damage areas, the constraint inversion module 300 filters out anomalous mesh regions with probability values ​​greater than a set probability threshold from the prior probability field matrix. As a preferred approach, this probability threshold can be set between 0.65 and 0.85. Its specific value is obtained by statistically calibrating the false alarm rate based on historical data accumulated by the constraint inversion module 300 under long-term operation in a damage-free baseline state, with the aim of filtering out false high-probability points caused by conventional background noise or accidental interference.

[0062] For the selected anomalous mesh regions, the constraint inversion module 300 schedules the piezoelectric ceramic nodes located at the edge of the region to synchronously transmit two center frequencies into the internal concrete medium. and High-frequency ultrasonic continuous waves. Based on the frequency modulation mechanism in nonlinear acoustic theory, when these two high-frequency ultrasonic waves penetrate a closed microcrack, the nonlinear contact and friction behavior at the crack interface under the periodic action of sound pressure will induce frequency aliasing. This physical modulation effect will generate frequencies of [missing information] in the transmitted wave. The low-frequency difference frequency component. To ensure that the generated low-frequency difference frequency signal can effectively penetrate the local material attenuation region and is easy to filter and separate from the high-frequency carrier in subsequent signal processing, the frequency difference between the two input carriers is usually strictly controlled within the range of 1kHz to 15kHz.

[0063] After the receiving piezoelectric ceramic node acquires an aliased signal containing fundamental and harmonic frequencies, the constrained inversion module 300 performs local oscillator modulation and low-pass filtering on the aliased signal to filter out unwanted high-frequency carrier components. Further, the constrained inversion module 300 performs a Hilbert transform on the retained low-frequency signal after filtering. Specifically, the constrained inversion module 300 extracts a slowly varying difference-frequency envelope time series that directly reflects the periodic opening and closing effect of the microcrack interface by constructing the complex analytical form of the one-dimensional real signal and calculating its magnitude. This slowly varying difference-frequency envelope time series contains the dynamic characteristics of the crack contact state evolving over time, and is used as the observation vector for solving the subsequent inversion equations.

[0064] S302, the generation rule for the cross-scale influence line kernel function matrix from macro to micro. Within the same analysis time window for extracting the aforementioned observation vectors, the constraint inversion module 300 receives in real-time the instantaneous load and position information of the gantry crane wheel assembly from the terminal's programmable logic controller. The continuous movement of the gantry crane on the terminal surface will induce a dynamic stress field in the underlying concrete structure, which will be transmitted downwards and modulate the degree of closure of microcracks. To establish this mapping relationship from macroscopic surface load to microscopic internal acoustic response, the constraint inversion module 300 constructs a cross-scale influence line kernel function matrix based on the principles of elasticity.

[0065] In this embodiment, to achieve numerical solution by computer, the constraint inversion module 300 equates the instantaneous load of the gantry crane wheel assembly to a concentrated moving force acting on a semi-infinite elastic space surface. Based on the Buschnesk equation for force attenuation in elasticity, the module calculates the influence coefficient of the normal stress excited by the surface load within the discretized grid of the target region. The calculation formula is as follows: ; In the formula, The first in the kernel function matrix of the influence line The time sampling point for the first time sampling point Stress influence coefficient of each spatial grid point; For the first The instantaneous load magnitude of the gantry crane wheel assembly obtained at each time sampling point; For the first The known equivalent physical burial depth of each spatial grid point; and The first The center of the load at time and the first The formula represents the difference in the relative abscissa and relative ordinate of a spatial grid point on a horizontal projection plane; π is the constant of pi. The physical purpose of this formula is to quantify the local stress increment generated at a grid point at a specific depth underground when a large piece of ground-supported machinery moves.

[0066] By traversing all load sampling points within the current time window and all spatial points within the abnormal grid region, the constraint inversion module 300 generates a complete influence line kernel function matrix. This matrix characterizes the degree of influence of the surface load on the normal stress increment of each spatial grid point at any given time. Based on a preset interface contact response mapping relationship, the constraint inversion module 300 uses the normal stress increment as the excitation input and the crack interface contact stiffness state quantity of each spatial grid point as the state quantity to be estimated. It performs first-order linearization processing on the difference frequency envelope response, thereby constructing a linear mapping relationship between the observation vector Y and the target state vector K, and generating the corresponding forward modeling observation operator H.

[0067] S303, Sparse Regularized Constrained Inversion Formula for Fusion of Spatial Prior Probability Field Matrix. Based on the aforementioned acquired observation vectors and forward physical model, the constrained inversion module 300 constructs a sparse regularized equation to solve for the unknown contact stiffness distribution at the crack interface. Given that the limited amount of observation data on the surface is much smaller than the number of unknowns in the internal spatial grid, the acoustic inversion problem inherently possesses ill-posedness. Directly solving for the inverse of the matrix easily leads to numerical divergence; therefore, a regularization penalty term must be introduced to stabilize the computation process and constrain the solution space.

[0068] The constrained inversion module 300 uses the slowly varying difference frequency envelope time series extracted by S301 as the observation vector, the column vector of the crack interface contact stiffness distribution to be solved as the target state vector, the forward observation operator generated by S302 as the mapping matrix, and simultaneously constructs a spatial adaptive penalty matrix using the prior probability field matrix generated by S203; where... This represents the current target state vector to be estimated. The following is based on the forward observation operator. The calculated model predicts the observation vector. The specific general formula for solving the sparse regularization equation is defined as follows: ; In the formula, For observation vectors; Forward observation operator; The target state vector; This is a regularization parameter used to balance the weights between the data fitting residuals and the sparsity of the model solution; This is a diagonal penalty matrix; The L2 norm of a vector is used to calculate the squared error between measured data and model predictions. The first norm of a vector is used to ensure that the solution has sparse characteristics.

[0069] For diagonal penalty matrix The generation logic, constraint inversion module 300, transforms the mesh prior damage probability value output from S203 into penalty weights. The assignment formula for its main diagonal elements is as follows: ; In the formula, The first of the diagonal penalty matrix One main diagonal element; The first in the prior probability field matrix The prior damage probability value of each grid point; For the additional minimal positive number. Set the minimal positive number. The technical objective is to limit the upper limit of the penalty weight calculation, preventing algorithmic dead zones where the denominator of division equals zero when the prior damage probability of some grid points approaches or equals zero. Based on this allocation rule, the higher the prior damage probability value of a grid point in the early stage, the smaller its corresponding penalty weight value in the equation. This spatial adaptive adjustment mechanism guides the inversion algorithm to prioritize searching for non-zero stiffness anomalies in regions with higher prior probabilities, effectively reducing the solution space of unknowns and avoiding the algorithm from getting trapped in local optima.

[0070] In actual computation, the regularization parameter The value of is determined based on the L-curve curvature maximization criterion, and is usually around 10. -2 The threshold value is dynamically adjusted between 1 and 1 to ensure the numerical convergence of the equation under different signal-to-noise ratio conditions. For the specific numerical solution of this sparse regularization equation, those skilled in the art can use the conventional fast iterative shrinkage threshold algorithm, which is a well-known technique in the field and will not be elaborated here.

[0071] The contact stiffness distribution vector at the crack interface is obtained through iterative solution. Subsequently, the constraint inversion module 300 uses the crack interface contact stiffness distribution vector as a basis for... The model prediction data is calculated using the forward modeling operator, and the fitting error between the measured observation vector and the model prediction data is extracted as the inversion residual output. Simultaneously, the constraint inversion module 300 executes multi-dimensional anomaly judgment logic. The constraint inversion module 300 not only extracts isolated mesh elements with stiffness values ​​below the healthy baseline threshold (the healthy baseline threshold is the threshold corresponding to the baseline contact stiffness distribution pre-calibrated and stored after performing S301 to S303 on the target area in the healthy state during the initial installation stage), but also further performs spatial clustering analysis on the stiffness connectivity of its surrounding adjacent meshes. Only when the cumulative area of ​​continuous low-stiffness meshes exceeds the preset lower limit scale of physical damage does the constraint inversion module 300 finally output its corresponding two-dimensional mesh coordinates as stiffness anomaly coordinates; the lower limit scale of physical damage is preferably set to the area corresponding to 3 to 5 continuous mesh units. This judgment logic, which integrates thresholding and spatial clustering, avoids noise-induced misjudgments caused by relying solely on a single numerical extreme value, enabling the output coordinates to more reliably indicate the spatial location of actual internal microcracks. The abnormal coordinates are then transmitted to the thermal hysteresis verification module 400 and the collaborative reconstruction module 500 as reference data for engineering repair and structural safety assessment.

[0072] Reference Figure 5 S400, based on the thermal hysteresis verification module 400, retrieves gantry crane passage events where stiffness anomaly coordinates are at different thermal phases, extracts the difference frequency envelope peak value for load normalization, calculates the thermal hysteresis loop envelope area enclosed by the heating trajectory line and the cooling trajectory line, and determines the thermal response hysteresis attribute of the crack and the structural crack risk. Specifically, it includes the following sub-steps: S401, Cross-Thermal Phase Historical Monitoring Event Retrieval and Transient Load Normalization Processing. In this embodiment, the thermal hysteresis verification module 400 receives the stiffness anomaly coordinates output by the constrained inversion module 300 and retrieves historical monitoring events matching the current anomaly across the thermal phase in the monitoring history database. The monitoring history database is independent of the historical baseline database in S201; the former stores gantry crane passage monitoring events and their thermal state information for each time period, while the latter only stores non-damage state reference events. To ensure the physical comparability of multi-condition data, the thermal hysteresis verification module 400 executes event matching tolerance rules for vehicle speed, coordinates, and surface temperature ranges.

[0073] In practical operation, the thermal hysteresis verification module 400 filters historical gantry crane passage events where the distance deviation between the gantry crane's passage coordinates and the coordinates of the stiffness anomaly is within 0.5 meters, and the difference in gantry crane speed is within 10%. The filtered historical gantry crane passage events are then divided into temperature bins based on the surface temperature. Within the same temperature bin, heating phase events and cooling phase events are extracted separately to form a response comparison under the same surface temperature. A "same temperature bin" refers to a surface temperature difference of no more than 2℃, but can also be set according to the diurnal temperature range of the monitoring area, which is between 1℃ and 3℃.

[0074] This tolerance matching logic ensures that the acoustic responses extracted under different heating and cooling phases are subjected to consistent external mechanical load excitation, and avoids verification deviations caused by differences in gantry position, vehicle speed, and transient load.

[0075] After extracting the time series of the slow-varying differential frequency envelope corresponding to the aforementioned historical gantry crane passage events, the thermal hysteresis verification module 400 extracts the peak value of the differential frequency envelope in this time series. Since the actual counterweights of different batches of gantry cranes vary during operation, directly comparing the peak value of the differential frequency envelope would introduce load fluctuation errors. Therefore, the thermal hysteresis verification module 400 synchronously calls the instantaneous total load recorded by the programmable logic controller at the corresponding moment and performs division normalization processing on the peak value of the differential frequency envelope to output a normalized differential frequency response. The formula for calculating the normalized differential frequency response is as follows: ; In the formula, This is the normalized difference frequency response; The extracted difference frequency envelope peak value; This represents the instantaneous total load of the gantry crane at the corresponding moment; For the additional minimal positive number. Set the minimal positive number. The technical objective is to prevent the algorithm dead zone caused by a division denominator of 0 when the gantry crane is in an unloaded state and the system hardware error results in an instantaneous total load reading of 0. This normalization process eliminates the influence of the absolute magnitude of the macroscopic load on the acoustic nonlinear characteristics, laying a data foundation for subsequently reflecting the differences in the thermal response of the crack during the heating and cooling processes.

[0076] S402, Temperature Hysteresis Evolution Trajectory Fitting and Envelope Area Integration Calculation. In this embodiment, the thermal hysteresis verification module 400 correlates the normalized difference frequency response corresponding to each matching event with the surface thermal state parameters to generate a temperature hysteresis evolution trajectory. In specific implementation, the thermal hysteresis verification module 400 obtains the surface temperature of each gantry crane passage event by docking with the surface temperature sensing unit deployed on the surface of the wharf. The susceptance baseline slope drift extracted by the piezoelectric ceramic node in S102 is used as a node temperature compensation parameter to correct the local baseline waveform signal, and is not the sole source for the thermal hysteresis verification module 400 to obtain the surface temperature. Subsequently, the thermal hysteresis verification module 400 projects the discrete matching event data points into a two-dimensional coordinate system with the surface temperature as the abscissa and the normalized difference frequency response as the ordinate.

[0077] After projection, the thermal hysteresis verification module 400 divides the data points into heating and cooling sets based on the rate of change of the surface temperature over time. Data points corresponding to the temperature stable phase are not included in the calculation of the thermal hysteresis loop envelope area. The formula for determining the heating / cooling phase is as follows: ; In the formula, For thermal phase identification; Indicates the heating phase; Indicates the cooling phase; Indicates the temperature-steady phase; The surface temperature of the surface layer; The temperature change rate dead zone threshold is defined as 0.1℃ / h to 0.5℃ / h.

[0078] For curve fitting, those skilled in the art can use conventional least-squares polynomial smoothing techniques, preferably using first- or second-order polynomials. When there are fewer than three valid data points in the heating or cooling set, the thermal hysteresis verification module 400 temporarily does not output a structural crack risk label and waits for subsequent gantry crane passage events to supplement data. The thermal hysteresis verification module 400 performs curve fitting on the heating and cooling sets respectively to obtain heating trajectory lines and cooling trajectory lines. These two trajectory lines together close in a two-dimensional coordinate system to form a thermal hysteresis loop. Further, the thermal hysteresis verification module 400 calculates the envelope area of ​​the thermal hysteresis loop enclosed by the heating and cooling trajectory lines, using the following formula: ; In the formula, The area of ​​the thermal hysteresis loop envelope; and These are the lowest and highest surface temperatures of the surface layer, respectively, within the temperature range jointly covered by the heating and cooling trajectories during the thermal analysis cycle. The variable is the integral variable, representing the surface temperature of the surface layer; The heating trajectory line has a surface temperature of [temperature value missing]. The normalized difference frequency response function value at time; The cooling trajectory line has a surface temperature of [temperature value missing]. The normalized difference frequency response function value at time.

[0079] This formula quantifies the degree of asymmetric difference in the acoustic response of microcracks during the heating and cooling stages at the same surface temperature using definite integrals. The calculated thermal hysteresis loop envelope area will be used as a parameter for subsequent determination of the thermal response hysteresis attribute of cracks and the risk of structural cracks.

[0080] S403, a physical discrimination rule for structural cracks based on the difference between the thermal conduction hysteresis of concrete and the thermal expansion and contraction of cracks. Based on the obtained envelope area of ​​the thermal hysteresis loop, the thermal hysteresis verification module 400 executes the physical discrimination rule for structural cracks based on the difference between the thermal conduction hysteresis of concrete and the thermal expansion and contraction of cracks. At the physical mechanism level, the wharf surface layer is exposed to sunlight, wind cooling, and nighttime heat dissipation environment for a long time. The surface temperature of the concrete changes rapidly, while the internal temperature changes slowly, forming a non-steady-state temperature gradient along the depth direction. For shallow surface shrinkage cracks, the opening and closing of the cracks are mainly directly controlled by the surface temperature. The difference in normalized difference frequency response between the heating and cooling stages at the same surface temperature is small. For structural cracks extending into the surface layer, the crack tip and interface contact state are affected by the internal temperature hysteresis. The heating and cooling stages correspond to different degrees of interface closure at the same surface temperature, resulting in path dependence in the normalized difference frequency response, which is manifested as an increase in the envelope area of ​​the thermal hysteresis loop. Therefore, the envelope area of ​​the thermal hysteresis ring can be used to distinguish between shallow surface cracks and structural cracks with the risk of internal propagation.

[0081] Based on this physical causal relationship, the thermal hysteresis verification module 400 sets a threshold for the thermal hysteresis area of ​​structural cracks. As a preferred approach, this threshold is set as the average area of ​​the historical baseline thermal hysteresis loop envelope of the undamaged healthy area plus 3 to 6 times the standard deviation. The specific multiple is statistically calibrated based on the diurnal temperature range, surface layer thickness, and thermal conductivity of the concrete in the monitored area. To avoid potential misjudgments due to relying solely on the envelope area as a single parameter, the thermal hysteresis verification module 400 introduces a multi-indicator joint judgment logic. In addition to the area parameter, the module simultaneously calculates the maximum longitudinal difference between the heating trajectory and the cooling trajectory throughout the entire thermal analysis cycle, defining it as the extreme value of the thermal response difference. The formula for calculating the extreme value of the thermal response difference is as follows: ; In the formula, This represents the extreme value of the phase difference in thermal response; This is the normalized difference frequency response corresponding to the heating trajectory line; The normalized difference frequency response corresponds to the cooling trajectory line; the lower limit of the extreme value of the thermal response difference is preferably set as the mean of the extreme values ​​of the thermal response difference of the historical baseline of the undamaged healthy area plus 2 to 3 times the standard deviation.

[0082] In the final judgment stage, the thermal hysteresis verification module 400 determines that the crack corresponding to the current stiffness anomaly coordinate has a significant internal thermal response hysteresis only when the envelope area of ​​the thermal hysteresis loop is greater than the threshold of the thermal hysteresis area of ​​the structural crack, and the extreme value of the thermal response difference is greater than the preset lower limit of the extreme value. It then generates a structural crack risk label for this stiffness anomaly coordinate. If either of the above two conditions is not met, the thermal hysteresis verification module 400 generates a shallow surface crack label. The generated structural crack risk label, combined with the stiffness anomaly coordinate, provides a basis for subsequent assessment of crack propagation, bearing capacity degradation, and maintenance priority in the wharf surface layer.

[0083] Reference Figure 6 The S500, based on the collaborative reconstruction module 500, aggregates multi-dimensional observation indicators into a joint evidence vector to perform attribution arbitration, calculates the inferred information entropy of the logical conflict grid, and performs targeted supplementary calculations on the array transmit / receive topology for extreme information entropy regions. Specifically, it includes the following sub-steps: S501, Construction of Multidimensional Joint Evidence Vectors and Attribution Rules for Competitive Hypothesis. In this embodiment, the collaborative reconstruction module 500 receives multidimensional observation indicators output by the preceding modules. To ensure the consistency of multi-source physical data, the collaborative reconstruction module 500 uses the currently analyzed discrete spatial grid coordinates as a reference to perform spatiotemporal alignment operations on heterogeneous data. For the thermal hysteresis loop envelope area and structural crack risk label, the valid results generated when the corresponding spatial grid most recently completed thermal hysteresis verification are used.

[0084] After completing spatiotemporal alignment, the collaborative reconstruction module 500 concatenates five independently calculated physical characteristic parameters—quality factor drop, reciprocity residual, prior damage probability value, inversion residual, and the envelope area of ​​the thermal hysteresis loop—to construct a joint evidence vector for the current spatial grid. To achieve system-level diagnosis of the structural state, the collaborative reconstruction module 500 defines a set of competing hypotheses. This set of competing hypotheses specifically includes four mutually exclusive physical state inferences: node coupling degradation, environmental state transition, local material damage, and transient strong noise.

[0085] As one specific implementation, the collaborative reconstruction module 500 establishes a logical mapping rule between the joint evidence vector and the competing hypothesis set based on a preset expert physical mechanism. The specific attribution judgment logic is as follows: When the joint evidence vector shows that the quality factor drops and the reciprocity residual exceeds the set baseline, while the prior damage probability value and the thermal hysteresis loop envelope area remain within the healthy baseline range, the collaborative reconstruction module 500 attributes it to node coupling degradation, indicating that the sensor and concrete interface may debond.

[0086] When the prior damage probability value exceeds the benchmark threshold, the crack interface contact stiffness distribution has continuous low stiffness anomalies, and the inversion residual shows a low degree of fitting error, the collaborative reconstruction module 500 attributes it to local material damage. When the thermal hysteresis loop envelope area corresponding to the same spatial grid meets the structural crack risk discrimination condition, a structural crack risk label is further assigned.

[0087] When only the inversion residual in the joint evidence vector shows a pulse-like increase, and the other parameters do not undergo long-term shift, the collaborative reconstruction module 500 determines that it is affected by transient strong noise interference caused by external equipment.

[0088] When all indicators show a slow global shift and the reciprocal residuals do not show abrupt changes, the collaborative reconstruction module 500 attributes it to the migration of environmental conditions, which mainly stems from large-scale changes in the temperature or humidity field.

[0089] S502, Quantification of Information Entropy Uncertainty Based on Competitive Hypothesis Conflict. In actual monitoring operations, due to physical limitations in time synchronization and spatial resolution of multi-source observation data, the various indicators in the joint evidence vector often exhibit conflicting directions. To quantify the uncertainty of this system, the collaborative reconstruction module 500 combines the aforementioned mapping rules with the observation sample data to calculate the posterior probability distribution of various inference results in the competing hypothesis set under the current spatial grid.

[0090] The collaborative reconstruction module 500 establishes a Bayesian network containing evidence nodes and hypothetical state variables. The evidence nodes include quality factor drop, reciprocity residual, prior damage probability value, inversion residual, and thermal hysteresis loop envelope area. The hypothetical state variables include node coupling degradation, environmental state transition, local material damage, and transient strong noise.

[0091] This Bayesian network employs a two-layer directed structure. The state variable is assumed to reside in the first layer, and five evidence nodes are located in the second layer, each directed by the assumed state variable. The input values ​​of each evidence node are first normalized to the range of 0 to 1, and then categorized into three levels of evidence: low, medium, and high, based on the mean and standard deviation of the healthy baseline. The low level is defined as no greater than the healthy baseline mean plus two standard deviations; the medium level is defined as greater than the healthy baseline mean plus two standard deviations but no greater than the healthy baseline mean plus four standard deviations; and the high level is defined as greater than the healthy baseline mean plus four standard deviations.

[0092] The prior probabilities of each hypothetical state are obtained from the attribution sample database. The conditional probabilities of each level of evidence under different hypothetical states are obtained from the historical samples with hypothetical state labels in the attribution sample database. When the number of samples for a certain conditional probability is 0, Laplace smoothing is used, and the smoothing coefficient ranges from 0.001 to 1.

[0093] The attribution sample database includes non-destructive samples collected during the installation and calibration phase, artificially simulated node coupling degradation samples, environmental state migration samples collected under slowly changing temperature or humidity fields and verified by engineering to show no local material damage, local material damage samples confirmed by engineering verification, and transient strong noise samples collected under equipment interference conditions. Among them, non-destructive samples are used to determine the mean and standard deviation of the healthy baseline, while node coupling degradation samples, environmental state migration samples, local material damage samples, and transient strong noise samples correspond to four hypothetical state labels: node coupling degradation, environmental state migration, local material damage, and transient strong noise, respectively.

[0094] The collaborative reconstruction module 500 calculates the posterior probability distribution of each hypothetical state according to the following formula: ; In the formula, For the first Hypothetical state; For the first The evidence level of each evidence node; For the first The prior probability of a hypothetical state; For the first The evidence node at the _ The conditional probabilities under the four hypothetical states. The sum of the calculated posterior probabilities for the four hypothetical states is 1.

[0095] After obtaining the posterior probability distribution, the collaborative reconstruction module 500 uses information theory principles to calculate the inferred information entropy of the current spatial grid. This index reflects the degree of disorder in the system's internal state judgments, and its calculation formula is as follows: ; In the formula, For the first The inferred information entropy of a spatial grid, with the unit of bit; In this embodiment, the total number of hypothesis options in the competing hypothesis set is [number]. The value is 4; For the first The spatial grid that occurs the first The posterior probability values ​​of the competing inference hypotheses; For the additional minimal positive number.

[0096] As a preferred method, The value range is limited to 10. -6 Up to 10 -5 Set the smallest positive number. The technical purpose is to prevent the algorithmic dead zone caused by the undefined logarithmic function when the absolute value of the posterior probability of a certain type of inference hypothesis is 0.

[0097] From a mathematical and physical perspective, inference information entropy characterizes the degree of uncertainty in the attribution arbitration process. When the posterior probability of one inference hypothesis approaches 1 while the probabilities of the others approach 0, the inference information entropy... A value approaching 0 indicates extremely high certainty in the state inference; conversely, when the posterior probabilities of the four inference hypotheses tend to be equally distributed, the inference information entropy decreases. It reaches its maximum value. Therefore, the larger this value is, the more severe the logical conflict between the multidimensional observation indicators, indicating that the current spatial grid is in a region of high uncertainty.

[0098] S503, control logic and topology scheduling instructions driven by information entropy are issued. In this embodiment, the collaborative reconfiguration module 500 sets an information entropy judgment threshold, which is dynamically calibrated based on the average entropy value of historical conflict-free events, and is usually set to 1.5 to 2.0 times the average entropy value.

[0099] To avoid frequent hardware actions and system oscillations caused by single measurement errors, the collaborative reconfiguration module 500 introduces a time-window smoothing judgment logic. The collaborative reconfiguration module 500 not only evaluates the information entropy at the current moment but also statistically analyzes the entropy fluctuations over multiple consecutive analysis periods (e.g., three consecutive periods). Only when the information entropy of a spatial grid exceeds the information entropy judgment threshold for multiple consecutive periods is the collaborative reconfiguration module 500 extracted as a high-conflict grid. For the identified high-conflict grids, the collaborative reconfiguration module 500 generates topology scheduling instructions and sends them to the piezoelectric ceramic node array drive control unit via industrial Ethernet.

[0100] Furthermore, the core objective of this topology scheduling command is to prioritize the allocation of transmit / receive modes with distinguishable competitive assumptions to high-conflict grids. In practice, the topology scheduling command instructs piezoelectric ceramic nodes located at specific positions on the edge of the high-conflict grid to change their conventional omnidirectional divergent excitation mode. The piezoelectric ceramic node array drive control unit adjusts the transmission delay of each node according to the command parameters, switching to a directional acoustic beam focusing mode using the phased array interference principle.

[0101] The directional allocation of the acoustic beam energy can provide supplementary measurement data with higher signal-to-noise ratio and spatial resolution to high-conflict grid regions. The collaborative reconstruction module 500 uses the new round of observation data obtained from the supplementary measurement to refresh the joint evidence vector, thereby reducing the impact of the original evidence conflict on the state judgment, forming a complete control closed loop from software state inference to hardware behavior scheduling, and improving the fault tolerance and self-correction capability of the structural health monitoring system in complex service environments.

[0102] The present invention also provides a computer device, including: a processor and a memory, the memory storing a computer program executable by the processor, the computer program performing the method described above when executed by the processor.

[0103] The present invention also provides a storage medium storing a computer program, which is executed by a processor to perform the method described above.

[0104] The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0105] To further clarify and demonstrate the application process and technical effects of this invention in actual engineering conditions, a complete application example and comparative experimental data are presented, using a specific service scenario of an automated container terminal's front-end operation area.

[0106] Based on the aforementioned system architecture and methodological process, specific application embodiments of the present invention are as follows: In the gantry crane operation area of ​​a deep-water wharf, a 40m × 20m concrete surface layer was selected as the target area. Surface-mounted piezoelectric ceramic node arrays with a spacing of 2m were deployed within the target area, arranged in a 20×10 rectangular array, totaling 200 nodes.

[0107] When the programmable logic controller (PLC) predicts that a 1200-ton gantry crane will enter the area in 45 seconds, the edge scheduling module 100 issues a wake-up command. Within the no-load time window, the system completes wideband scanning and bidirectional ultrasonic reciprocity testing. The measured data shows that node 45 at coordinates (12m, 6m) was affected by the previous heavy-load impact, causing its main mechanical resonance peak quality factor to drop by 32% compared to the initial installation stage, and its bidirectional reciprocity residual with neighboring nodes exceeded the limit. Based on the piezoelectric ceramic node pair reliability fusion rule, the edge scheduling module 100 calculates that the reliability of the piezoelectric ceramic node pair along the relevant propagation path of this node is below the reliability threshold of 0.6. Therefore, it determines that the node's physical coupling has failed and performs logic masking to prevent degraded node data from participating in subsequent spatial difference calculations.

[0108] The gantry crane enters the target area. The common-mode stripping module 200 extracts the current surface temperature (32.4℃), heating phase, and other parameters and encodes them into a state vector. Using the Mahalanobis distance, it matches similar thermal conditions in the historical baseline database and extracts the virtual reference stretching factor. By performing weighted spatial difference calculations, the system maps the one-dimensional acoustic variability to a two-dimensional mesh.

[0109] Combination Figure 7 The contour map of the prior probability field matrix shown in section (a) depicts the physical dimensions of the target area on the horizontal and vertical axes. It can be seen that, influenced by the spatial exponential decay kernel function of the first Fresnel zone, after common-mode stripping calculations, a continuous high-value region with a prior damage probability higher than the surrounding area is formed within the coordinate range (14m–16m, 7m–9m). This region indicates the approximate extent of the potential damage, but the specific extension location of the crack is not yet clear.

[0110] against Figure 7 In the abnormal mesh region indicated by (a), the constraint inversion module 300 schedules the surrounding healthy nodes and synchronously injects continuous high-frequency waves of 150kHz and 155kHz. When the gantry crane wheel assembly passes over this region, the normal stress increment generated by the load of the gantry crane wheel assembly is transmitted downward, causing nonlinear contact and friction response at the closed microcrack interface. A 5kHz slow-varying difference frequency envelope time series is extracted from the transmitted wave.

[0111] The system call contains the influence line kernel function matrix of the Buschneiss stress attenuation law, and uses... Figure 7 The reciprocal of the prior probability value in (a) is used as the spatial adaptive penalty weight to perform sparse regularization equation iteration. Combined with... Figure 7The crack interface contact stiffness distribution map shown in section (b) has been further constrained after restricted inversion calculation, resulting in a wider high prior probability region. The map displays a band-shaped low-stiffness anomaly grid extending from 14.5 meters to 15.2 meters on the horizontal axis and 8.2 meters on the vertical axis. Combining the physical damage lower-limit scale clustering logic, the system outputs the stiffness anomaly coordinates at this location.

[0112] To distinguish whether the stiffness anomaly at this coordinate point is a shallow shrinkage crack or structural damage with a risk of penetrating downwards, the thermal hysteresis verification module 400 searches the historical database for gantry crane passage events at this coordinate point within the past 15 days that have a speed deviation of less than 10%. After normalizing the load and processing the difference frequency envelope peak value, the heating and cooling trajectories are fitted separately.

[0113] Combination Figure 8 The diagram shows the hysteresis evolution trajectory of the crack's thermal response. The horizontal axis represents the surface temperature of the surface layer, and the vertical axis represents the normalized difference frequency response. Solid dots represent heating phase events, and hollow squares represent cooling phase events. The fitted heating trajectory (lower solid line) and cooling trajectory (upper dashed line) do not overlap within the 25℃ to 35℃ range. Due to the hysteresis effect of internal heat conduction in concrete, a spindle-shaped thermal hysteresis loop is formed. The system calculates the envelope area of ​​this thermal hysteresis loop and combines it with the extreme value of the thermal response difference (0.40) for multi-indicator joint judgment. Since the envelope area of ​​the thermal hysteresis loop is greater than the threshold for the thermal hysteresis area of ​​structural cracks, and the extreme value of the thermal response difference is greater than the preset lower limit of the extreme value, the thermal hysteresis verification module determines that the crack corresponding to the stiffness anomaly coordinate has internal thermal response hysteresis characteristics and generates a structural crack risk label.

[0114] During a monitoring cycle, when the gantry crane was under heavy load and stopped suddenly, the joint evidence vector of the above-mentioned abnormal areas showed an index deviation: the prior damage probability value was too high, but the inversion residual showed abnormal pulse jumps.

[0115] Combination Figure 9 The bar chart showing the posterior probabilities of competing hypotheses and the inferred information entropy indicates that after the collaborative reconstruction module inputs the multidimensional indicators into the Bayesian network, the initial analysis period labeled T1 in the figure shows that the posterior probabilities of the two competing hypotheses, local material damage and transient strong noise, are close (42% and 45%, respectively), causing the inferred information entropy of this grid to rise to 1.85 bits.

[0116] After triggering the information entropy anti-oscillation smoothing judgment, the piezoelectric ceramic node switches to directional acoustic beam focusing mode by adjusting the emission delay to supplement the calculation of the high-collision grid. For example... Figure 9As shown in the post-reconstruction analysis cycle (T2), after acquiring new observation data and refreshing the evidence vector, the consistency of multi-source physical indicators improved, the posterior probability of local material damage increased to 89%, and the inferred information entropy decreased to below 0.6 bits. Under this closed-loop system, the impact of transient strong noise interference caused by sudden equipment shutdown on state judgment was reduced, and the affected area was attributed to local material damage.

[0117] To verify the actual effectiveness of this scheme, traditional ultrasonic wave velocity detection and conventional acoustic emission monitoring systems were simultaneously deployed in the aforementioned target area as control groups. A six-month comparative monitoring experiment was conducted, using artificial rock core samples and actual fracture evolution records detected by engineering geological radar as real references.

[0118] Combination Figure 10 The system's long-term operational performance comparison curves are shown, recording the trends of false alarm rate and structural crack false alarm rate for each method over 6 months. The false alarm rate is the proportion of events identified by the system as structural crack risks but not detected by manual review, relative to the total number of structural crack risk events output by the system. The false alarm rate is the proportion of events in the actual reference results that were confirmed to have structural cracks but were not identified by the system, relative to the total number of structural crack events in the actual reference results. The measured data confirms the following results: (1) Resistance to common-mode interference: During the seasonal transition from the 3rd to the 4th month, the diurnal temperature difference increases to over 15°C. The traditional ultrasonic wave velocity detection method is affected by wave velocity temperature drift, and the false alarm rate rises to 28%. The conventional acoustic emission monitoring system is affected by the thermal expansion and contraction of the gantry structure and friction noise, and the false alarm rate is about 22%. This invention reduces the impact of large-scale temperature field changes on damage judgment through similar working condition matching, common-mode stripping and thermal hysteresis verification. The false alarm rate remains below 3% during the experimental period.

[0119] (2) Early concealed damage identification capability: Traditional methods mainly rely on wave velocity changes or acoustic emission events caused by cracks. In the first two months, the control group had a false negative rate of over 40% for concealed micro-crack initiation points located 5 to 10 cm below the surface layer. This invention integrates the Buchnesk stress influence line kernel function generated by gantry motor loads through a sparse regularized equation and uses a low-frequency difference envelope to characterize the nonlinear contact response at the crack interface, reducing the false negative rate of structural cracks to about 5% and enabling the identification of degradation of internal contact stiffness in concrete.

[0120] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An adaptive monitoring and collaborative control system for cracks in the concrete surface layer of a wharf, characterized in that, include: The edge scheduling module is used to issue wake-up commands to the piezoelectric ceramic node array, and perform frequency domain scanning decoupling and bidirectional acoustic reciprocity test to generate the reliability of piezoelectric ceramic node pairs. The common-mode stripping module is used to obtain the current relative stretching factor and the virtual reference stretching factor, and combine the current relative stretching factor, the virtual reference stretching factor and the piezoelectric ceramic node to perform weighted spatial difference calculation on the reliability to construct a prior probability field matrix; The constrained inversion module is used to extract abnormal grid regions based on the prior probability field matrix, inject continuous high-frequency waves into the abnormal grid regions to extract the difference frequency envelope time series, construct a sparse regularized equation by combining the influence line kernel function matrix constructed based on the instantaneous load, solve the sparse regularized equation to obtain the contact stiffness distribution of the crack interface, and output the stiffness anomaly coordinates. The thermal hysteresis verification module is used to extract response features under different thermal phases based on the stiffness anomaly coordinates, generate heating trajectory lines and cooling trajectory lines, calculate the envelope area of ​​the thermal hysteresis loop enclosed by the heating trajectory lines and the cooling trajectory lines, and output structural crack risk labels. The collaborative reconstruction module is used to aggregate multi-dimensional observation indicators, including the envelope area of ​​the thermal hysteresis loop, into a joint evidence vector, perform attribution arbitration, and calculate the inferred information entropy. Based on the inferred information entropy, it issues a topology scheduling command to switch the transmit and receive modes of the piezoelectric ceramic node array.

2. The adaptive monitoring and collaborative control system for cracks in the concrete surface layer of a wharf according to claim 1, characterized in that, The edge scheduling module calculates the complex electromechanical admittance spectrum by applying voltage excitation to the piezoelectric ceramic node array and acquiring the current response, and extracts the quality factor of the main mechanical resonance peak from the conductance spectrum of the complex electromechanical admittance spectrum; the edge scheduling module schedules adjacent piezoelectric ceramic node pairs to transmit and receive low-frequency ultrasonic waves, and extracts the bidirectional reciprocity residuals of the positive and negative wake signals; the edge scheduling module calculates the reliability of the piezoelectric ceramic node pair based on the quality factor of the main mechanical resonance peak, the initial state calibration value of the quality factor of the main mechanical resonance peak, and the bidirectional reciprocity residuals.

3. The adaptive monitoring and collaborative control system for cracks in the concrete surface layer of a wharf according to claim 1, characterized in that, The common-mode stripping module constructs a current state vector, which includes the gantry section, travel direction, surface temperature, heating / cooling phase, surface temperature change rate, susceptance baseline slope drift, and a binary vector of effective piezoelectric ceramic node availability. The common-mode stripping module calculates the Mahalanobis distance between the current state vector and historical state vectors in the historical baseline database. The common-mode stripping module selects the stretching factor corresponding to the historical state vector that satisfies the matching threshold and has the smallest Mahalanobis distance as the virtual reference stretching factor.

4. The adaptive monitoring and collaborative control system for cracks in the concrete surface layer of a wharf according to claim 1, characterized in that, The common-mode stripping module calculates the cross-correlation coefficient between the current wake signal and the local baseline waveform signal under different time axis stretching ratios, and uses the stretching ratio corresponding to the maximum value of the cross-correlation coefficient as the current relative stretching factor. The common-mode stripping module calculates the difference between the current relative stretching factor and the virtual reference stretching factor, uses the piezoelectric ceramic nodes to weight the difference based on reliability, and maps the weighted difference onto a two-dimensional grid to construct the prior probability field matrix through an exponential decay kernel function based on the first Fresnel zone.

5. The adaptive monitoring and collaborative control system for cracks in the concrete surface layer of a wharf according to claim 1, characterized in that, The constrained inversion module selects grids with probability values ​​greater than the prior probability field matrix as anomalous grid regions by setting a probability threshold. The constrained inversion module schedules piezoelectric ceramic nodes at the edge of the anomalous grid regions to synchronously transmit two continuous high-frequency waves with different center frequencies, and collects aliased signals that penetrate the anomalous grid regions. The constrained inversion module performs low-pass filtering on the aliased signals to remove high-frequency carrier components, and performs Hilbert transform on the filtered signals to extract the envelope modulus to generate the difference frequency envelope time series, and uses the difference frequency envelope time series as the observation vector of the sparse regularization equation.

6. The adaptive monitoring and collaborative control system for cracks in the concrete surface layer of a wharf according to claim 1, characterized in that, The constrained inversion module calculates the influence coefficient of the instantaneous load of the gantry crane wheel group on the normal stress of each spatial grid point in the discretized grid of the target area based on the Buschnesk spatial force attenuation equation, combines them to generate the influence line kernel function matrix, and generates the forward modeling observation operator of the sparse regularization equation based on the influence line kernel function matrix. The constraint inversion module converts the prior damage probability values ​​of each grid point in the prior probability field matrix into penalty weights to construct a diagonal penalty matrix. The diagonal penalty matrix is ​​used as a regularization penalty term to solve the sparse regularization equation to obtain the target state vector, and the target state vector is used as the contact stiffness distribution of the crack interface.

7. The adaptive monitoring and collaborative control system for cracks in the concrete surface layer of a wharf according to claim 1, characterized in that, After obtaining the contact stiffness distribution at the crack interface, the constraint inversion module extracts mesh elements with stiffness values ​​lower than a preset healthy baseline threshold and performs spatial clustering analysis on the stiffness connectivity of adjacent meshes. When the cumulative area of ​​continuous low-stiffness meshes exceeds a preset physical damage lower limit scale, the constraint inversion module outputs the two-dimensional mesh coordinates corresponding to the continuous low-stiffness meshes as the stiffness anomaly coordinates.

8. The adaptive monitoring and collaborative control system for cracks in the concrete surface layer of a wharf according to claim 1, characterized in that, The thermal hysteresis verification module filters historical monitoring events where the distance deviation between the gantry crane's travel coordinates and the stiffness anomaly coordinates, as well as the difference in the gantry crane's travel speed, are within a preset range. It then divides the filtered historical monitoring events into temperature bins based on the surface temperature, extracting heating phase events and cooling phase events within the same temperature bin. The thermal hysteresis verification module divides the extracted difference frequency envelope peak value by the instantaneous total load at the corresponding moment to obtain the normalized difference frequency response as the response feature. The thermal hysteresis verification module uses a least-squares polynomial to fit the normalized difference frequency response under the heating and cooling phases to generate the heating trajectory and the cooling trajectory. When the calculated thermal hysteresis loop envelope area is greater than the structural crack thermal hysteresis area threshold and the maximum longitudinal difference between the heating trajectory and the cooling trajectory is greater than the lower limit of the extreme value, the thermal hysteresis verification module generates a structural crack risk label for the stiffness anomaly coordinates.

9. The adaptive monitoring and collaborative control system for cracks in the concrete surface layer of a wharf according to claim 1, characterized in that, The multi-dimensional observation indicators collected by the collaborative reconstruction module include the quality factor drop of the main mechanical resonance peak, bidirectional reciprocity residual, prior damage probability value, inversion residual, and the envelope area of ​​the thermodynamic hysteresis loop; the collaborative reconstruction module establishes a Bayesian network containing the hypothetical states and the evidence nodes corresponding to the multi-dimensional observation indicators, and calculates the posterior probability distribution of each hypothetical state using the conditional probability statistics of the attribution sample database; based on the posterior probability distribution of each hypothetical state, the collaborative reconstruction module calculates the inferred information entropy of the logical conflict grid according to the principle of information entropy.

10. The adaptive monitoring and collaborative control system for cracks in the concrete surface layer of a wharf according to claim 1, characterized in that, The collaborative reconstruction module statistically analyzes the fluctuation state of the inferred information entropy over multiple consecutive analysis periods, and extracts spatial grids whose inferred information entropy is greater than the judgment threshold over multiple consecutive analysis periods as high-conflict grids. The collaborative reconstruction module sends the topology scheduling command to the piezoelectric ceramic node array drive control unit. The topology scheduling command instructs the piezoelectric ceramic nodes at the edge of the high-conflict grid to adjust their transmission delay, switch to directional acoustic beam focusing mode using the phased array interference principle, perform supplementary calculations in the high-conflict grid region, and refresh the joint evidence vector.

Images