A method and sensor for stepwise tomographic imaging of structural defects

Abstract

This invention provides a stepwise tomographic imaging method and sensor for engineering structure defects, comprising: attaching a flexible sensor to the monitoring area of ​​the engineering structure; deploying 12 edge electrodes and 4 internal electrodes in the sensing layer; acquiring initial voltage data and establishing a reference voltage matrix; acquiring a mixed voltage matrix for the current measurement cycle according to adjacent edge excitation, edge-internal excitation, and internal inter-excitation, and calculating a differential voltage matrix with the reference voltage matrix; extracting the inter-edge differential voltage matrix from the differential voltage matrix; establishing a uniform coarse grid and an edge sensitivity matrix; performing full-field coarse imaging and determining the region of interest; refining the local grid within the region of interest; constructing an extended sensitivity matrix; calculating stable initial values ​​using the total differential voltage matrix; and obtaining local fine imaging results through correction iterations; and fusing the coarse imaging results with the local fine imaging results using distance weighting.

Description

Technical Field

[0001] This invention relates to the field of engineering structure health monitoring and intelligent sensor technology, and in particular to a step-by-step tomographic imaging method for engineering structure defects and an intelligent sensor. Background Technology

[0002] During long-term service, engineering structures are affected by load cycles, temperature and humidity changes, material aging, and water seepage erosion. Cracks, voids, debonding, water seepage channels, and localized material deterioration areas can easily form in concrete bridges, tunnel linings, underground utility tunnels, and building walls. In the early stages, these defects usually manifest as subtle changes in local electrical conductivity, contact state, or strain state. If they are not identified in time, they may lead to a decrease in load-bearing capacity or a deterioration in waterproof performance.

[0003] Existing detection methods include manual inspection, image recognition, ultrasonic testing, radar testing, single-point strain sensor monitoring, and resistive sensing monitoring. However, manual inspection and image recognition mainly rely on visible surface defects, making it difficult to detect hidden cavities, internal debonding, and early seepage channels in a timely manner. Ultrasonic testing and radar testing are mostly used for phased detection, making it difficult to establish long-term continuous monitoring. Single-point strain sensors or resistive sensors have limited monitoring ranges and cannot directly output spatial distribution images of the defect area. Therefore, this invention proposes a step-by-step tomographic imaging method for engineering structure defects and an intelligent sensor.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may contain prior art information that is not common knowledge to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by providing a step-by-step tomographic imaging method for engineering structural defects and an intelligent sensor, thereby solving the technical problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The first part of this invention provides a step-by-step tomographic imaging method for structural defects in engineering structures, comprising the following steps: S1. Attach a flexible sensor to the engineering structure monitoring area to record the effective imaging boundary of the sensing layer. Lay edge electrodes and internal electrodes on the sensing layer and connect the electrode switching unit, constant current excitation unit, voltage acquisition unit and data processing unit to collect initial voltage data and establish a reference voltage matrix. S2. The data processing unit collects the edge measurement data, edge and interior measurement data and interior measurement data of the current measurement cycle according to the adjacent edge excitation, edge and interior excitation and interior excitation, summarizes them into a hybrid voltage matrix, and calculates the differential voltage matrix with the reference voltage matrix; S3. Extract the inter-edge differential voltage matrix from the differential voltage matrix, establish a uniform coarse grid and edge sensitivity matrix, perform full-field coarse imaging, and determine the disease area of ​​interest based on the coarse imaging results; S4. Read the area of ​​interest in the disease, refine the local mesh to form a hybrid mesh, establish an extended sensitivity matrix, calculate the stable initial value using the total differential voltage matrix, and correct and iterate from the stable initial value to obtain the local fine imaging results. S5. Interpolate the coarse imaging results to the hybrid grid, and perform distance-weighted fusion with the local fine imaging results to obtain the final fused conductivity perturbation image. After filtering, enhancement and segmentation, output the disease tomography image and disease parameters.

[0007] S1 specifically includes: receiving the monitoring area record of the engineering structure, recording the monitoring area number, boundary coordinates, attachment direction and effective imaging boundary of the sensing layer; after cleaning the surface of the monitoring area, attaching a flexible sensor composed of an encapsulation layer, a sensing layer and a protective layer, and generating a sensor deployment record; deploying edge electrodes at the edge of the sensing layer and internal electrodes in the central area of ​​the sensing layer; connecting the electrodes to the electrode switching unit, constant current excitation unit, voltage acquisition unit and data processing unit, and generating an effective electrode status table through electrode numbering, wire continuity and electrode contact verification; after the flexible sensor stabilizes, executing the measurement cycle, acquiring initial voltage data, writing it into the reference voltage matrix according to the unified channel sequence number, and generating an effective measurement channel table.

[0008] S2 specifically includes: reading the valid measurement channel table, controlling the electrode switching unit to connect the excitation electrode pair according to the adjacent edge excitation sequence of the edge electrodes, injecting a constant micro current, scanning the measurement electrode pair according to the channel number, generating an edge excitation measurement record, retesting channels that are out of range or have abnormal fluctuations, and if they are still abnormal, deleting them from the valid data of the current measurement cycle, and simultaneously deleting the data corresponding to the same channel number in the reference voltage matrix. After completing the excitation of adjacent edges, the internal electrodes are reused as active excitation electrodes to perform edge-to-internal and inter-internal excitation, collect enhanced measurement data of the central region and mark the excitation type; the inter-edge measurement data, edge-to-internal measurement data and inter-internal measurement data are summarized into a hybrid voltage matrix according to a unified channel number, and the differential voltage matrix is ​​calculated after being aligned with the reference voltage matrix.

[0009] S3 specifically includes: extracting the inter-edge differential voltage matrix from the differential voltage matrix; generating a uniform coarse grid covering the monitoring area of ​​the engineering structure based on the effective imaging boundary of the sensing layer; assigning numbers to each coarse grid unit; and establishing an edge sensitivity matrix based on the actual coordinates of the edge electrodes, so that its rows correspond to edge-to-edge measurement channels and its columns correspond to coarse grid units; using the zero matrix as the initial coarse grid conductivity perturbation image, performing iterative reconstruction using the edge sensitivity matrix and the inter-edge differential voltage matrix to generate coarse imaging results; performing threshold segmentation and connected component analysis based on the conductivity perturbation amplitude of the coarse imaging results; deleting isolated candidate regions; and calculating the center coordinates and boundary coordinates of the retained candidate regions and expanding them to generate a data table of regions of interest.

[0010] S4 specifically includes: reading the data table of the disease region of interest, locally refining the coarse grid cells within the disease region of interest, retaining the coarse grid outside the disease region of interest to form a hybrid grid, and constructing an extended sensitivity matrix based on the actual placement of edge electrodes and internal electrodes, incorporating the internal electrodes as electrode subdomains with conductive properties into the modeling; reading the total differential voltage matrix and aligning the channels according to the row order of the extended sensitivity matrix, obtaining stable initial values ​​for local fine tomography through regularization; executing a modified Landweber iteration starting from the stable initial values, correcting the local fine conductivity perturbation image according to the adaptive step size, and outputting the local fine imaging results when the residual threshold or the upper limit of the number of iterations is met.

[0011] S5 specifically includes: reading coarse imaging results, the region of interest, the hybrid grid, and local fine imaging results; interpolating the coarse imaging results into the hybrid grid; setting fusion weights based on the position of grid points relative to the boundary of the region of interest; and generating the final fused conductivity perturbation image. The final fused conductivity perturbation image is sequentially subjected to median filtering, linear contrast stretching, and threshold segmentation to extract the disease connected components and generate a disease connected component data table. Based on the disease connected component data table, the disease location, disease area, disease length, disease width, and disease level are calculated, and the tomographic image of the engineering structure disease and disease parameters are output. It is then determined whether the quasi-steady-state condition is met. The quasi-steady-state condition is that the mean change rate of the edge-to-edge voltage is less than a preset change rate threshold within a continuous preset time, and no new disease connected components appear. When the quasi-steady-state condition is met, the average value of the mixed voltage matrix collected within the quasi-steady-state interval is updated to a new reference voltage matrix.

[0012] The second part of this invention provides an intelligent sensor, comprising: an encapsulation layer, a sensing layer, and a protective layer arranged sequentially from bottom to top; the encapsulation layer is a flexible insulating bearing layer attached to the surface of an engineering structure; The sensing layer is a resistance response layer formed by a composite of a flexible substrate material and a conductive filler. Multiple edge electrodes are arranged along the edges of the sensing layer, and multiple internal electrodes are arranged in the central region of the sensing layer. The protective layer covers the sensing layer. Both the edge electrodes and the internal electrodes are connected to an external electrode switching unit via wires.

[0013] The beneficial effects of this invention are as follows: This invention enhances the electric field coverage of the central region of the sensing layer by deploying 12 edge electrodes and 4 internal electrodes in the sensing layer and collecting edge-to-edge, edge-to-internal, and internal measurement data. This reduces the insensitivity of traditional edge electrode imaging to central defects and improves the accuracy of locating cracks, voids, and debonding areas. First, a coarse full-field imaging is performed using the differential voltage matrix between edges. Then, the region of interest is determined based on the coarse imaging results. This approach narrows the subsequent fine inversion range while ensuring full-field coverage, reducing the computational burden caused by direct inversion of the entire fine mesh.

[0014] This invention improves the clarity of disease boundaries and reduces false crack widening, edge fracturing, and local artifacts by refining the local mesh within the region of interest and combining this with an extended sensitivity matrix, stable initial values, and a modified Landweber iteration for localized finer tomography. By using distance-weighted fusion of coarse and fine local imaging results, the fine imaging results are retained within the region of interest, while the overall coarse imaging background is preserved outside the region, creating a continuous transition at the region boundaries. This reduces the risk of boundary jumps and missegmentation caused by directly stitching coarse and fine meshes.

[0015] This invention, through filtering, enhancement, thresholding, and connected component analysis of the final fused conductivity perturbation image, can output the location, area, length, width, and level of defects, transforming the monitoring results of engineering structure defects from single-point responses into spatialized, parameterized tomographic images. In long-term monitoring, updating the reference voltage matrix by combining the mean change rate of edge-to-edge voltages and the status of newly added defect connected components can reduce the long-term drift effects caused by temperature, humidity, contact impedance, and material aging, while preserving the traceability of the defect evolution process. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of a step-by-step tomographic imaging method for engineering structural defects according to the present invention. Detailed Implementation

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

[0018] Example 1: As Figure 1 As shown, this embodiment provides a step-by-step tomographic imaging method for structural defects in engineering structures, including the following steps: S1. Attach a flexible sensor to the engineering structure monitoring area to record the effective imaging boundary of the sensing layer. Arrange 12 edge electrodes and 4 internal electrodes in the sensing layer and connect the electrode switching unit, constant current excitation unit, voltage acquisition unit and data processing unit to collect initial voltage data and establish a reference voltage matrix. S2. The data processing unit collects the edge measurement data, edge and interior measurement data and interior measurement data of the current measurement cycle according to the adjacent edge excitation, edge and interior excitation and interior excitation, summarizes them into a hybrid voltage matrix, and calculates the differential voltage matrix with the reference voltage matrix; S3. Extract the inter-edge differential voltage matrix from the differential voltage matrix, establish a uniform coarse grid and edge sensitivity matrix, perform full-field coarse imaging, and determine the disease area of ​​interest based on the coarse imaging results; S4. Read the area of ​​interest in the disease, refine the local mesh to form a hybrid mesh, establish an extended sensitivity matrix, calculate the stable initial value using the total differential voltage matrix, and correct and iterate from the stable initial value to obtain the local fine imaging results. S5. Interpolate the coarse imaging results to the hybrid grid, and perform distance-weighted fusion with the local fine imaging results to obtain the final fused conductivity perturbation image. After filtering, enhancement and segmentation, output the disease tomography image and disease parameters.

[0019] S1 specifically includes the following sub-steps: S110. Determine the monitoring area of ​​the engineering structure and attach flexible sensors.

[0020] The data processing unit receives engineering structure monitoring area records, which are data sets used to define the attachment range of flexible sensors and subsequent tomographic imaging range. These records include at least the monitoring area number, engineering structure type, monitoring area boundary coordinates, sensor attachment direction, effective imaging boundary of the sensing layer, and installation time. The engineering structure monitoring area records are imported from field survey data, engineering structure design drawings, or manually calibrated data. The data processing unit uses the monitoring area boundary coordinates as a unified spatial reference for subsequent electrode coordinates, grid coordinates, and defect parameter coordinates.

[0021] After the monitoring area of ​​the engineering structure is determined, the construction personnel clean the surface of the monitoring area of ​​the engineering structure to remove dust, oil stains, loose particles and visible water film, so that the encapsulation layer can form a continuous bonding interface with the surface of the engineering structure. After cleaning, the data processing unit records the four corner coordinates and surface status indicators of the monitoring area of ​​the engineering structure. The surface status indicators are used to distinguish three types of status: "cleaned", "re-cleaned", and "adhesion prohibited".

[0022] When the surface status is marked as "cleaned", the data processing unit allows the generation of flexible sensor attachment records; when the surface status is marked as "re-cleaning required" or "attachment prohibited", the data processing unit does not generate flexible sensor attachment records to avoid subsequent data acquisition from unstable attachment interfaces.

[0023] The flexible sensor comprises, from bottom to top, an encapsulation layer, a sensing layer, and a protective layer. The encapsulation layer is a flexible insulating load-bearing layer that is directly bonded to the surface of the engineering structure. It is used to isolate the sensing layer from direct wear caused by roughness defects on the surface of the engineering structure and to provide a stable adhesion interface for the sensing layer. The sensing layer is a resistance response layer formed by a composite of flexible substrate material and conductive filler. It is used to convert changes in local conductivity, local tensile deformation, or local contact state caused by defects in the engineering structure into a measurable voltage response. A protective layer covers the sensing layer to block moisture, dust, and construction disturbances. In practice, the encapsulation layer thickness is set to 0.5 mm, the sensing layer thickness to 1 mm, and the protective layer thickness to 2 mm, ensuring that the flexible sensor, after attachment, balances adhesion stability, conductive response sensitivity, and external protection.

[0024] During flexible sensor attachment, the boundary of the encapsulation layer is aligned with the boundary of the monitoring area of ​​the engineering structure, and the effective imaging boundary of the sensing layer is located within the monitoring area of ​​the engineering structure. After attachment, the data processing unit generates a sensor deployment record, which includes at least the sensor number, the coordinates of the encapsulation layer boundary, the coordinates of the effective imaging boundary of the sensing layer, the coverage area of ​​the protective layer, the attachment time, and the curing completion time. If the sensor deployment record lacks the sensor number, the coordinates of the effective imaging boundary of the sensing layer, or the curing completion time, the data processing unit marks the sensor deployment record as invalid and prohibits it from entering the electrode deployment confirmation process of S120.

[0025] The sensor deployment record and effective imaging boundary of the sensing layer output by S110 are used by S120 to determine the electrode coordinates and the spatial position of the measurement link.

[0026] Specifically, the sensing layer is composed of a flexible substrate material (such as polydimethylsiloxane PDMS or waterborne polyurethane) and a conductive filler (such as multi-walled carbon nanotubes, nano-graphite powder or silver nanowires) in a mass ratio of 100:3 to 100:15.

[0027] During preparation, the conductive filler is added to the solvent and ultrasonically dispersed for 60 minutes. Then, it is thoroughly stirred with the flexible substrate and a curing agent is added. After degassing in a vacuum environment, a 1 mm thick sensing layer is formed by scraping or screen printing and cured at 60-80℃ for 2-4 hours.

[0028] S120. Edge electrodes and internal electrodes are deployed on the sensing layer and connected to the measurement link.

[0029] Twelve edge electrodes are arranged at the edge of the sensing layer. The edge electrode number is prefixed with E to indicate the edge electrode. The twelve edge electrodes are numbered E1 to E12 in a clockwise direction. Among them, E1, E2, and E3 are located at the upper edge of the sensing layer, E4, E5, and E6 are located at the right edge of the sensing layer, E7, E8, and E9 are located at the lower edge of the sensing layer, and E10, E11, and E12 are located at the left edge of the sensing layer.

[0030] Four internal electrodes are arranged in the central region of the sensing layer. The internal electrode number is prefixed with I to indicate that it is an internal electrode. The four internal electrodes are numbered I1 to I4 in 2 rows and 2 columns. The row spacing and column spacing of the internal electrodes are recorded in the electrode arrangement record. In an effective imaging area of ​​the sensing layer of 300mm×300mm, the row spacing and column spacing of the internal electrodes are set to 40mm to form an electric field path from the edge to the center in the central region of the sensing layer.

[0031] Each electrode uses a copper foil electrode, and the electrode size, electrode thickness, electrode center coordinates, and wire number are recorded in the electrode layout record. Each electrode is connected to an electrode switching unit via a wire. The electrode switching unit is connected to a constant current excitation unit and a voltage acquisition unit, which in turn is connected to a data processing unit.

[0032] The electrode switching unit is used to switch the conduction state between different excitation electrode pairs and measurement electrode pairs; the constant current excitation unit is used to inject a constant microcurrent into the selected excitation electrode pair; the voltage acquisition unit is used to acquire the differential voltage between the selected measurement electrode pairs; and the data processing unit is used to generate measurement channels, store voltage data, and perform tomographic calculations.

[0033] The magnitude of the constant microcurrent output by the constant current excitation unit is determined based on the initial impedance of the sensing layer, and in practice, the value ranges from 0.1mA to 5mA. For a sensing area of ​​300mm×300mm, it is preferable to inject a 1mA AC constant current signal or a DC constant current signal to ensure a sufficient signal-to-noise ratio while avoiding impedance drift caused by excessive Joule heating leading to local temperature rise in the sensing layer.

[0034] After the electrodes are deployed, the data processing unit performs electrode validity verification. Electrode validity verification includes electrode number verification, wire continuity verification, and electrode contact verification. Electrode number verification confirms that E1 to E12 and I1 to I4 each have a unique number; wire continuity verification confirms that a continuous conductive path is formed between each electrode and the electrode switching unit. Electrode contact verification confirms that the effective contact area between the electrode and the sensing layer is not less than a preset ratio, which is set to 80% during implementation. If any electrode is missing a number, has an open wire, or has an effective contact area less than 80%, the data processing unit marks the electrode status corresponding to that electrode as abnormal and blocks the measurement channel containing the abnormal electrode when generating the measurement channel. S120 outputs an electrode layout record and an effective electrode status table, which are used by S130 to generate an effective measurement channel table.

[0035] S130. Acquire initial voltage data and establish a reference voltage matrix.

[0036] After the flexible sensor is attached, cured, and the electrode wiring is completed, the data processing unit performs reference acquisition after the sensor has stabilized for a preset time. The sensor stabilization preset time refers to the waiting time required for the sensing layer, encapsulation layer, protective layer, and engineering structure surface to reach a stable electrical response after installation; in practice, this is set to 30 minutes. After 30 minutes, the data processing unit controls the electrode switching unit to execute at least three complete measurement cycles. Each complete measurement cycle includes voltage acquisition between edge electrodes, between edge electrodes and internal electrodes, and between internal electrodes.

[0037] For voltage values ​​continuously acquired from the same measurement channel, the data processing unit calculates the fluctuation ratio. If the fluctuation ratio does not exceed 2%, the average voltage value of the measurement channel is taken as the reference voltage value. If the fluctuation ratio exceeds 2%, the data processing unit acquires the voltage value of the measurement channel again. If the value still does not meet the requirements after the second acquisition, the measurement channel is marked as an invalid channel.

[0038] The data processing unit writes the reference voltage values ​​of each valid measurement channel into the reference voltage matrix according to a unified channel number. A measurement channel is a data unit jointly determined by the excitation electrode pair, the measurement electrode pair, the excitation current value, and the acquisition sequence. Its recording fields include at least the channel number, excitation positive electrode number, excitation negative electrode number, measurement positive electrode number, measurement negative electrode number, excitation current value, reference voltage value, acquisition time, and channel status.

[0039] For newly constructed test components, the reference voltage matrix corresponds to a defect-free state; for existing engineering structures, the reference voltage matrix corresponds to the existing structural state at the time of flexible sensor installation. Cracks, voids, or material inhomogeneities already present in this state are used as the reference background for subsequent monitoring and are not interpreted as new defects in this monitoring cycle. S130 outputs the reference voltage matrix and the effective measurement channel table, which together serve as the alignment basis for S230 to calculate the differential voltage matrix.

[0040] S2 specifically includes the following sub-steps: S210: Control the adjacent edge electrodes to perform constant current excitation and voltage scanning in sequence.

[0041] The data processing unit reads the valid measurement channel table output by S130 and controls the electrode switching unit to perform voltage scanning according to the adjacent edge excitation sequence. The adjacent edge excitation sequence includes E1-E2, E2-E3, E3-E4, E4-E5, E5-E6, E6-E7, E7-E8, E8-E9, E9-E10, E10-E11, E11-E12, and E12-E1, where each group of numbers represents a pair of adjacent edge electrodes. Taking E1-E2 as an example, the data processing unit controls the electrode switching unit to connect E1 to the positive terminal of the constant current excitation unit and E2 to the negative terminal of the constant current excitation unit, and controls the constant current excitation unit to inject a constant microcurrent into E1-E2.

[0042] In each group of adjacent edge excitation states, the voltage acquisition unit sequentially scans the measurement electrode pairs according to the valid measurement channel table, excluding the measurement items corresponding to the current excitation electrode pair. After each scan, the data processing unit generates an edge excitation measurement record, which includes at least the excitation electrode pair number, the measurement electrode pair number, the excitation current value, the current voltage value, the acquisition time, the number of samplings, and the channel status.

[0043] If the current voltage value exceeds the range of the voltage acquisition unit, or if the absolute value of the voltage difference between two consecutive samples of the same measurement channel exceeds 5% of the absolute value of the reference voltage for that measurement channel, the data processing unit marks that measurement channel as a channel to be retested. When the absolute value of the reference voltage is lower than the lower limit of the voltage acquisition unit's resolution, that lower limit is used as the comparison benchmark. If the channel to be retested still fails the retest, it is deleted from the valid data of the current measurement cycle, and the data corresponding to the same channel number in the reference voltage matrix is ​​deleted simultaneously, so that the current measurement cycle and the reference measurement cycle maintain consistency in channel dimension.

[0044] S220, reuse internal electrodes to perform enhanced measurements in the central region.

[0045] After completing 12 sets of adjacent edge excitations, the data processing unit reuses the internal electrodes as active excitation electrodes and performs central region enhancement measurements according to two sub-modes: edge-to-internal excitation and internal excitation. Edge-to-internal excitation involves injecting a constant microcurrent between an edge electrode and an internal electrode and measuring the differential voltage between other effective electrode pairs; internal excitation involves injecting a constant microcurrent between two internal electrodes and measuring the differential voltage between other effective electrode pairs. After the internal electrodes actively participate in the excitation, the electric field path passes through the central region of the sensing layer, causing conductivity disturbances caused by cracks, voids, or debonding in the central region to be incorporated into the voltage response data.

[0046] To avoid misidentifying localized high-gradient electric fields near the internal electrodes as defects, the data processing unit sets current constraints and channel markings for the internal electrode excitation data. The internal electrode excitation current is not higher than the excitation current of the adjacent edge, and each internal electrode excitation channel records the excitation type, internal electrode number, and measurement electrode number.

[0047] The data generated by the internal electrode excitation is not output as the final imaging result alone. Instead, it is written into the total differential voltage matrix along with the inter-edge measurement data and modeled as a known measurement channel when the extended sensitivity matrix is ​​constructed in S410. Thus, the fixed electric field perturbation caused by the internal electrodes is incorporated into the system matrix processing, rather than being processed by manually deleting image regions after imaging.

[0048] S230. Summarize the mixed voltage matrix and calculate the differential voltage matrix.

[0049] The data processing unit summarizes the inter-edge measurement data obtained in S210, the inter-edge and inter-interior measurement data obtained in S220, and the inter-interior measurement data into a mixed voltage matrix for the current measurement cycle according to a unified channel sequence number. The mixed voltage matrix refers to a data matrix composed of the voltage values ​​of all valid measurement channels within the same measurement cycle, and its row order is completely consistent with the valid measurement channel table in S130.

[0050] During implementation, a complete measurement cycle generates approximately 109 sets of valid measurement data, including inter-edge measurement data, inter-edge and inter-interior measurement data, and inter-interior measurement data. If there are invalid channels that have been deleted in the current measurement cycle, the data processing unit synchronously masks the same channel number in the reference voltage matrix, ensuring that the hybrid voltage matrix and the reference voltage matrix are consistent in terms of row number, channel order, and excitation current value.

[0051] The data processing unit calculates the differential voltage matrix according to the following formula: in, This is a differential voltage matrix, representing the voltage change in the current measurement cycle relative to the reference measurement cycle; This is a hybrid voltage matrix, representing the voltage values ​​of all valid measurement channels in the current measurement cycle; This is the reference voltage matrix, representing the reference voltage values ​​stored in S130 under the same channel number.

[0052] After the differential voltage matrix is ​​generated, the data processing unit splits it into an inter-edge differential voltage matrix and a total differential voltage matrix. The inter-edge differential voltage matrix consists of effective measurement channels where both the excitation and measurement electrodes are edge electrodes, and is used in S310 to establish the edge sensitivity matrix and perform full-field coarse imaging. The total differential voltage matrix consists of three types of effective measurement channels: edge-to-edge, edge-to-interior, and interior-to-interior, and is used in S420 to calculate stable initial values ​​for local fine tomography. This split does not change the channel numbers, but only the range of data read in subsequent calculations.

[0053] S3 specifically includes the following sub-steps: S310. Extract the differential voltage matrix between edges and establish the edge sensitivity matrix.

[0054] The data processing unit extracts the inter-edge differential voltage matrix from the differential voltage matrix obtained from S230. The inter-edge differential voltage matrix refers to the differential voltage matrix where both the excitation electrode and the measuring electrode belong to the edge electrodes, and it is used to quickly identify the approximate location of structural defects across the entire field. The data processing unit generates a uniform coarse grid covering the entire monitoring area based on the effective imaging boundary of the sensing layer, and assigns a unique cell number to each coarse grid cell.

[0055] During implementation, the number of coarse grid cells is set to 800 to 1200. In the effective imaging area of ​​the 300mm×300mm sensing layer, a stable balance can be achieved between computational load and overall positioning accuracy when the number of coarse grid cells is 1024.

[0056] The data processing unit establishes an edge sensitivity matrix based on a uniform coarse grid and the actual coordinates of 12 edge electrodes. Each row of the edge sensitivity matrix corresponds to an effective measurement channel in the inter-edge differential voltage matrix, and each column corresponds to a coarse grid cell. Each element in the matrix represents the degree of influence of a unit conductivity perturbation in the corresponding coarse grid cell on the voltage change of the corresponding edge-to-edge measurement channel.

[0057] After the edge sensitivity matrix is ​​established, the data processing unit binds and stores the edge sensitivity matrix, the inter-edge differential voltage matrix, and the coarse grid cell number table as input for the S320 coarse imaging iteration.

[0058] S320: A coarse imaging result is generated using an iterative reconstruction method.

[0059] The data processing unit uses the zero matrix as the initial coarse-grid conductivity perturbation image and performs iterative reconstruction based on the edge sensitivity matrix and the inter-edge differential voltage matrix to generate coarse imaging results. The zero matrix indicates that at the beginning of the iteration, no conductivity perturbation has occurred in any coarse-grid cell relative to the reference state; during the iteration, the data processing unit gradually corrects the conductivity perturbation value of each coarse-grid cell according to the residual between the calculated voltage and the measured differential voltage.

[0060] The coarse imaging process is performed according to the following formula: in, For the first The coarse-grid conductivity perturbation image obtained in the second iteration; This is the coarse-grid conductivity perturbation image obtained in the k-th iteration; k is the number of the coarse imaging iteration. This is the coarse imaging iteration step size; This is the edge sensitivity matrix; This is the differential voltage matrix between edges.

[0061] In implementation, the coarse imaging iteration step size is set to 0.01, and the number of iterations is set to 30. When the average change between two adjacent coarse grid conductivity perturbation images is less than a preset change threshold, the data processing unit stops the iteration early and uses the final output coarse grid conductivity perturbation image as the coarse imaging result. The coarse imaging result is not directly used as the final tomographic image output of the engineering structure defects, but only as the basis for S330 to determine the region of interest. Through this setting, the coarse imaging stage focuses on rapid full-field localization, avoiding direct high-density grid inversion across the entire field.

[0062] S330. Threshold segmentation is performed on the coarse imaging results to determine the disease area of ​​interest.

[0063] The data processing unit reads the coarse imaging results obtained from S320 and calculates the conductivity perturbation amplitude of each coarse grid cell in the coarse imaging results. Using 0.4 times the maximum perturbation amplitude of the coarse imaging results as the coarse segmentation threshold, the data processing unit marks coarse grid cells with conductivity perturbation amplitudes not lower than the coarse segmentation threshold as candidate disease cells and performs connected component analysis on adjacent candidate disease cells. Connected component analysis refers to the process of grouping candidate disease cells with adjacent boundaries or vertices into the same candidate region, used to organize discrete grid cells into spatially continuous suspected disease regions.

[0064] The data processing unit filters the candidate regions, deleting isolated candidate regions with an area less than 3 coarse grid cells and retaining candidate regions with an area of ​​not less than 3 coarse grid cells. For the retained candidate regions, the data processing unit calculates their geometric center, boundary coordinates, and circumscribed rectangle range, and expands the candidate region boundary by 1 to 2 coarse grid cells to generate the region of interest.

[0065] The region of interest (ROI) refers to the calculated range of suspected defects determined by coarse imaging results and used for local fine tomographic imaging. Its data fields include at least the region number, region center coordinates, region boundary coordinates, outer width, corresponding coarse grid cell number, and region status.

[0066] S330 outputs a data table of diseased areas of interest, which serves as the direct input for S410 to construct a hybrid mesh.

[0067] S4 specifically includes the following sub-steps: S410: Read the disease area of ​​interest and construct a hybrid grid and extended sensitivity matrix.

[0068] The data processing unit reads the data table of regions of interest (ROIs) output by S330 and performs local mesh refinement within each ROI. Local mesh refinement refers to the process of subdividing coarse mesh cells within the ROI into multiple fine mesh cells; in practice, each coarse mesh cell is subdivided into 4 sub-cells, and when the edge of the ROI is elongated, each coarse mesh cell is subdivided into 8 sub-cells.

[0069] The coarse grid cells outside the region of interest (ROI) retain their original scale, thus forming a hybrid grid where the ROI region has fine grid cells and the ROI region has coarse grid cells. The data processing unit assigns a unique number to each grid cell in the hybrid grid and establishes a mapping relationship between coarse and fine grid cells.

[0070] The data processing unit constructs an extended sensitivity matrix based on the actual placement of the 12 edge electrodes and 4 internal electrodes. When constructing the extended sensitivity matrix, the internal electrodes are not simplified to zero-size point sources; instead, electrode subdomains corresponding to the actual sizes of the internal electrodes are divided within the hybrid grid, and these subdomains are assigned the conductive properties of copper electrodes.

[0071] At the boundary between the electrode subdomain and the sensing layer subdomain, potential continuity and normal current continuity conditions are set to enable finite element calculations to reflect the equipotential body effect, flow-around effect, and convergence effect of the internal electrodes on the local electric field. Each row of the extended sensitivity matrix corresponds to an effective hybrid measurement channel in S230, and each column corresponds to a grid cell in the hybrid mesh. The matrix elements represent the degree of influence of a unit conductivity perturbation on the voltage change of the corresponding hybrid measurement channel in the corresponding grid cell.

[0072] After the extended sensitivity matrix is ​​established, the fixed electric field distortion caused by the internal electrodes is written into the extended sensitivity matrix as a known system characteristic. In the subsequent inversion process, the data processing unit uses this extended sensitivity matrix to solve for the conductivity perturbation corresponding to structural defects, rather than directly identifying the high-conductivity region around the internal electrodes as the defect area. S410 outputs a hybrid mesh, a hybrid mesh element number table, and the extended sensitivity matrix, which serve as input for the regularization solution in S420.

[0073] When constructing the extended sensitivity matrix, a complete electrode model (CEM) is used for the internal electrodes. This not only sets the potential continuity condition in the electrode subdomain, but also introduces contact impedance parameters at the interface between the electrode and the sensing layer. Its boundary conditions are expressed as: in Let be the potential of the l-th electrode, and u be the potential of the continuous domain of the sensing layer. This represents the interface normal current density. Contact impedance. It is estimated by subtracting the intrinsic impedance of the material from the total impedance of the electrode pair measured at the time of system installation.

[0074] S420, fuses mixed measurement data and calculates stable initial values ​​for localized refined tomography.

[0075] The data processing unit reads the total differential voltage matrix obtained in S230 and aligns the channels according to the row order of the extended sensitivity matrix in S410. The total differential voltage matrix includes the inter-edge differential voltage matrix, the edge-internal differential voltage matrix, and the internal-internal differential voltage matrix. If any channel in the total differential voltage matrix cannot find a corresponding row in the extended sensitivity matrix, the data processing unit deletes that channel; if there is a row in the extended sensitivity matrix that is not matched by the total differential voltage matrix, the data processing unit deletes that row, ensuring a one-to-one correspondence between the total differential voltage matrix and the extended sensitivity matrix.

[0076] The data processing unit uses regularization to obtain stable initial values ​​for localized refined tomography. Stable initial values ​​refer to the initial conductivity perturbation image obtained after fusing all effective mixed measurement data. Their function is to provide a starting image close to the actual disease distribution for the correction iteration of S430, reducing the impact of measurement noise, environmental drift, and contact impedance fluctuations on the direction of subsequent iterations.

[0077] The stable initial value is calculated according to the following formula: in, This serves as a stable initial value for localized refined tomography. To expand the sensitivity matrix; This is the transpose of the extended sensitivity matrix; For regularization parameters; It is a discrete Laplacian operator used to constrain the smoothness of conductivity perturbations between adjacent hybrid mesh elements; Let be the transpose of the discrete Laplace operator; This is the total differential voltage matrix.

[0078] The regularization parameter is determined by the L-curve method, and is set to 1.2 × 10^-5 in practice. The data processing unit stores the stable initial value, the regularization parameter, the residual initial value, and the mixed mesh element number table together. The stable initial value is not used as the final tomographic image output of the engineering structure defects, but only as the starting point for the S430 correction iteration.

[0079] S430: Starting from a stable initial value, perform correction iterations and output local fine imaging results.

[0080] The data processing unit uses the stable initial value obtained from S420 as the initial image to perform a modified Landweber iteration. The modified Landweber iteration refers to an inversion process that introduces an adaptive step size based on the basic Landweber iteration, used to recover high-frequency details of the lesion edges based on the stable initial value. In each iteration, the data processing unit first calculates the voltage residual corresponding to the current image, then calculates the Barzilai-Borwein step size based on the image changes and gradient changes between two adjacent iterations, and finally uses this step size to modify the current image.

[0081] The correction iteration is performed according to the following formula: in, For the first The local fine conductivity perturbation image obtained by the second correction iteration; This is the local fine conductivity perturbation image obtained in the j-th correction iteration; j is the correction iteration number. Let be the adaptive step size for the j-th correction iteration; To expand the sensitivity matrix; This is the transpose of the extended sensitivity matrix; This is the total differential voltage matrix.

[0082] The adaptive step size is calculated using the following formula: in, Let be the adaptive step size for the j-th correction iteration; For the j-th and the th The difference vector between images in each correction iteration; Difference vector The transpose of ; For the j-th and the th The difference vector between the gradients of each correction iteration; Difference vector The transpose of .

[0083] To avoid abnormal step size amplification caused by local noise, the data processing unit sets upper and lower limits for the adaptive step size. If the calculated adaptive step size is less than the lower limit, the lower limit is used as the step size for this iteration; if the calculated adaptive step size is greater than the upper limit, the upper limit is used as the step size for this iteration. The data processing unit stops iterating when the relative residual is less than 5 × 10^-4 or the number of correction iterations reaches 40.

[0084] The relative residual is calculated using the following formula: in, Let be the relative residual of the j-th correction iteration; Let L2 be the voltage residual L2 of the j-th correction iteration; The norm of the total differential voltage matrix; To expand the sensitivity matrix; This is the local fine conductivity perturbation image obtained from the j-th correction iteration; This is the total differential voltage matrix.

[0085] After the correction iteration is completed, the data processing unit outputs the local fine imaging result. The local fine imaging result includes at least the hybrid grid cell number, the local fine conductivity perturbation value of each grid cell, the final relative residual, the actual number of iterations, and the number of effective channels. This local fine imaging result serves as the input for S510 distance weight fusion.

[0086] S5 specifically includes the following sub-steps: S510. Perform distance-weighted fusion of coarse imaging results and local fine imaging results.

[0087] The data processing unit reads the coarse imaging result obtained in S320, the region of interest (ROI) obtained in S330, the mixed grid obtained in S410, and the local fine imaging result obtained in S430. Since the coarse imaging result is located on a uniform coarse grid, and the local fine imaging result is located on a mixed grid, the data processing unit first interpolates the coarse imaging result to the mixed grid. During interpolation, nearest-neighbor interpolation is used outside the ROI to maintain overall background stability; linear interpolation is used near the boundary of the ROI to avoid numerical abrupt changes at the boundary between coarse and fine grids.

[0088] The data processing unit sets fusion weights based on the position of the mixed grid points relative to the boundary of the region of interest (ROI). Grid points located inside the ROI use locally refined imaging results; grid points located outside the ROI and far from the boundary use coarse imaging interpolation results; and grid points located in the transition zone outside the ROI are smoothly fused using distance weights. The width of the transition zone is set to two coarse grid cell sizes; when the coarse grid cell size is 5mm, the transition zone width is set to 10mm to reduce closed-loop artifacts caused by direct splicing of coarse and fine grids.

[0089] The final fused conductivity perturbation image is calculated according to the following formula: in, represents the final merged conductivity perturbation value at grid point x; x is a grid point in the hybrid grid. The fusion weight at grid point x; This represents the local fine-scale imaging value at grid point x; This is the coarse imaging interpolation result at grid point x.

[0090] The fusion weights are calculated using the following formula: in, The fusion weight at grid point x; Let x be the distance from the boundary of the region of interest when grid point x is outside the region of interest, and the distance when grid point x is inside the region of interest. Set to 0; The internal fidelity boundary distance is set to 0 during implementation; The distance to the external transition boundary is taken as the size of two coarse mesh cells during implementation.

[0091] Through the distance-weighted fusion described above, the local fine imaging results output by S430 are retained inside the region of interest (ROI), while the coarse full-field imaging background output by S320 is retained outside the ROI, and a continuous transition zone is formed outside the ROI. The final fused conductivity perturbation image output by S510 is used as input for filtering, enhancement, and segmentation by S520.

[0092] S520: Filter, enhance, and segment the final fused conductivity perturbation image.

[0093] The data processing unit reads the final fused conductivity perturbation image output by the S510 and first performs 3×3 median filtering. 3×3 median filtering uses the 3×3 neighborhood formed by the current grid point and its adjacent grid points as the filtering window, and uses the median value within this window as the updated value for the current grid point. This is used to remove isolated noise points and preserve the abrupt changes in the edge of the fault. For grid points at the boundary of the mixed grid that have less than a 3×3 neighborhood, the data processing unit uses a boundary replication method to complete the filtering window, preventing boundary grid points from being incorrectly weakened.

[0094] After median filtering, the data processing unit performs linear contrast stretching on the filtered image. Linear contrast stretching refers to mapping the conductivity perturbation value of the effective grid cells to a normalized range of 0 to 1, making the grayscale difference between the diseased area and the background area easier for threshold segmentation.

[0095] Linear contrast stretching is performed according to the following formula: in, This represents the normalized conductivity perturbation value at grid point x; This represents the final fused conductivity perturbation value at grid point x; The minimum effective conductivity perturbation value in the final fused conductivity perturbation image; represents the maximum effective conductivity perturbation value in the final fused conductivity perturbation image; x represents a grid point in the hybrid grid. At that time, the data processing unit determines that there is no segmentable perturbation region in the current final fused conductivity perturbation image, and stops the extraction of the disease connected domain in this cycle.

[0096] The data processing unit then performs threshold segmentation, with the threshold being the mean of the normalized conductivity perturbation values ​​plus twice the standard deviation. Threshold segmentation is performed according to the following formula: in, The disease determination threshold used for threshold segmentation; The mean of the normalized conductivity perturbation values; This represents the standard deviation of the normalized conductivity perturbation value.

[0097] The data processing unit marks grid cells with normalized conductivity perturbation values ​​not lower than the disease determination threshold as diseased grid cells and performs connected component analysis on these diseased grid cells. Connected components with an area less than two diseased grid cells are identified as isolated noise regions and deleted; connected components with an area not less than two fine grid cells are retained as diseased connected components. The diseased connected component data table includes at least the connected component number, the number of the disease region of interest to which it belongs, the number of the grid cells containing it, the coordinates of the connected component boundary, the area of ​​the connected component, and the maximum normalized conductivity perturbation value.

[0098] S520 outputs a table of connected components of the disease, which is used as input for S530 to calculate the disease parameters.

[0099] S530 outputs tomographic images and parameters of structural defects, and performs a reference voltage update.

[0100] The data processing unit reads the defect connectivity data table output by S520 and calculates the defect parameters of the engineering structure based on the defect connectivity. The defect location is determined by the conductivity perturbation weighting center of the defect connectivity; the defect area is obtained by the sum of the surface areas of all grid cells within the defect connectivity; the defect length is determined by the projected length of the defect connectivity along the principal direction; the defect width is determined by the projected width of the defect connectivity perpendicular to the principal direction; the defect level is determined by the defect area, defect length, defect width, and the maximum normalized conductivity perturbation value. The principal direction of the defect connectivity refers to the direction with the largest coordinate variance of the grid cells within the connectivity, used to characterize the extension direction of slender defects such as cracks.

[0101] The location of the disease is calculated using the following formula: in, The conductivity perturbation weighting center of the diseased connected domain; q represents the number of grid cells within the disease-connected domain; q represents the grid cell number within the disease-connected domain. Normalized conductivity perturbation value The value at the center of the q-th grid cell; Let q be the center coordinates of the q-th grid cell.

[0102] The affected area is calculated using the following formula: in, Area of ​​disease; q represents the number of grid cells within the disease-connected domain; q represents the grid cell number within the disease-connected domain. Let q be the area of ​​the q-th grid cell.

[0103] For crack-type defects, the data processing unit calculates the crack length along the main direction of the defect's connected domain and the crack width along the direction perpendicular to the main direction; for defects such as cavities, debonding, or seepage channels, the data processing unit outputs the defect range based on the defect area, circumscribed contour, and maximum normalized conductivity perturbation value.

[0104] The data processing unit determines the disease level according to a preset disease level threshold table. The preset disease level threshold table includes at least a disease area threshold, a disease length threshold, a disease width threshold, and a maximum normalized conductivity perturbation value threshold, and classifies the disease level into three levels: low risk, medium risk, and high risk. When any one of the disease area, disease length, disease width, or maximum normalized conductivity perturbation value reaches the corresponding level threshold, the data processing unit marks the disease connected region as the corresponding level.

[0105] The data processing unit will ultimately fuse the conductivity perturbation image, the boundary of the disease connectivity domain, the location of the disease, the area of ​​the disease, the length of the disease, the width of the disease, and the level of the disease into a tomographic image of the engineering structure disease, and write the tomographic image of the engineering structure disease and the disease parameters into the historical database.

[0106] In long-term monitoring scenarios, the data processing unit updates the reference voltage based on the voltage stability and changes in fault connectivity over a continuous monitoring period. If the mean change rate of edge-to-edge voltage is less than 2% over 7 consecutive days, and no new fault connectivity appears in the fault connectivity data table output by S520, the data processing unit determines that the engineering structure is in a quasi-steady state and updates the average value of the mixed voltage matrix collected within this quasi-steady state interval to a new reference voltage matrix.

[0107] Before the update, the data processing unit writes the original reference voltage matrix, update time, tomographic image of the structural defects before the update, the updated reference voltage matrix, and corresponding defect parameters into the historical database. If new defect connectivity components appear within 7 consecutive days, or the mean change rate of edge-to-edge voltages is not less than 2%, the data processing unit keeps the original reference voltage matrix unchanged and continues to use the original reference voltage matrix in subsequent differential voltage matrix calculations. Thus, this method, while preserving the traceability of defect evolution, reduces the impact of long-term drift caused by temperature, humidity, and material aging on the identification of structural defects.

[0108] The determination of quasi-steady-state conditions also includes monitoring changes in ambient temperature and humidity. The data processing unit synchronously reads data from environmental sensors. When the ambient temperature fluctuation is within ±2℃ and the humidity fluctuation is within ±5% for 7 consecutive days, and the average voltage change rate is less than 2%, it is confirmed that the current impedance change is mainly caused by material creep or contact degradation rather than new defects. At this time, the reference voltage matrix is ​​updated.

[0109] Example 2: This example provides an intelligent sensor, including: an encapsulation layer, a sensing layer, and a protective layer arranged sequentially from bottom to top; the encapsulation layer is a flexible insulating bearing layer attached to the surface of an engineering structure; The sensing layer is a resistance response layer formed by a composite of a flexible substrate material and a conductive filler. Multiple edge electrodes are arranged along the edges of the sensing layer, and multiple internal electrodes are arranged in the central region of the sensing layer. The protective layer covers the sensing layer. Both the edge electrodes and the internal electrodes are connected to an external electrode switching unit via wires.

[0110] The flexible substrate material of the sensing layer is polydimethylsiloxane or waterborne polyurethane, and the conductive filler is multi-walled carbon nanotubes, nano-graphite powder or silver nanowires; there are 12 edge electrodes, which are evenly distributed around the perimeter of the sensing layer; there are 4 internal electrodes, which are distributed in a 2x2 matrix in the central area of ​​the sensing layer.

[0111] All the above formulas are performed using dimensionless numerical calculations; the relevant formulas are based on empirical models that approximate the real situation, obtained through extensive data collection and software simulation fitting. The preset parameters and thresholds involved in the formulas can be conventionally set and adjusted by those skilled in the art according to the physical constraints of the actual application scenario.

[0112] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0113] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0114] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A step-by-step tomographic imaging method for structural defects in engineering structures, characterized in that, Includes the following steps: S1. Attach a flexible sensor to the engineering structure monitoring area to record the effective imaging boundary of the sensing layer. Lay edge electrodes and internal electrodes on the sensing layer and connect the electrode switching unit, constant current excitation unit, voltage acquisition unit and data processing unit to collect initial voltage data and establish a reference voltage matrix. S2. The data processing unit collects the edge measurement data, edge and interior measurement data and interior measurement data of the current measurement cycle according to the adjacent edge excitation, edge and interior excitation and interior excitation, summarizes them into a hybrid voltage matrix, and calculates the differential voltage matrix with the reference voltage matrix; S3. Extract the inter-edge differential voltage matrix from the differential voltage matrix, establish a uniform coarse grid and edge sensitivity matrix, perform full-field coarse imaging, and determine the disease area of ​​interest based on the coarse imaging results; S4. Read the area of ​​interest in the disease, refine the local mesh to form a hybrid mesh, establish an extended sensitivity matrix, calculate the stable initial value using the total differential voltage matrix, and correct and iterate from the stable initial value to obtain the local fine imaging result.

2. The method for step-by-step tomographic imaging of structural defects according to claim 1, characterized in that, Also includes: S5. Interpolate the coarse imaging results to the hybrid grid, and perform distance-weighted fusion with the local fine imaging results to obtain the final fused conductivity perturbation image. After filtering, enhancement and segmentation, output the disease tomography image and disease parameters.

3. The method for step-by-step tomographic imaging of structural defects according to claim 1, characterized in that, S1 specifically includes: Receive the monitoring area record of the engineering structure, record the monitoring area number, boundary coordinates, attachment direction and effective imaging boundary of the sensing layer, clean the surface of the monitoring area and attach a flexible sensor composed of an encapsulation layer, sensing layer and protective layer, and generate a sensor deployment record; Edge electrodes are arranged at the edge of the sensing layer, and internal electrodes are arranged in the central area of ​​the sensing layer. The electrodes are connected to the electrode switching unit, constant current excitation unit, voltage acquisition unit and data processing unit. A valid electrode status table is generated by verifying electrode number, wire continuity and electrode contact. After the flexible sensor stabilizes, a measurement cycle is executed to collect initial voltage data, which is written into the reference voltage matrix according to a unified channel number, and a table of valid measurement channels is generated.

4. The method for step-by-step tomographic imaging of structural defects according to claim 1, characterized in that, S2 specifically includes: Read the valid measurement channel table, control the electrode switching unit to turn on the excitation electrode pair according to the adjacent edge excitation sequence of the edge electrodes, inject a constant micro current, and scan the measurement electrode pair according to the channel number to generate an edge excitation measurement record; For channels that exceed the range or have abnormal fluctuations, retest them. If they are still abnormal, delete them from the valid data of the current measurement cycle and simultaneously delete the data corresponding to the same channel number in the reference voltage matrix. After completing the excitation of adjacent edges, the internal electrodes are reused as active excitation electrodes to perform edge-to-internal and internal excitation, acquire enhanced measurement data of the central region and label the excitation type.

5. The method for step-by-step tomographic imaging of structural defects according to claim 4, characterized in that, Also includes: The measurement data between edges, between edges and interiors, and between interiors are aggregated into a hybrid voltage matrix according to a unified channel number. After aligning with the reference voltage matrix, the differential voltage matrix is ​​calculated.

6. The method for step-by-step tomographic imaging of structural defects according to claim 1, characterized in that, S3 specifically includes: The differential voltage matrix between edges is extracted from the differential voltage matrix. A uniform coarse grid covering the monitoring area of ​​the engineering structure is generated according to the effective imaging boundary of the sensing layer. Each coarse grid unit is assigned a number, and an edge sensitivity matrix is ​​established based on the actual coordinates of the edge electrodes, so that its rows correspond to edge-to-edge measurement channels and its columns correspond to coarse grid units. Using the zero matrix as the initial coarse grid conductivity perturbation image, iterative reconstruction is performed using the edge sensitivity matrix and the inter-edge differential voltage matrix to generate coarse imaging results; Based on the conductivity perturbation amplitude of the coarse imaging results, threshold segmentation and connected component analysis are performed to delete isolated candidate regions. The center coordinates and boundary coordinates of the remaining candidate regions are calculated and expanded to generate a data table of disease regions of interest.

7. The method for step-by-step tomographic imaging of structural defects according to claim 1, characterized in that, S4 specifically includes: Read the data table of the disease area of ​​interest, refine the local mesh of the coarse mesh cells within the disease area of ​​interest, retain the coarse mesh outside the disease area of ​​interest to form a hybrid mesh, and construct an extended sensitivity matrix based on the actual placement of the edge electrodes and internal electrodes, and include the internal electrodes as electrode subdomains with conductive properties in the modeling. Read the total differential voltage matrix and align the channels according to the row order of the extended sensitivity matrix. Then, obtain the stable initial value for local fine tomography by regularization. Starting with a stable initial value, perform a modified Landweber iteration, modify the local fine conductivity perturbation image according to the adaptive step size, and output the local fine imaging result when the residual threshold or the upper limit of the number of iterations is met.

8. The method for step-by-step tomographic imaging of structural defects according to claim 1, characterized in that, S5 specifically includes: Read the coarse imaging results, the region of interest, the mixed mesh, and the local fine imaging results. Interpolate the coarse imaging results into the mixed mesh and set the fusion weights according to the position of the mesh points relative to the boundary of the region of interest to generate the final fused conductivity perturbation image. The final fused conductivity perturbation image is sequentially subjected to median filtering, linear contrast stretching, and threshold segmentation to extract the disease connected components and generate a disease connected component data table. Based on the disease connected component data table, the disease location, disease area, disease length, disease width, and disease level are calculated, and the tomographic image of the engineering structure disease and disease parameters are output.

9. The method for step-by-step tomographic imaging of structural defects according to claim 8, characterized in that, Also includes: Determine whether the quasi-steady-state condition is met. The quasi-steady-state condition is: the rate of change of the average edge-to-edge voltage within a continuous preset time is less than a preset rate of change threshold, and no new faulty connected regions appear. When the quasi-steady-state condition is met, update the average value of the mixed voltage matrix collected within the quasi-steady-state interval to a new reference voltage matrix.

10. A smart sensor for use in the method of any one of claims 1 to 9, comprising: The encapsulation layer, sensing layer, and protective layer are arranged sequentially from bottom to top; The encapsulation layer is a flexible insulating load-bearing layer that is attached to the surface of the engineering structure; The sensing layer is a resistance response layer formed by a composite of a flexible substrate material and a conductive filler. Multiple edge electrodes are arranged along the edges of the sensing layer, and multiple internal electrodes are arranged in the central region of the sensing layer. The protective layer covers the sensing layer. Both the edge electrodes and the internal electrodes are connected to an external electrode switching unit via wires.

Images