Concrete damage classification method combining ae and dic technology

Abstract

The application describes a concrete damage grading discrimination method combining AE and DIC technologies, and relates to the field of civil engineering material testing technology. The method comprises the following steps: obtaining macro mechanical response data, surface image data and internal acoustic data synchronously collected for a concrete test piece; extracting stiffness degradation characteristics of each cyclic loading and unloading stage based on the macro mechanical response data, and constructing a macro damage parameter; extracting surface feature parameters based on the surface image data to construct a first microscopic damage parameter; extracting acoustic feature parameters based on the internal acoustic data to construct a second microscopic damage parameter; combining the macro damage parameter and the two microscopic damage parameters to optimize the evolution demarcation point and determine the damage evolution threshold; and dividing the damage grade based on the damage evolution threshold. The application realizes the multi-source data fusion of the macro bearing framework, the surface crack localization and the internal microscopic crack evolution, reduces the subjectivity of the artificially defined damage stage, and improves the reliability of the concrete damage evaluation.

Description

Technical Field

[0001] This invention belongs to the field of civil engineering material testing technology, and in particular to a method for classifying concrete damage using a combination of AE and DIC technologies. Background Technology

[0002] During long-term service, concrete materials experience the accumulation of microscopic defects and performance degradation. Tracking and describing the evolution of concrete materials from microcrack initiation and localization to macroscopic through-crack failure can advance the development of damage diagnosis and failure early warning for concrete structures.

[0003] Current mainstream methods for diagnosing material damage typically rely on monitoring a single physical quantity. For example, traditional methods mainly obtain macroscopic stress-strain or load-displacement curves using conventional mechanical testing machines, with the degree of degradation of overall stiffness as the sole evaluation index; or they deploy acoustic emission instruments separately to collect transient elastic wave signals generated by material fracture, and infer the internal degradation by statistically analyzing the number of acoustic impact events.

[0004] The technical system for classifying damage stages still needs improvement due to the reliance on human experience and the significant subjective interference. Summary of the Invention

[0005] The purpose of this invention is to provide a method for classifying concrete damage by combining AE and DIC technologies, in view of the above-mentioned problems in the prior art.

[0006] According to one aspect of this application, a method for classifying concrete damage using a combination of AE and DIC technologies includes:

[0007] Acquire macroscopic mechanical response data, surface image data, and internal acoustic data of concrete specimens simultaneously;

[0008] Based on macroscopic mechanical response data, stiffness degradation characteristics of each cyclic loading and unloading stage are extracted to construct macroscopic damage parameters;

[0009] Based on surface image data, surface feature parameters characterizing the degree of spatial localization evolution of surface cracks are extracted to construct the first microscopic damage parameter;

[0010] Based on internal acoustic data, acoustic feature parameters characterizing the degree of internal crack activity are extracted to construct a second microscopic damage parameter.

[0011] The evolution boundary point is optimized by combining macroscopic damage parameters, first microscopic damage parameters, and second microscopic damage parameters to determine the damage evolution threshold.

[0012] The damage level of concrete specimens is classified based on the damage evolution threshold.

[0013] According to another aspect of this application, macroscopic mechanical response data of concrete specimens acquired simultaneously are obtained, including:

[0014] Three-point bend cyclic loading and unloading tests were performed on concrete specimens; the loading phase of the three-point bend cyclic loading and unloading test was controlled by the crack opening displacement, and the unloading phase was controlled by the load.

[0015] Macroscopic mechanical response data were collected during the three-point bend cyclic loading and unloading test.

[0016] According to another aspect of this application, based on macroscopic mechanical response data, stiffness degradation characteristics of each cyclic loading and unloading stage are extracted to construct macroscopic damage parameters, including:

[0017] The initial stiffness, load value and crack opening displacement value corresponding to the unloading point in each cyclic loading and unloading stage, and residual load value and residual crack opening displacement value corresponding to the unloading end point are extracted from the macroscopic mechanical response data.

[0018] Based on the difference between the load value and the residual load value, and the difference between the crack opening displacement value and the residual crack opening displacement value, the cyclic stiffness of the corresponding cyclic loading and unloading stage is determined.

[0019] The stiffness degradation characteristics are determined based on the ratio between cyclic stiffness and initial stiffness, and these characteristics are used as macroscopic damage parameters.

[0020] According to another aspect of this application, based on surface image data, surface feature parameters characterizing the degree of spatial localization evolution of surface cracks are extracted to construct a first microscopic damage parameter, including:

[0021] Calculate the surface strain field of the corresponding test area based on surface image data;

[0022] Extract strain concentration regions characterizing local deformation from the surface strain field;

[0023] Based on preset image size parameters, the strain concentration area is converted into the actual physical area, and the current physical area corresponding to the current test time and the failure physical area corresponding to the failure time of the concrete specimen are obtained respectively.

[0024] Based on the ratio of the current physical area to the failed physical area, surface characteristic parameters are determined and used as the first microscopic damage parameter.

[0025] According to another aspect of this application, extracting strain concentration regions characterizing local deformation from a surface strain field includes:

[0026] The surface strain field is segmented using an image segmentation algorithm to obtain a preliminary binarized image containing concentrated deformation features.

[0027] By using connected component analysis, connected regions are filtered in the binarized image to remove noise regions with area features smaller than the preset filtering conditions, thus obtaining the denoised strain concentration region.

[0028] According to another aspect of this application, based on internal acoustic data, acoustic characteristic parameters characterizing the degree of internal crack activity are extracted to construct a second microscopic damage parameter, including:

[0029] The current cumulative number of events corresponding to the current test time of the internal acoustic data is counted, as well as the cumulative number of failure events corresponding to the time of failure of the concrete specimen;

[0030] Based on the ratio of the current cumulative number of events to the cumulative number of failure events, acoustic characteristic parameters are determined and used as the second microscopic damage parameter.

[0031] According to another aspect of this application, optimization of the evolutionary boundary point is performed by combining macroscopic damage parameters, a first microscopic damage parameter, and a second microscopic damage parameter, including:

[0032] Construct a multi-segmented piecewise regression model to represent the first correlation between the first microscopic damage parameter and the macroscopic damage parameter, and the second correlation between the second microscopic damage parameter and the macroscopic damage parameter, respectively;

[0033] Among them, the multi-segment piecewise regression model divides the data into multiple continuous linear intervals by setting at least two levels of dividing points.

[0034] According to another aspect of this application, the optimization of evolutionary boundary points is carried out, specifically including:

[0035] Within the range of values ​​for the macroscopic damage parameter, iterate through the combinations of candidate boundary points;

[0036] For each candidate combination, the corresponding parameter data is divided into multiple segments for linear fitting calculation to obtain the overall fitting residual evaluation index.

[0037] The combination of the boundary points corresponding to the minimum value of the fitting residual evaluation index is determined as the optimal mutation boundary point.

[0038] According to another aspect of this application, the determination of the damage evolution threshold includes:

[0039] The optimal mutation boundary points corresponding to the first microscopic damage parameter and the second microscopic damage parameter after optimization are extracted respectively. The optimal mutation boundary points include the first-level boundary point and the second-level boundary point.

[0040] Calculate the mean value of the first limit of the first-level boundary point corresponding to the two microscopic damage parameters, and the mean value of the second limit of the corresponding second-level boundary point;

[0041] The mean of the first and second limits are used as the thresholds for damage evolution.

[0042] According to another aspect of this application, the surface strain field of the corresponding test area is calculated based on surface image data, specifically including:

[0043] Using a preset image matching window, the degree of matching between subsets of surface image data before and after loading deformation is evaluated based on the cross-correlation coefficient algorithm, and the surface strain field is calculated.

[0044] Beneficial effects: This invention achieves multi-source data fusion of macroscopic load-bearing skeleton, localized surface cracks, and internal micro-crack evolution, reducing the subjectivity of manually defining damage stages and improving the reliability of concrete damage assessment. Attached Figure Description

[0045] Figure 1 A flowchart illustrating a concrete damage grading method combining AE and DIC technologies, provided as an embodiment of this application.

[0046] Figure 2 This is a flowchart illustrating an example of acquiring macroscopic mechanical response data of a concrete specimen simultaneously, as provided in an embodiment of this application.

[0047] Figure 3 An example provided in this application is based on macroscopic mechanical response data, which extracts stiffness degradation characteristics of each cyclic loading and unloading stage and constructs a flowchart of macroscopic damage parameters.

[0048] Figure 4 This application provides an example of a flowchart for extracting surface feature parameters that characterize the degree of spatial localization evolution of surface cracks based on surface image data, and constructing a first microscopic damage parameter.

[0049] Figure 5 This is a flowchart illustrating an example of extracting strain concentration regions characterizing local deformation from a surface strain field, as provided in an embodiment of this application.

[0050] Figure 6 This application provides an example of a flowchart for constructing a second microscopic damage parameter based on internal acoustic data, which extracts acoustic feature parameters characterizing the degree of internal crack activity.

[0051] Figure 7 This is a flowchart illustrating the optimization of the evolution boundary point using a combination of macroscopic damage parameters, a first microscopic damage parameter, and a second microscopic damage parameter, as provided in an embodiment of this application.

[0052] Figure 8 This is a flowchart illustrating an example of evolutionary boundary point optimization provided in an embodiment of this application.

[0053] Figure 9 A flowchart illustrating the determination of a damage evolution threshold, provided as an embodiment of this application. Detailed Implementation

[0054] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.

[0055] It should be noted that the terms "first," "second," etc., in the specification and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in sequences other than those illustrated or described herein. Furthermore, the terms "including" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0056] To overcome the aforementioned problems in the existing technology, the inventors, through literature review and technical analysis, verified that existing related solutions generally have the following shortcomings:

[0057] Among them, the existing schemes rely solely on macroscopic stiffness response, making it difficult to sensitively capture the early microscopic localized crack development of materials, while single acoustic statistical methods lack intuitive comparison of surface spatial deformation distribution.

[0058] Based on this, when actually defining the critical points of materials in different stages of decay, existing methods generally rely on the subjective manual interpretation of the trend of a single curve by testers, which is prone to introducing random human error and makes it difficult to maintain a unified evaluation standard.

[0059] In summary, existing methods struggle to simultaneously verify both macroscopic load-bearing degradation and microscopic localized development when dealing with material degradation mechanisms, and they lack rigid criteria for defining the boundaries of damage stages.

[0060] To overcome the shortcomings of existing technologies, referring to Figures 1-9 The implementation of this application is described in detail using the following embodiments.

[0061] To standardize the wording throughout the text and eliminate ambiguity in terminology, this embodiment adopts the following unified conventions for some technical terms:

[0062] AE stands for acoustic emission.

[0063] DIC stands for Digital Image Correlation.

[0064] Before conducting specific tests and evaluations, a comprehensive experimental platform is pre-built, including a three-point bending loading device, digital image acquisition hardware, and acoustic emission acquisition hardware; suitable for concrete specimens of different grades such as C30, C40, or C50.

[0065] During the testing and operation phase, cyclic loads were applied to the assembled specimens, and multi-source characteristics were simultaneously monitored using the integrated experimental platform.

[0066] To overcome the above shortcomings, combined with Figures 1 to 9 The present application will be further illustrated with reference to specific embodiments.

[0067] On the one hand, a method for classifying concrete damage using a combination of AE and DIC technologies is proposed, which specifically includes the following steps:

[0068] Step 101: Acquire macroscopic mechanical response data, surface image data, and internal acoustic data of the concrete specimens simultaneously.

[0069] Accordingly, a three-point bending loading device was used to record the time, load and displacement mapping relationship of the specimen during the stress process, forming macroscopic mechanical response data.

[0070] Furthermore, a pre-installed industrial camera is used to record the speckle displacement sequence on the specimen surface, forming surface image data. Simultaneously, an acoustic emission probe is used to capture transient acoustic voltage signals generated by the initiation and propagation of microcracks inside the specimen, forming internal acoustic data.

[0071] The above can be used to provide data input sources for damage status diagnosis.

[0072] For example, an industrial camera can be selected with a resolution of 5 megapixels and a frame rate of 2 frames per second.

[0073] Alternatively, a comprehensive selection can be made based on the accuracy requirements under actual working conditions.

[0074] Step 102: Based on the macroscopic mechanical response data, extract the stiffness degradation characteristics of each cyclic loading and unloading stage, and construct macroscopic damage parameters;

[0075] In conclusion, under cyclic loading, the microscopic defects inside the specimen will gradually accumulate and be reflected in the degradation of the macroscopic load-bearing frame.

[0076] Optionally, the data sequence that maps time, load and displacement is read, and the cyclic stiffness change trend corresponding to each loading and unloading action is extracted;

[0077] As the number of cyclic loading cycles increases, the cyclic stiffness of the specimen exhibits a monotonically decreasing evolution law;

[0078] The ratio of this attenuation value to the initial stiffness of the specimen in its initial undamaged state is calculated to obtain a dimensionless macroscopic damage parameter, which maps the physical weakening of the material's overall resistance to deformation and can serve as a unified evolution benchmark for multidimensional data comparison in the following text.

[0079] Step 103: Extract surface feature parameters that characterize the degree of spatial localization evolution of surface cracks based on surface image data, and construct the first microscopic damage parameter;

[0080] For the speckle displacement change sequence continuously acquired by industrial cameras, computer vision technology is used to extract pixel-level displacement deformation information.

[0081] Next, the surface strain contour map is calculated, from which localized deformation regions exhibiting high strain gradients are separated. As the external load continues to be applied, the microcracks in the material evolve from a dispersed random state to a locally connected state, and the actual physical area of ​​the aforementioned high strain gradient regions will show an expanding trend;

[0082] The ratio of the area of ​​the region at the current monitoring moment to the total area corresponding to the failure of the specimen is extracted. This ratio is used as a surface characteristic parameter to establish the first microscopic damage parameter, which is used to quantify the spatial localization development level of surface cracks.

[0083] Step 104: Extract acoustic feature parameters that characterize the degree of internal crack activity based on internal acoustic data, and construct the second microscopic damage parameter;

[0084] The captured transient acoustic voltage signal can be converted into discrete acoustic event count values ​​after being filtered by a preset decision time parameter.

[0085] Damage to the internal microstructure will release acoustic emission signals outward, and the total number of valid acoustic events accumulated within the current monitoring period will be statistically calculated;

[0086] The cumulative value is normalized to the total number of events when the specimen loses its load-bearing capacity, resulting in the second microscopic damage parameter, which can reflect the activity of crack initiation that is hidden inside the specimen.

[0087] Step 105: Combine macroscopic damage parameters, first microscopic damage parameters, and second microscopic damage parameters to optimize the evolution boundary point and determine the damage evolution threshold.

[0088] For example, macroscopic damage parameters can be used as horizontal evolution reference coordinates, and the first microscopic damage parameter and the second microscopic damage parameter can be used as vertical comparison parameters to establish a multi-dimensional evolutionary correlation mapping relationship.

[0089] Since the deterioration of concrete materials has a stage-based abrupt change characteristic, the algorithm search in this mapping relationship can be used to find the abrupt change inflection point that can minimize the overall linear fitting error, and this point is determined as the evolution boundary point.

[0090] By combining the mutation limits obtained through optimization of visual and acoustic features, the mean limit index after fusion is calculated and used as the threshold for judging the evolution of concrete damage.

[0091] Step 106: Classify the damage level of concrete specimens based on the damage evolution threshold;

[0092] In one specific embodiment, the damage level of concrete specimens can also be classified based on the comparison between macroscopic damage parameters and damage evolution thresholds.

[0093] Accordingly, after determining a comprehensive boundary scale, the evolution process of the measured microscopic parameters is dynamically compared under this scale. Based on the predetermined numerical range of the measured parameters, the physical and mechanical state of the test specimen is classified into progressively higher risk tiers, thus outputting specific damage level labels.

[0094] In other embodiments, the damage level corresponding to the concrete specimen can also be output as a condition assessment result, and a health monitoring or safety early warning strategy for the corresponding target concrete structure can be executed based on the condition assessment result.

[0095] For example, after completing the feature calibration and state classification at the laboratory level, the assessment conclusion based on damage grading is packaged and transmitted to the application terminal at the actual engineering site.

[0096] The main control equipment deployed at the target engineering site, such as bridges, dams, or roads, receives the result and uses it as a safety judgment benchmark for the corresponding load-bearing structure.

[0097] When on-site equipment collects data in real time based on similar sensor configurations, and the calculated damage parameters approach or cross the critical instability level range, a safety early warning strategy will be directly triggered. This could include automatically activating audible and visual alarm hardware or issuing load-limiting isolation commands to the dispatch system, thereby achieving proactive safety protection throughout the service life of the concrete structure. The above requires parameter adaptation based on the actual structural characteristics.

[0098] In some embodiments, acquiring macroscopic mechanical response data synchronously collected from concrete specimens may include:

[0099] Step 201: Perform a three-point bend cyclic loading and unloading test on the concrete specimen;

[0100] In the three-point bend cyclic loading and unloading test, the loading phase is controlled by the crack opening displacement, and the unloading phase is controlled by the load.

[0101] For mechanical testing of brittle materials, a single control mode for the testing machine can easily lead to testing failure. When a concrete specimen enters the softening stage under stress exceeding the peak load, internal cracks propagate rapidly. During this stage, the system sends displacement control commands to the servo motor based on the crack opening displacement to maintain the stable and slow development of cracks, preventing brittle fracture and instability of the specimen.

[0102] Once the preset target displacement is reached, the test enters the unloading phase.

[0103] At this point, the load control mode can be switched to gradually reduce the applied external force at a constant unloading rate. Using load control can return the specimen to the set baseline low load state at the end of each cycle, thereby extracting the irreversible residual deformation of the concrete after the current cycle.

[0104] Step 202: Collect macroscopic mechanical response data during the three-point bend cyclic loading and unloading test.

[0105] Alternatively, during the operating cycle of alternating dual control modes, the electrical signals of the sensor are recorded synchronously at high frequency; the sensor can specifically be a piezoelectric load sensor.

[0106] For example, the output voltage signal is converted into a real-time load value, while the deformation displacement recorded by the clamp extensometer is converted into a real-time crack opening displacement value.

[0107] It should be understood that the one-dimensional time sequence, which includes timestamps, continuous loads, and corresponding opening displacements, together constitutes the required macroscopic mechanical response data, which can be used for subsequent damage characteristic calculations.

[0108] On the other hand, based on macroscopic mechanical response data, stiffness degradation characteristics of each cyclic loading and unloading stage are extracted to construct macroscopic damage parameters, including:

[0109] Optionally, the initial stiffness, as well as the load value and crack opening displacement value corresponding to the unloading point in each cyclic loading and unloading stage, and the residual load value and residual crack opening displacement value corresponding to the unloading end point are extracted from the macroscopic mechanical response data.

[0110] Correspondingly, peaks and troughs are identified in continuous one-dimensional time series, thereby dividing the data stream into multiple independent loading and unloading cycles;

[0111] Extract the slope of the data segment that exhibits a linear elastic rise during the first loading cycle, and use it as the initial stiffness.

[0112] At the peak position of each subsequent cycle, identify the trigger point where the command switches from displacement control to load control, extract the value corresponding to that point, and use it as the unloading point load value and crack opening displacement value for the current cycle stage.

[0113] When the unloading process continues until the load reading drops to the set lower unloading limit benchmark, the data corresponding to the trough position is extracted as the residual load value and the residual crack opening displacement value.

[0114] Optionally, the cyclic stiffness of the corresponding cyclic loading and unloading stage is determined based on the difference between the load value and the residual load value, and the difference between the crack opening displacement value and the residual crack opening displacement value.

[0115] For example, the secant method combined with residual parameters can be used to calculate the mechanical response of the current cycle. That is, the effective rebound load is obtained by subtracting the residual load at the unloading endpoint from the load at the unloading peak.

[0116] The effective springback deformation is obtained by subtracting the residual displacement from the displacement of the unloading vertex.

[0117] The ratio of effective rebound load to effective rebound deformation is calculated using the secant method, and the output value is the cyclic stiffness of the current cycle. The corresponding calculation process follows the formula:

[0118] K _i =(P _i_u -P _i_r ) / (CMOD _i_u -CMOD _i_r );

[0119] In the formula, K _i Let P be the cyclic stiffness corresponding to the loading / unloading phase of the i-th loop. _i_u Let P be the load value corresponding to the unloading point in the i-th cycle phase. _i_r CMOD represents the residual load value corresponding to the end point of unloading in this stage. _i_u To be with P _i_u The corresponding crack opening displacement value, CMOD _i_r To be with P _i_r The corresponding residual crack opening displacement value.

[0120] As an optional solution, if the testing equipment does not have a dual control switching function and uses pure displacement control for unloading, a preset tolerance threshold close to zero can be set. The tolerance threshold can be determined according to the accuracy of the testing machine.

[0121] When the real-time load value falls within the tolerance threshold range, the displacement value at this time is actively captured to replace the standard residual crack opening displacement value for subsequent calculations.

[0122] Optionally, stiffness degradation characteristics are determined based on the ratio of cyclic stiffness to initial stiffness, and these characteristics are used as macroscopic damage parameters.

[0123] In this process, after obtaining the cyclic stiffness corrected by residual parameters at each cyclic stage, it is compared with a reference benchmark in which the specimen is in good condition to quantify the proportion of load-bearing capacity loss.

[0124] Next, the ratio of the current cyclic stiffness to the initial stiffness is calculated to obtain the proportion of residual stiffness of the material. Subtracting this proportion from 1 outputs a dimensionless value characterizing the current damaged state, thus completing the construction of the macroscopic damage parameters. The corresponding formula is as follows:

[0125] D _a =1-(K _i / K _0 );

[0126] In the formula, D _a K is a macroscopic damage parameter characterizing stiffness degradation. _i K is the cyclic stiffness of the loading / unloading phase in the i-th loop. _0 The initial stiffness is extracted beforehand. The calculated output D is... _a The value always ranges from 0 to 1. The gradual increase in the value reflects the process of macroscopic resistance degradation caused by the continuous accumulation of microcracks inside the specimen.

[0127] In a further embodiment, the specific visual processing mechanism for obtaining the first microscopic damage parameter is described in detail below:

[0128] In this process, speckle images are pre-prepared on the surface area of ​​the concrete specimen to be observed. These speckle images are used to provide a comparative pattern for the surface image data, characterizing the random texture features of the material surface.

[0129] For example, before starting the three-point bend cyclic loading and unloading test, the surface of the concrete specimen to be observed can be smoothed and random speckles can be created using a black and white spray painting process.

[0130] Specifically, the bottom layer can be sprayed with a white matte paint surface as a background, and the top layer can be sprayed with black paint dots as feature markers.

[0131] Controlling the spraying distance and duration ensures that the speckle particles have an appropriate size to meet the matching accuracy requirements of digital image correlation algorithms;

[0132] In this embodiment, the diameter of the speckle particles is approximately 3 to 5 pixels.

[0133] In this case, the prepared speckle pattern is attached to the surface of the specimen. Its randomly distributed texture undergoes spatial displacement synchronously with the surface of the specimen during the deformation process, thereby providing a pixel tracking basis with high contrast and high information entropy for digital image correlation algorithms.

[0134] In this embodiment, surface feature parameters characterizing the degree of spatial localization evolution of surface cracks are extracted based on surface image data to construct a first microscopic damage parameter, including:

[0135] Step 301: Calculate the surface strain field of the corresponding test area based on the surface image data, specifically including:

[0136] Using a preset image matching window, the degree of matching between subsets of surface image data before and after loading deformation is evaluated based on the cross-correlation coefficient algorithm, and the surface strain field is calculated.

[0137] After receiving continuous frames of surface image data, the initial undeformed image before loading is used as a reference, and the deformed image during the loading process is used as the target image. A subset of reference images of a predetermined size is set in the reference image, and a traversal search is performed in the target image using a preset step size window.

[0138] In this embodiment, the size of the reference image subset can be configured to 31×31 pixels, and the preset step size can be configured to 15×15 pixels.

[0139] Based on this, the cross-correlation coefficient algorithm can be used to calculate the gray-level matching index between the reference image subset and the target image candidate subset.

[0140] When the cross-correlation value reaches an extreme value, the subset of the target image is determined to be the matching region, and then the center pixel displacement bias of the two is extracted;

[0141] Gradient calculation is performed on the discrete displacement vector field over the entire domain, and the two-dimensional surface strain field matrix corresponding to the current stress stage is output.

[0142] Step 302: Extract strain concentration regions characterizing local deformation from the surface strain field, including:

[0143] For example, an image segmentation algorithm is used to segment the surface strain field to obtain a preliminary binarized image containing concentrated deformation features.

[0144] The two-dimensional surface strain field matrix exhibits a continuous gray-level gradient distribution, and it is necessary to extract the high-strain regions that represent crack initiation and propagation.

[0145] For example, a pre-configured image segmentation algorithm can be invoked to evaluate the matrix pixel by pixel. Specifically, an adaptive threshold segmentation algorithm can be used to perform statistical calculations on the global strain histogram to find the segmentation boundary value that maximizes the inter-class variance.

[0146] Pixels with strain values ​​higher than the segmentation threshold are designated as target foreground pixels and assigned a Boolean value of 1, while pixels with strain values ​​lower than the threshold are designated as background pixels and assigned a Boolean value of 0. The output matrix dimension is the same as the original. Figure 1 Binarized image.

[0147] In some alternative implementations, when the application scenario has the computing power to support it, the segmentation process can also be replaced by a lightweight deep learning segmentation model based on a fully convolutional network architecture.

[0148] In this scheme, well-known semantic segmentation networks such as U-Net can be used, with labeled strain field images as training samples for transfer learning. The two-dimensional surface strain field matrix is ​​input into the model, and the network directly outputs a binarized image through forward inference, thereby improving the stability of capturing high-strain regions under complex stray lighting interference.

[0149] For example, connected component analysis is used to filter connected regions in a binarized image, removing noise regions whose area features are smaller than preset filtering conditions, thus obtaining the denoised strain concentration region.

[0150] In the above text, the binarized image is often mixed with isolated noise caused by device thermal noise or local coating peeling.

[0151] Furthermore, based on the eight-neighborhood connectivity criterion, all foreground pixel sets in the binarized image are traversed to extract interconnected independent pixel clusters.

[0152] Calculate the number of pixels contained in each pixel cluster. Set an area threshold parameter, mark pixel clusters with fewer pixels than the area threshold parameter as environmental noise, and correct their corresponding Boolean values ​​to 0;

[0153] The remaining pixel clusters are retained and integrated into the output as the noise-reduced strain concentration region.

[0154] The area threshold parameter can be determined by statistically analyzing the typical area distribution of background noise in the unloaded state, based on the image resolution and speckle particle size.

[0155] Step 303: Based on the preset image size parameters, the strain concentration area is converted into the actual physical area, and the current physical area corresponding to the current test time and the failure physical area corresponding to the failure time of the concrete specimen are obtained respectively.

[0156] Optionally, the calibration height data in the preset parameters can be extracted. Further, the ratio of the measured physical height of the calibration reference object to the total number of vertical pixels occupied in the image coordinate system can be calculated to obtain the actual physical side length corresponding to a single pixel.

[0157] The actual physical side length is squared to output the actual physical area corresponding to a single pixel.

[0158] Traverse the strain concentration region matrix at the current test time and count the total number of valid pixels retained in it.

[0159] The current physical area is calculated by multiplying the total number of effective pixels by the actual physical area corresponding to a single pixel.

[0160] Continue this recording process, and when the specimen loses its load-bearing capacity and loading stops, execute the same calculation process to obtain the failed physical area.

[0161] Step 304: Based on the ratio of the current physical area to the failed physical area, determine the surface characteristic parameters and use the surface characteristic parameters as the first microscopic damage parameter.

[0162] The spatial localization of surface cracks expands as material damage intensifies. Therefore, the current physical area is extracted as the numerator parameter, and the failed physical area is extracted as the denominator parameter, and a division operation is performed.

[0163] The calculated dimensionless ratio is used as a surface characteristic parameter and directly defined as the first microscopic damage parameter to characterize the damage evolution process of the material in the two-dimensional observation plane. That is:

[0164] D _DIC =A _i / A _max ;

[0165] In the formula, D _DIC As the first microscopic damage parameter, A _i Let A be the current physical area corresponding to the i-th test time. _max This represents the physical area of ​​failure when the specimen loses its load-bearing capacity at the end of the entire loading cycle.

[0166] In another embodiment, the acoustic event identification mechanism and the calculation of the second microscopic damage parameter are further explained.

[0167] Among them, the acoustic events associated with the current cumulative number of events and the cumulative number of failure events in the internal acoustic data can be jointly identified and obtained through preset peak determination time parameters, impact determination time parameters, and impact locking time parameters.

[0168] As an example, the hardware operating parameters of the acoustic emission acquisition system can be configured during a three-point bend cyclic loading and unloading test run to shield against environmental noise and extract valid signals.

[0169] In one possible scenario, the sampling frequency can be set to 1MHz, and the signal acquisition threshold can be set to 45dB.

[0170] Alternatively, the acquisition threshold can be determined by technicians based on the background noise level of the test environment, using conventional methods such as pencil-breaking simulation source calibration.

[0171] Furthermore, for the original transient voltage oscillation waveform that exceeds this threshold, the joint identification mechanism is invoked to define the event.

[0172] In one possible scenario, the peak determination time (PDT) parameter is set to 50 μs, which is used to identify the true peak amplitude of the acoustic impact after the waveform crosses the threshold.

[0173] The impact determination time (HDT) parameter is set to 150 μs, which is used to determine the end point of a single acoustic impact event during the descent of the signal envelope.

[0174] The impact lockout time (HLT) parameter was set to 300 μs, which limits the silent window time for the system to stop signal acquisition response after confirming a single impact, in order to filter out false signal interference caused by the reflection of the specimen boundary.

[0175] Based on this, continuous time-domain voltage waveform data can be extracted and encapsulated into independent discrete acoustic events.

[0176] In this embodiment, acoustic feature parameters characterizing the degree of internal crack activity are extracted based on internal acoustic data to construct a second microscopic damage parameter, which can specifically be:

[0177] Step 401: Calculate the current cumulative number of events corresponding to the internal acoustic data at the current test time, and the cumulative number of failure events corresponding to the time of failure of the concrete specimen;

[0178] With the cyclic input of external loads, microscopic defects within the material continuously excite discrete acoustic events during friction and propagation. A time-axis accumulator can be built in memory to monotonically increment the count of valid acoustic events output according to the timestamp sequence.

[0179] When the target is being monitored, extract the current count value recorded by the accumulator and define it as the current cumulative event count.

[0180] When the macroscopic bearing capacity curve is detected to fall to the preset lower limit of failure, that is, when the concrete specimen is determined to have reached the mechanical failure point, the status update of the accumulator is frozen, the final count value at this moment is read and retained, and it is defined as the cumulative failure event number.

[0181] Step 402: Determine the acoustic characteristic parameters based on the ratio of the current cumulative number of events to the cumulative number of failure events, and use the acoustic characteristic parameters as the second microscopic damage parameter.

[0182] Among them, the cumulative total number of discrete acoustic events is nonlinearly positively correlated with the volume expansion increment of internal cracks in the material.

[0183] Furthermore, the current cumulative number of events is obtained as the numerator parameter, and the cumulative number of failed events is obtained as the denominator parameter, and an algebraic division operation is performed.

[0184] The calculated dimensionless characteristic value reflects the relative proportion of the activity of invisible internal cracks throughout the entire service life; this characteristic value can be directly output as the second microscopic damage parameter.

[0185] The specific operational relationships can be expressed as follows:

[0186] D _AE =N _AE_i / N _AE_max ;

[0187] In the above formula, D _AE N is the second microscopic damage parameter output. _AE_i N represents the current cumulative number of events counted at the i-th test node. _AE_max The cumulative number of failure events at the failure nodes of concrete specimens.

[0188] The above provides input conditions for subsequent cross-scale data fusion modules.

[0189] According to the embodiments of this application, damage characterization parameters were obtained from three physical dimensions: macroscopic mechanics, surface optics, and internal acoustics. However, the evolution rate of each parameter differs at different stages. If the stage boundaries are defined by manually observing the inflection point of a single curve, the interpretation results of different operators may be biased.

[0190] Therefore, this embodiment establishes a multi-segmented piecewise regression mathematical model with physical constraints, transforming the determination of stage boundaries into a mathematical optimization problem of minimizing global residuals, thereby reducing the reliance on subjective interpretation in damage threshold setting. The specific implementation is as follows:

[0191] Optionally, a multi-segmented piecewise regression model is constructed to represent the first correlation between the first microscopic damage parameter and the macroscopic damage parameter, and the second correlation between the second microscopic damage parameter and the macroscopic damage parameter, respectively.

[0192] Among them, the multi-segment piecewise regression model divides the data into multiple continuous linear intervals by setting at least two levels of dividing points.

[0193] For example, a two-dimensional coordinate mapping system can be established with macroscopic damage parameters as independent variables and microscopic damage parameters as dependent variables. To address the nonlinear abrupt changes in the material degradation process, a multi-segmented regression model can be configured within this coordinate system.

[0194] Furthermore, by setting two unknown boundary breakpoints, the entire coordinate domain is divided into three independent linear response intervals along the horizontal axis. This model is used to fit the surface image feature evolution data stream and the internal acoustic feature evolution data stream, respectively, thereby transforming the physical evolution process into a piecewise linear combination.

[0195] Accordingly, D _micro =a _k ×D _a +b _k ;

[0196] Among them, D _micro The dependent variable is D, representing either the first or second microscopic damage parameter. _a The independent variable is denoted by , representing the macroscopic damage parameter, k is the index number of the segmented interval, and a is denoted by . _k Let b be the slope parameter of the linear function in the k-th segment. _k Let be the intercept parameter of the k-th segment of the linear function.

[0197] In this embodiment, the case of setting two levels of boundary points and dividing the linear interval into three segments is used as an example for explanation. It can be understood that when a more refined stage division is required, the number of boundary points can be expanded to three or more levels, and the number of segments can be increased accordingly. The continuity constraint condition must be met at each adjacent boundary point, and the optimization method is similarly applicable.

[0198] In some embodiments, the multi-segmented piecewise regression model satisfies preset physical boundary conditions and continuity constraints, specifically including:

[0199] At the boundary point corresponding to the intersection of adjacent linear intervals, the calculation results of two adjacent linear function segments are constrained to be equal;

[0200] When the macroscopic damage parameter represents a zero initial damage state, the intercept parameter of the corresponding first linear function is limited to 0.

[0201] Accordingly, physical boundary conditions and continuity constraint equations are injected into the multi-segmented piecewise regression model.

[0202] At the starting point of the first linear interval, i.e., under the state of zero initial damage where the macrostructure is not subjected to stress degradation, the dependent variable reading is forced to zero, thereby locking the ordinate of the first regression line.

[0203] Simultaneously, at the set coordinates of each boundary point, equality constraints are established between adjacent intervals, forcing the endpoints of the two function segments to coincide in two-dimensional space. The specific formulas for the constraint conditions are as follows:

[0204] a _1 ×D' _1 +b _1 =a _2 ×D' _1 +b _2 ;

[0205] a _2 ×D' _2 +b _2 =a _3 ×D' _2 +b _3 ;

[0206] b _1 =0;

[0207] In the above formula, D' _1 This is the first-level candidate boundary point;

[0208] D' _2 This is the designated second-level candidate boundary point;

[0209] a _1 With b _1 These are the parameters of the first linear function segment;

[0210] a _2 With b _2 These are the parameters of the second linear function.

[0211] a _3 With b _3 The parameters of the third linear function;

[0212] The equation holds true if the ending value of the preceding stage is equal to the starting value of the following stage.

[0213] In some embodiments, the specific operations for optimizing evolutionary boundary points include:

[0214] Within the range of values ​​for the macroscopic damage parameter, iterate through the combinations of candidate boundary points;

[0215] For each candidate combination, the corresponding parameter data is divided into multiple segments for linear fitting calculation to obtain the overall fitting residual evaluation index.

[0216] The combination of the boundary points corresponding to the minimum value of the fitting residual evaluation index is determined as the optimal mutation boundary point.

[0217] In this step, within the legal numerical range of the macroscopic damage parameters, a candidate boundary point coordinate network for the entire set is generated according to the set step size.

[0218] The above step size can be adjusted by technicians according to the actual data sampling density and computing resources.

[0219] In this case, for each candidate breakpoint combination in the network, the measured scatter data is divided into corresponding data subsets, and least squares regression is performed in each subset;

[0220] Calculate the sum of squared distances of all data points from the fitted line, and use the sum as the overall residual evaluation index to determine the goodness of fit of the current combination;

[0221] Perform a full-domain traversal comparison to extract a predetermined combination of breakpoints that causes the overall evaluation index to converge to a minimum value, and output it as the optimal abrupt change boundary point characterizing the transformation of the internal and external damage evolution mechanism of the material.

[0222] For example, the calculation of the fitting residual evaluation index can be expressed as the following formula:

[0223] RSS=∑((D _micro_i -D _predict_i ) 2 );

[0224] Where RSS corresponds to the overall fit residual evaluation index, D _micro_i The corresponding extracted measured micro-damage parameter values, D _predict_i The corresponding predicted value is calculated based on the current candidate boundary point combination and the regression model output. ∑ is the summation operation performed on all discrete sample points within multiple intervals.

[0225] As an alternative, when the large amount of input sample data leads to excessive computational resources being consumed by grid traversal search, heuristic search strategies such as genetic algorithms or particle swarm optimization algorithms can be used to replace exhaustive traversal. The RSS value is used as an evaluation term of the fitness function, and the optimal mutation boundary coordinates are approximated through population iteration and convergence to improve analysis efficiency.

[0226] In some embodiments, determining a damage evolution threshold and classifying the damage level of concrete specimens based on the damage evolution threshold includes:

[0227] The optimal mutation boundary points corresponding to the first microscopic damage parameter and the second microscopic damage parameter after optimization are extracted respectively. The optimal mutation boundary points include the first-level boundary point and the second-level boundary point.

[0228] Calculate the mean value of the first limit of the first-level boundary point corresponding to the two microscopic damage parameters, and the mean value of the second limit of the corresponding second-level boundary point;

[0229] Using the first and second limit mean values ​​as damage evolution thresholds, the damage state of concrete specimens is divided into three progressively increasing damage level intervals.

[0230] Among these features, optical surface features and acoustic internal features exhibit phase asynchrony at the inflection point of material degradation. To address this, the first set of double-boundary coordinates obtained through image branch optimization and the second set of double-boundary coordinates obtained through acoustic branch optimization can be read.

[0231] Mean smoothing is performed on the coordinate sequence at the same level to output a comprehensive boundary parameter that balances multi-source biases, and this parameter is fixed as the damage evolution threshold. That is:

[0232] D _threshold_1 =(D' _1_opt +D'' _1_opt ) / 2;

[0233] D _threshold_2 =(D' _2_opt +D'' _2_opt ) / 2;

[0234] In the above formula:

[0235] D _threshold_1 This is the calculated first bound mean;

[0236] D _threshold_2 The mean of the second limit obtained from the calculation;

[0237] D' _1_opt With D' _2_opt The boundary point between the first and second levels is used to optimize the output of the first microscopic damage parameter;

[0238] D'' _1_opt With D'' _2_opt The boundary point between the first and second levels is used for optimizing the output of the second microscopic damage parameter.

[0239] Based on this, the scale coordinate axis is defined as three continuous and independent intervals, corresponding to different degradation tiers that characterize the service life of the specimen.

[0240] According to embodiments of this application, the three progressively increasing damage level intervals may specifically include:

[0241] The initial damage level range characterizing the dispersion and accumulation of micro-damage within the material, i.e., level I;

[0242] The damage development level range, i.e., Level II, characterizes the localized interconnection evolution trend of microcracks inside the material;

[0243] The critical instability level range, or Level III, is used to characterize the rapid propagation and penetration of the main crack inside the material.

[0244] In other words, the damage state is determined based on the mapping relationship of the measured parameters.

[0245] That is, when the measured macroscopic damage parameter value is between zero and the mean of the first limit, the initial damage level range instruction is output.

[0246] At this point, the overall stiffness of the structure decreases slightly, and acoustic events and strain concentration phenomena are in a gradual growth phase. The damage mechanism is mainly characterized by the dispersed nucleation of microcracks, and the concrete maintains a high safety redundancy.

[0247] When the value of this parameter crosses the first limit mean and is below the second limit mean, output the damage development level range instruction;

[0248] In this state, the slope of the regression curve becomes steeper, acoustic events occur more frequently, and surface strain shows connectivity, indicating that material damage is evolving from a dispersed state to a localized state, and the risk level is adjusted upward accordingly.

[0249] When the measured values ​​exceed the upper limit of the second boundary average, it is classified as a critical instability level range. Parameters in all dimensions show rapid growth, microcracks coalesce to form macroscopic main cracks, the internal load-bearing skeleton of the concrete collapses, triggering the alarm response logic of the corresponding control system.

[0250] In this application, macroscopic mechanical feedback, internal acoustic events and surface deformation images are simultaneously integrated to construct a cross-scale damage parameter system covering internal and external dimensions, which maps the evolution process of materials from microscopic initiation to macroscopic failure.

[0251] Furthermore, a multi-segment regression mathematical network based on physical boundaries and continuity constraints was established. This network automatically captures the abrupt change boundary corresponding to the minimization of the overall fitting error through a global search, and further calibrates using the boundary mean of multi-source parameters. This mechanism transforms the determination of the damage threshold into a mathematical solution process, thereby providing a safety early warning benchmark for the damage state assessment of concrete structures.

[0252] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for classifying concrete damage using a combination of AE and DIC technologies, characterized in that, include: Acquire macroscopic mechanical response data, surface image data, and internal acoustic data of concrete specimens simultaneously; Based on macroscopic mechanical response data, stiffness degradation characteristics of each cyclic loading and unloading stage are extracted to construct macroscopic damage parameters; Based on surface image data, surface feature parameters characterizing the degree of spatial localization evolution of surface cracks are extracted to construct the first microscopic damage parameter; Based on internal acoustic data, acoustic feature parameters characterizing the degree of internal crack activity are extracted to construct a second microscopic damage parameter. The evolution boundary point is optimized by combining macroscopic damage parameters, first microscopic damage parameters, and second microscopic damage parameters to determine the damage evolution threshold. The damage level of concrete specimens is classified based on the damage evolution threshold.

2. The method according to claim 1, characterized in that, Acquire macroscopic mechanical response data of concrete specimens collected simultaneously, including: Three-point bend cyclic loading and unloading tests were performed on concrete specimens; the loading phase of the three-point bend cyclic loading and unloading test was controlled by the crack opening displacement, and the unloading phase was controlled by the load. Macroscopic mechanical response data were collected during the three-point bend cyclic loading and unloading test.

3. The method according to claim 1, characterized in that, Based on macroscopic mechanical response data, stiffness degradation characteristics of each cyclic loading and unloading stage are extracted to construct macroscopic damage parameters, including: The initial stiffness, load value and crack opening displacement value corresponding to the unloading point in each cyclic loading and unloading stage, and residual load value and residual crack opening displacement value corresponding to the unloading end point are extracted from the macroscopic mechanical response data. Based on the difference between the load value and the residual load value, and the difference between the crack opening displacement value and the residual crack opening displacement value, the cyclic stiffness of the corresponding cyclic loading and unloading stage is determined. The stiffness degradation characteristics are determined based on the ratio between cyclic stiffness and initial stiffness, and these characteristics are used as macroscopic damage parameters.

4. The method according to claim 1, characterized in that, Based on surface image data, surface feature parameters characterizing the degree of spatial localization evolution of surface cracks are extracted to construct the first microscopic damage parameters, including: Calculate the surface strain field of the corresponding test area based on surface image data; Extract strain concentration regions characterizing local deformation from the surface strain field; Based on preset image size parameters, the strain concentration area is converted into the actual physical area, and the current physical area corresponding to the current test time and the failure physical area corresponding to the failure time of the concrete specimen are obtained respectively. Based on the ratio of the current physical area to the failed physical area, surface characteristic parameters are determined and used as the first microscopic damage parameter.

5. The method according to claim 4, characterized in that, From the surface strain field, strain concentration regions characterizing local deformation are extracted, including: The surface strain field is segmented using an image segmentation algorithm to obtain a preliminary binarized image containing concentrated deformation features. By using connected component analysis, connected regions are filtered in the binarized image to remove noise regions with area features smaller than the preset filtering conditions, thus obtaining the denoised strain concentration region.

6. The method according to claim 1, characterized in that, Based on internal acoustic data, acoustic feature parameters characterizing the degree of internal crack activity are extracted to construct a second microscopic damage parameter, including: The current cumulative number of events corresponding to the current test time of the internal acoustic data is counted, as well as the cumulative number of failure events corresponding to the time of failure of the concrete specimen; Based on the ratio of the current cumulative number of events to the cumulative number of failure events, acoustic characteristic parameters are determined and used as the second microscopic damage parameter.

7. The method according to claim 1, characterized in that, The evolutionary boundary point is optimized by combining macroscopic damage parameters, first microscopic damage parameters, and second microscopic damage parameters, including: Construct a multi-segmented piecewise regression model to represent the first correlation between the first microscopic damage parameter and the macroscopic damage parameter, and the second correlation between the second microscopic damage parameter and the macroscopic damage parameter, respectively; Among them, the multi-segment piecewise regression model divides the data into multiple continuous linear intervals by setting at least two levels of dividing points.

8. The method according to claim 1, characterized in that, Finding the optimal evolutionary boundary point specifically includes: Within the range of values ​​for the macroscopic damage parameter, iterate through the combinations of candidate boundary points; For each candidate combination, the corresponding parameter data is divided into multiple segments for linear fitting calculation to obtain the overall fitting residual evaluation index. The combination of the boundary points corresponding to the minimum value of the fitting residual evaluation index is determined as the optimal mutation boundary point.

9. The method according to claim 1, characterized in that, Determining the damage evolution threshold includes: The optimal mutation boundary points corresponding to the first microscopic damage parameter and the second microscopic damage parameter after optimization are extracted respectively. The optimal mutation boundary points include the first-level boundary point and the second-level boundary point. Calculate the mean value of the first limit of the first-level boundary point corresponding to the two microscopic damage parameters, and the mean value of the second limit of the corresponding second-level boundary point; The mean of the first and second limits are used as the thresholds for damage evolution.

10. The method according to claim 4, characterized in that, The surface strain field of the corresponding test area is calculated based on surface image data, specifically including: Using a preset image matching window, the degree of matching between subsets of surface image data before and after loading deformation is evaluated based on the cross-correlation coefficient algorithm, and the surface strain field is calculated.

Images