A method for testing and analyzing the solidification performance of a soil stabilizer

Abstract

The application discloses a kind of soil solidifying agent solidification performance test analysis method, it is related to soil solidifying agent solidification performance test field, comprising: the present application obtains load-displacement curve and micro video stream in quasi-static loading process, and the load value and corresponding time stamp of characteristic inflection point in load-displacement curve are extracted, which represent crack initiation, stable expansion and unstable expansion feature;Identify the pixel coordinates of crack initiation point, and the pixel coordinates of crack initiation point are matched with the position of pore, interface defect in background structure feature map in space, to judge the defect type that crack initiation originates from;By obtaining the crack path from crack initiation point to unstable expansion, the tortuosity and branch number are extracted, combined with crack tip strain field concentration degree and crack origin type under different loading stages, the dominant cause leading to sample failure is analyzed, the collaborative analysis of loading mechanical behavior and micro deformation evolution is realized, and the failure mechanism of solidified soil can be identified.

Description

Technical Field

[0001] This invention belongs to the technical field of soil stabilizer solidification performance testing, and relates to a method for testing and analyzing the solidification performance of soil stabilizers. Background Technology

[0002] Soil stabilizers, as materials capable of improving soil engineering properties, have broad application prospects in geotechnical engineering fields such as roadbed reinforcement, slope protection, and site treatment. The quality of their solidification performance directly affects the safety and durability of the project; therefore, establishing scientific solidification performance testing methods is crucial.

[0003] Existing technologies have proposed macroscopic testing and optimization schemes for the engineering performance of solidified soil. For example, the invention patent with publication number CN118464574A proposes a performance analysis method for solidified soil buffer materials based on soil reinforcement technology. Through soil compaction, unconfined compressive strength, direct shear and rockfall impact simulation tests, the influence of solidifying agent on the macroscopic mechanical properties of soil is analyzed. This method treats the soil as a homogeneous black box, analyzes the terminal mechanical response, and then makes a performance evaluation of the solidifying agent.

[0004] Although the existing methods have achieved results in macroscopic performance evaluation, they still have the following shortcomings: First, the existing technology relies entirely on the peak stress, shear strength parameters, and peak impact force at the final failure to evaluate the solidification performance. This method treats the soil as a black box, only observing the final relationship between its load and displacement and damage, while completely lacking observation of the internal damage initiation, accumulation, and evolution behavior of the soil from the beginning of bearing to the final failure process. It cannot reveal the damage development law and failure physical mechanism of the material before reaching the bearing limit, and therefore cannot fully assess its failure toughness.

[0005] Secondly, the existing performance evaluation of curing technology stops at the level of comparing the levels of macroscopic mechanical parameters. It cannot establish a causal relationship between macroscopic mechanical properties and the internal microstructural changes that cause these properties. This results in the performance evaluation conclusions remaining at the level of phenomenon description. Since it is impossible to establish a causal chain between macroscopic performance failure and microstructural defects, it is difficult to provide feedback with clear physical direction for substantial improvement of material properties, thus limiting its value in in-depth material diagnosis. Summary of the Invention

[0006] In view of this, in order to solve the problems mentioned in the background art, the present invention provides a method for testing and analyzing the solidification performance of soil stabilizers.

[0007] The objective of this invention can be achieved through the following technical solution: a method for testing and analyzing the solidification performance of a soil stabilizer, comprising: acquiring the load-displacement curve and microscopic video stream of a solidified soil sample during a quasi-static continuous loading process, and extracting the load values ​​and corresponding timestamps of the inflection points characterizing crack initiation, stable propagation and unstable propagation in the load-displacement curve.

[0008] Based on the timestamp, the image frames corresponding to the time moment are extracted from the microscopic video stream, the pixel coordinates of the crack initiation point are identified, and spatial matching is performed with the position coordinates of pores and interface defects in the background structural feature image of the solidified soil sample to determine the type of defect from which the crack initiation originates.

[0009] The crack path from the crack initiation point to unstable propagation is obtained from the microscopic video stream, and the tortuosity and number of branches of the crack path are obtained.

[0010] Based on the crack path, the tracking scale of the computational region is adjusted according to the time interval corresponding to the inflection point of the load characteristics to obtain the degree of strain field concentration at the crack tip under different loading stages.

[0011] Based on the defect type of crack initiation, the tortuosity and number of branches of the crack path, and the degree of strain field concentration at the crack tip, the dominant cause leading to the final failure of the specimen is determined.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) The present invention obtains the load-displacement curve and microscopic video stream during the quasi-static loading process, and extracts the load value and corresponding timestamp of the inflection point of the load-displacement curve that characterizes the crack initiation, stable propagation and unstable propagation. This enables continuous observation of the internal damage initiation, propagation and evolution behavior of solidified soil from the beginning of bearing to the final failure process, and overcomes the limitation of the prior art that regards the soil as a black box and only focuses on the final failure parameters.

[0013] (2) This invention identifies the pixel coordinates of the crack initiation point and spatially matches the pixel coordinates of the crack initiation point with the positions of pores and interface defects in the background structural feature diagram to determine the type of defect from which the crack initiation originates. This establishes a clear causal relationship between the macroscopic mechanical properties and the internal microstructure that causes the change in these properties. It not only reveals the whole process law of soil damage development and the physical mechanism of failure, and realizes the assessment of damage toughness, but also elevates the performance evaluation from phenomenon description to the microstructure level.

[0014] (3) By obtaining the crack path from the crack initiation point to the unstable propagation, extracting the tortuosity and the number of branches, and combining the degree of strain field concentration at the crack tip and the type of crack origin under different loading stages, this invention analyzes the dominant cause of the sample failure, realizes the structured diagnosis of the dominant cause of the final failure of the sample, breaks through the limitation of relying on the comparison of a single macroscopic parameter, and provides feedback basis with clear physical orientation for the performance improvement of soil stabilizers. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a diagram illustrating the implementation steps of the method of the present invention.

[0017] Figure 2 This is a flowchart of the crack initiation point pixel coordinate recognition process of the present invention.

[0018] Figure 3 This is a flowchart illustrating the process of obtaining the crack initiation to unstable propagation path in this invention. Detailed Implementation

[0019] 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.

[0020] Please see Figure 1 As shown, the present invention provides a method for testing and analyzing the solidification performance of soil stabilizers, including: S1, obtaining the load-displacement curve and microscopic video stream of the solidified soil sample during the quasi-static continuous loading process, and extracting the load values ​​and corresponding timestamps of the inflection points characterizing crack initiation, stable propagation and unstable propagation in the load-displacement curve.

[0021] The load-displacement curve and microscopic video stream are obtained through the following steps: First, the solidified soil sample to be tested is installed in the fixture of the material testing machine, and the solidified soil sample is subjected to quasi-static continuous loading. The loading speed can be controlled within 0.05 mm / min to 2 mm / min to ensure that the sample response is quasi-static.

[0022] Next, during the loading process, the load value of the solidified soil sample is collected in real time by the load sensor integrated into the material testing machine; at the same time, the displacement of the loading point is collected in real time by the displacement sensor of the material testing machine; the load and displacement data are recorded synchronously, and the load-displacement curve is plotted in real time.

[0023] Then, during the loading of the solidified soil sample, the sample surface was continuously observed using an optical microscope equipped with a high-speed camera. Microscopic images of the sample surface were continuously captured during the loading process to form a microscopic video stream.

[0024] Finally, to ensure the correspondence between mechanical data and image data, one of the following synchronization control methods can be adopted: use a periodic synchronization pulse signal of the material testing machine as an external trigger signal and connect it to the trigger port of the high-speed camera to control the camera to start recording at the same time as loading begins, so as to achieve synchronization of the two based on the same hardware clock; or connect the data acquisition of the material testing machine and the control of the high-speed camera to the same clock source to achieve data alignment afterward.

[0025] The steps for extracting the load values ​​and corresponding timestamps representing the inflection points of crack initiation, stable propagation, and unstable propagation in the load-displacement curve are as follows: filter the load-displacement curve and calculate the first and second derivatives of the load with respect to the displacement after filtering.

[0026] The above filtering process uses a smoothing filter to filter the original load-displacement curve. The filter window width is selected as a sliding window containing an odd number of data points, and is set according to the sampling frequency and signal characteristics. For example, it can be set to 1% to 5% of the total number of data points. Local least squares fitting is performed using a second-order polynomial to obtain a smooth load-displacement curve, thereby suppressing noise interference while preserving the true shape of the curve's characteristic inflection points. The first and second derivative sequences of the load with respect to displacement can be calculated using the central difference method in practical discrete data processing.

[0027] The starting point where the first derivative sequence first shows a sustained monotonically decreasing trend is identified as a candidate point for crack initiation. The mean value of the second derivative load before and after the candidate point is calculated. If the mean value is less than zero, the point is confirmed as the first feature inflection point, and its timestamp and load value are recorded.

[0028] When identifying candidate points for crack initiation, several data points can be set as 5 consecutive data points.

[0029] Since crack initiation usually corresponds to irreversible micro-damage starting inside the material, the first derivative of the load-displacement curve begins to decrease continuously.

[0030] To eliminate misjudgments caused by local fluctuations due to noise, candidate points need to be determined: take a preset time window before and after the candidate point. The preset time window can be set to 1% to 2% of the total test time. Calculate the mean of the second derivative sequence within these two windows. If the mean is less than zero, the curve will show an upward convex shape in this interval, thus confirming that the point is the first feature inflection point.

[0031] The point where the global maximum value of the load in the load-displacement curve is identified as the second characteristic inflection point, and its timestamp and load value are recorded.

[0032] During the testing of curing agents, the maximum load point is usually the critical precursor for cracks to shift from stable propagation to unstable propagation, corresponding to the maximum load point in the load-displacement curve.

[0033] In the load decrease segment after the second characteristic inflection point, the curvature sequence is calculated based on the first and second derivatives of the load-displacement curve. The extreme point with the largest absolute value in the curvature sequence is identified as the third characteristic inflection point, and its timestamp and load value are recorded.

[0034] It should be noted that since the first derivative of the load decrease segment after the second characteristic inflection point is usually negative and has a large absolute value, the curvature calculation formula can effectively amplify the curvature change.

[0035] When instability occurs, the crack propagates rapidly, and the load often drops sharply, causing the curve to show a significant inflection point. Mathematically, this is represented by the curvature reaching its maximum absolute value, which is usually a negative minimum point, indicating that the curve bends sharply.

[0036] S2. Extract the image frames corresponding to the timestamp from the microscopic video stream, identify the pixel coordinates of the crack initiation point, and spatially match them with the position coordinates of pores and interface defects in the background structural feature image of the solidified soil sample to determine the type of defect from which the crack initiation originates.

[0037] See Figure 2 As shown, in order to define the location of the crack initiation point, the steps to determine the initial spatial location of crack initiation are as follows: extract the image frame corresponding to the timestamp of the first feature inflection point, and the average image of the previous multiple consecutive frames to obtain the current frame and the background frame. This background frame represents the static texture of the specimen surface in the crack-free state, separating the dynamic changes from the static background.

[0038] It should be added that, in order to ensure uniform and stable illumination on the specimen surface and to minimize the interference of shadows and specular reflections on image analysis, the illumination system should use uniform and stable coaxial light or dual-sided cold light sources.

[0039] The background frame is constructed using a series of consecutive images, corresponding to a time period prior to the crack initiation point determined by the load-displacement curve. Furthermore, no new surface changes are observed in the microscopic video stream within this time period in the observed area of ​​the sample. Preferably, 20 to 50 consecutive images are selected before the load value reaches 70% of the crack initiation load value to calculate the average background frame.

[0040] The current frame and the background frame are subjected to image difference, enhancement, and binarization to identify all connected regions.

[0041] The image difference processing between the current frame and the background frame is the difference between the gray value of each pixel in the current frame and the gray value of the corresponding pixel in the background frame. The absolute value of the difference is taken to obtain a difference image that represents the gray value difference between the two frames. The brighter areas in the difference image represent new features that are different from the background.

[0042] The contrast-limited enhancement process for the difference image specifically includes: first, dividing the difference image into context regions; then, cropping the grayscale histogram of each context region to limit its distribution slope; then, equalizing the cropped histogram; and finally, smoothing the boundaries between adjacent regions using bilinear interpolation to obtain a locally contrast-enhanced image, thereby improving the visibility of low-contrast cracks and ensuring that microcracks are not missed.

[0043] The binarization process uses Otsu's method to automatically calculate the global threshold: it iterates through all possible grayscale thresholds, calculates the inter-class variance between candidate cracks and the background, and takes the threshold that maximizes the inter-class variance as the global threshold. Based on the global threshold, the enhanced image is converted into a black and white binary image: pixels with grayscale values ​​greater than or equal to the global threshold are set to white, representing the foreground target, and the rest are set to black, representing the background.

[0044] Finally, in the binary image, all sets of interconnected white pixels are marked as independent connected regions.

[0045] The ratio of the principal axis length to the secondary axis length of each connected region is calculated as the aspect ratio, and the ratio of the area of ​​the connected region to the area of ​​its smallest bounding rectangle is calculated as the linearity.

[0046] Aspect ratio describes the length-to-width ratio along the direction of extension, while linearity describes the degree to which a connected region fills its circumscribed rectangle. Real cracks typically have high aspect ratio and high linearity, distinguishing them from interference from particles, scratches, noise, etc.

[0047] Sort all connected regions in descending order of aspect ratio and linearity value, and select connected regions that are in the top quartile of both aspect ratio and linearity as candidate regions.

[0048] Considering that real cracks typically exhibit long, continuous, or slightly curved linear characteristics in their early stages, a high aspect ratio indicates a narrow and elongated region, while high linearity indicates a high degree of filling of its circumscribed rectangle and a regular shape. Therefore, both a high aspect ratio and high linearity are required to geometrically match the physical characteristics of the crack, thereby eliminating circular pores, irregular clusters of noise, or diffuse damage areas.

[0049] Under different samples, lighting conditions, or imaging conditions, the absolute distribution range of connected region feature values ​​may vary greatly. By sorting and selecting the top 25%, the most prominent morphological features of the region are automatically selected within each image. This avoids over- or under-screening problems caused by improperly setting a fixed threshold, and improves the universality of the method for different testing scenarios.

[0050] Calculate the geometric center coordinates of the earliest appearing region in the candidate region, and use them as the pixel coordinates of the crack initiation point.

[0051] Crack initiation is a sequential process that gradually expands outward from a single initial point, rather than an instantaneous change occurring simultaneously in multiple regions. The region that appears earliest among the candidate regions corresponds to the initial source of crack initiation, and subsequent candidate regions are extensions formed from that initial point.

[0052] Furthermore, the geometric center coordinates of the earliest appearing region in the candidate region are selected as the representative point because the germination region may be small and irregular in shape in the initial stage, and its geometric center can represent the overall spatial position of the region, and the coordinates are not sensitive to local shape changes of the region.

[0053] The method for determining the defect type from which crack initiation originates includes: extracting the geometric center coordinates of each defect from the background structural feature map and calculating its equivalent radius.

[0054] Before mechanical testing, the background structural feature map is obtained by three-dimensional imaging of the solidified soil sample using micro-computed tomography. Then, the coordinate system of the background structural feature map is unified with the pixel coordinate system of the microscopic video. At least three positioning marks that can be recognized in both imaging modes are prepared on the sample surface. The spatial alignment of the two coordinate systems is achieved by perspective transformation based on the coordinates of these marks.

[0055] The process of identifying pores and interface defects in the background structural feature image is as follows: First, the background structural feature image is segmented, and the global grayscale threshold is calculated using the Otsu method. All pixels with grayscale values ​​lower than the global grayscale threshold are classified as potential defect regions, and the remaining pixels are classified as matrix material regions, thus obtaining a binary image. Second, 8-connected component analysis is performed on the binary image to identify all independent potential defect regions. If a connected region is completely surrounded by matrix material pixels, it is classified as a pore defect. If a connected region is in contact with or intertwined with the boundary of the region representing soil particles or aggregates in the image, it is classified as an interface defect.

[0056] For each defect region after classification, the arithmetic mean of the coordinates of all pixels within the defect region is calculated to obtain its centroid coordinates, which are the geometric center coordinates of the defect.

[0057] Each defect is approximated as a circle. The pixel area of ​​the defect region that is approximately circular is counted. The square root of the ratio of the pixel area of ​​the defect region to pi is calculated and denoted as the equivalent radius.

[0058] Using the initial pixel coordinates of the crack initiation point as the query point, calculate the Euclidean distance from it to the coordinates of the geometric center of each defect.

[0059] The ratio of the square of the Euclidean distance to the square of the equivalent radius of the corresponding defect is denoted as the weight value of each defect.

[0060] Crack initiation depends not only on the spatial proximity to a defect, but also on the size of the defect itself. Under the same conditions, larger defects, due to their greater stress concentration effect, may be more dangerous and more likely to become crack initiators, even at greater distances, than smaller, adjacent defects.

[0061] Specifically, the square of the Euclidean distance represents spatial proximity; a smaller value indicates that the initiation point is closer to the defect in location. The square of the defect's equivalent radius represents the size influence weight of the defect; a larger value indicates that the defect is more likely to become a crack initiation point under the same conditions. In fracture mechanics, the degree of stress concentration induced by defects is usually proportional to the square of its characteristic size. Therefore, the smaller the weight value of each defect, the stronger the correlation between the defect and crack initiation.

[0062] If the base point with the smallest attribution weight is selected, the defect corresponding to that base point is determined to be the origin defect associated with the crack initiation space.

[0063] Selecting the base point with the smallest attribution weight value indicates that it is either close enough to the crack initiation point, or its size is large enough and its influence is strong enough, or both, so that the crack initiation point is most likely located in the stress field or damage influence zone dominated by the defect.

[0064] S3. Obtain the crack path from the crack initiation point to the unstable propagation from the microscopic video stream, and obtain the tortuosity and number of branches of the crack path.

[0065] See Figure 3 As shown, the steps for obtaining the crack path from the crack initiation point to the unstable propagation are as follows: Starting from the image frame corresponding to the timestamp of the crack initiation point, the following steps are iteratively executed until the timestamp corresponding to the third feature inflection point is reached: (a) Calculate the dense optical flow field between the current frame and the next frame.

[0066] The dense optical flow field is calculated using the Farneback dense optical flow algorithm. The specific process is as follows: First, the local neighborhood grayscale distribution of each pixel in the image is modeled as a quadratic polynomial surface. Then, the polynomial coefficients of the corresponding neighborhood in the current frame and the next frame are fitted using the weighted least squares method. Next, based on the assumption of constant brightness, a constraint equation is established to allow the polynomial coefficients to change with displacement. Finally, by solving this constraint equation, the two-dimensional displacement vector of each pixel in the image from the current frame to the next frame is calculated, thereby obtaining the dense optical flow field and acquiring the motion vector of each pixel in the image between the current frame and the next frame.

[0067] (b) Predict the candidate position of the crack tip in the next frame based on the motion vector of the current crack tip position.

[0068] Specifically, a small neighborhood of k×k pixels is defined centered on the current crack tip position, for example, k=5. The motion vectors of all pixels within this small neighborhood are extracted from the optical flow field obtained in step (a).

[0069] Calculate the median vector of these motion vectors To resist interference from local noise or erroneous optical flow vectors, the median vector is calculated as follows: First, taking the current crack tip position as the center, collect all motion vectors provided by the optical flow field in its surrounding neighborhood.

[0070] Suppose we obtain M vectors, each containing a horizontal displacement component u and a vertical displacement component v. Then, we extract the horizontal and vertical displacement components of these M vectors to form two independent datasets.

[0071] Then, the set of horizontal displacement components is sorted, and the value in the middle position is found as the median of the horizontal displacement components. If M is odd, the median value after sorting is taken; if M is even, the arithmetic mean of the two middle values ​​after sorting is taken. The same process is performed on the set of vertical displacement components to obtain the median of the vertical displacement components.

[0072] Finally, the median of the horizontal component and the median of the vertical component are combined to form the final median vector, which represents a robust center estimate of the pixel motion trend within the neighborhood.

[0073] Predicted candidate position P pred The calculation is as follows: ,in, , These are the horizontal and vertical coordinates of the current crack tip position.

[0074] (c) Perform local image feature matching in the neighborhood of the candidate position in the next frame, correct and determine the new position of the crack tip, and add the new position to the crack path point sequence.

[0075] Using the crack tip position in the current frame and the predicted position in the next frame as the center, an a×a pixel image block is extracted as the template and an n×n pixel search area, for example, a=15, n=31, and n>a.

[0076] The above local image feature matching steps are as follows: slide the template within the search area, calculate the normalized cross-correlation coefficient at each possible position, find all positions with a normalized cross-correlation coefficient greater than 0.7, and each position corresponds to a successfully matched candidate feature region, effectively correcting the error of optical flow prediction.

[0077] Since the original tip splits into multiple new tips after the crack bifurcates, and each new tip has image features similar to the atomic region template in the next frame, the above correction and determination of the new position of the crack tip can be achieved through the following steps: First, when there is no bifurcation, if only one candidate feature region has a normalized cross-correlation coefficient exceeding 0.7 and is the maximum value, then the center position of that region is determined as the new position of the crack tip; when bifurcation occurs, if two or more spatially separated candidate feature regions with normalized cross-correlation coefficients exceeding 0.7 are identified, then it is determined that crack bifurcation has occurred in the current step.

[0078] Next, the tip of the main branch is updated to the center of the region with the highest normalized cross-correlation coefficient with the previous position.

[0079] Finally, each new branch starts at the center of the candidate region, starts at the next frame, and is added to the activity tracking branch list.

[0080] It should be added that if no matching feature region is identified within the neighborhood, the search range is expanded. If the location of the crack tip still cannot be determined, the tracking of the current branch is terminated and the tracking failure time is recorded.

[0081] During the matching process in step (c), if multiple feature regions that meet the matching conditions are identified in the neighborhood, it is determined that a crack bifurcation has occurred, and a new tracking branch is created, and iterative steps are executed in parallel for each tracking branch.

[0082] Connect the sequence of crack path points obtained in all iterative steps to form a crack path from the initiation point to the unstable propagation.

[0083] The process of obtaining the tortuosity of the crack path includes: performing skeletonization processing on the crack path and extracting the center path line of the crack path skeleton.

[0084] The skeletonization process for the crack path is performed as follows: First, the binary image of the crack path is used as input, where the crack pixel value is 1 and the background value is 0. Then, a thinning algorithm is used to iteratively peel away the outer pixels of the crack region layer by layer. In each iteration, all crack pixels are traversed, and it is determined whether they are deletable boundary points based on the configuration status of their eight neighboring pixels, ensuring that the deleted boundary points do not destroy the topological connectivity of the crack path. Then, all pixels marked as deletable are deleted. The above iterative process is repeated until no pixels can be deleted. Finally, a skeleton center path line image with a single pixel width and maintaining the original path topology is obtained.

[0085] Extract the coordinates of all skeleton pixels from the skeleton center path line image and sort them by connectivity to form the skeleton center path line.

[0086] The ratio of the total length of the path line at the center of the skeleton to the straight-line distance between its first and last endpoints is used as the tortuosity of the crack.

[0087] Torque is an index that reflects a material's ability to resist linear crack propagation and induces a crack deflection toughening effect. When the path is straight, the tortuosity is 1; the more curved the path, the greater the tortuosity.

[0088] The steps for obtaining the number of branches are as follows: count the number of pixels belonging to the skeleton centerline within the eight neighborhoods of each pixel in the skeleton centerline image, and use this as the connection number of that pixel.

[0089] Pixels with a connection count of 1 are designated as endpoints, and pixels with a connection count of 3 or more are designated as branch points.

[0090] It should be noted that an isolated endpoint has only one skeleton neighbor and its connection count is 1; the midpoint of a straight line has two skeleton neighbors and its connection count is 2; and a T-shaped or Y-shaped intersection has three skeleton neighbors and its connection count is 3.

[0091] Tracing from the endpoint of the crack initiation point along the centerline of the skeleton to the crack tip at the moment of unstable propagation, when encountering a bifurcation point, the direction with the longest cumulative path length is taken as the main path.

[0092] Since the main crack usually extends further and consumes more energy, and the longest cumulative path length reflects the physical fact mentioned above, this standard can be used to distinguish and track the main crack propagation trajectory.

[0093] Except for the crack initiation point and the crack tip at the moment of unstable propagation, trace back along the skeleton centerline towards the main path, and the path that first reaches the main pixel is recorded as a branch.

[0094] The total number of branches is counted as the number of branches in the crack path.

[0095] S4. Using the crack path as a reference, adjust the tracking scale of the calculation region according to the time interval corresponding to the inflection point of the load characteristics to obtain the degree of strain field concentration at the crack tip under different loading stages.

[0096] The step of adjusting the tracking scale of the calculation region according to the time interval corresponding to the load feature inflection point includes: extracting the current crack tip position and its adjacent path points from the crack path within the time interval corresponding to each load feature inflection point.

[0097] For the crack tip position corresponding to the crack path at any given time, extracting several adjacent path points before and after the current time can smooth out position fluctuations caused by tracking or image noise.

[0098] Specifically, the number b of adjacent path points before and after the current moment is extracted, obtained by dynamically calculating the number based on the video capture frame rate f using a preset time window length ΔT. In one specific embodiment, ΔT is set to 10 milliseconds to remove minute fluctuations in the tip position caused by image noise within a single frame. This ensures that the calculated step size accurately reflects the instantaneous or near-instantaneous crack propagation behavior, avoiding over-averaging due to excessive time and masking rapid changes in the propagation rate. The value of b should be no less than 1, and the time span of the selected local path point set should not cross different propagation stages defined by the load characteristic inflection point.

[0099] The average Euclidean distance between adjacent path points is calculated as the local step size for the current crack propagation. This local step size reflects the instantaneous crack propagation rate at the current moment. The step size may be small and uniform during the stable propagation phase, and may increase significantly as instability approaches. This parameter is a sensitive indicator of the crack's dynamic behavior.

[0100] Using a fixed multiple of the local step size as the current calculation region window, and taking the current crack tip position as the center, a square calculation region is drawn on the sample surface with the calculation region window as the side length, to obtain the concentration analysis of the strain field at the crack tip at that moment.

[0101] The above method allows for the adaptive determination of the optimal window size and position for strain field analysis. When crack propagation is slow, the window automatically shrinks to improve spatial resolution and capture minute strain concentrations during the initiation stage; when crack propagation is rapid, the window automatically expands to ensure that a larger process region is accommodated and to avoid information loss.

[0102] The fixed factor α is an empirical constant whose value should ensure that the calculation area covers the main strain-affected region at the crack tip. This can be obtained through the following steps: First, divide the crack propagation process into three stages—initiation, stable propagation, and unstable propagation—based on the load characteristic inflection point; then, under the same test conditions at each stage, perform crack path tracing and digital image-related strain field measurements, adjusting the fixed factor until the calculation window can completely cover the main strain concentration region at the crack tip for that stage; finally, record the calibrated fixed factors for each stage.

[0103] Because the strain field gradient at the crack tip is relatively gentle in the initiation stage, a larger window is needed to capture it; in the unstable stage, the strain is highly concentrated in the region very close to the tip, and a smaller window can capture the peak value. For example, a fixed multiple of 8 to 12 is used in the crack initiation stage, 4 to 6 in the stable propagation stage, and 2 to 3 in the unstable propagation stage.

[0104] The degree of strain field concentration at the crack tip includes: performing digital image correlation analysis on the microscopic video stream image sequence within the defined computational region to obtain the maximum principal strain field within the computational region.

[0105] Specifically, an image sequence of the computational region centered on the crack tip is extracted. The first frame in the image sequence is defined as the reference image, and each subsequent frame used for analysis is defined as a deformed image. A computational grid is arranged on the reference image, and a reference sub-region is defined with each computation point as its center. Within the corresponding search region of the deformed image, the displacement field is obtained through zero-mean normalized correlation matching and inverse combination Gaussian-Newton optimization. Based on this displacement field, the displacement gradient of each point is calculated using the least squares method, and then the strain tensor is obtained according to the Cauchy strain formula. Finally, by performing eigenvalue decomposition on the strain tensor of each point, the maximum eigenvalue is extracted, thereby obtaining the maximum principal strain field of the entire computational region.

[0106] Calculate the average strain of all pixels with positive strain values ​​in the maximum principal strain field, and use it as an indicator of the concentration of the strain field at the crack tip.

[0107] The average strain value is insensitive to noise in individual pixels, resulting in more stable results and reflecting the average severity of the entire tensile strain concentration area, which is more representative of the overall mechanical state of the region than a single peak value.

[0108] S5. Based on the defect type of crack initiation, the tortuosity and number of branches of the crack path, and the degree of strain field concentration at the crack tip, determine the dominant cause leading to the final failure of the specimen.

[0109] The determination of the dominant cause of the final failure of the sample includes: if the crack initiation originates from interface defects, and the tortuosity of the crack path and the number of branches are respectively lower than the corresponding thresholds, then the weak interface bonding is determined to be the dominant cause of failure.

[0110] It should be noted that the specific steps for determining the corresponding threshold are as follows: First, from the completed historical soil stabilizer performance test experiments, all sample test data whose crack initiation points were confirmed by microscopic observation to originate from interface defects are selected; then, the measured values ​​of crack path tortuosity and branch number corresponding to each sample in this set of data are extracted; finally, the statistical upper limit of the tortuosity value and the statistical upper limit of the branch number are calculated respectively, such as the top 90% quantile, and these two statistical values ​​are set as the tortuosity threshold and branch number threshold for interface defect judgment, respectively.

[0111] When cracks initiate at interface defects, they will preferentially propagate along the weakly bonded interface. Their path is usually relatively straight and it is difficult to induce secondary microcracks in the matrix, thus exhibiting low tortuosity and few branches in morphology. Therefore, it can be determined that weak interface bonding is the dominant cause of failure.

[0112] If crack initiation originates from pore defects, then separate quantitative standards are established for the tortuosity of the crack path, the number of branches, and the degree of strain field concentration at the crack tip.

[0113] The tortuosity, number of branches, and strain concentration at the crack tip are classified into low, medium, and high levels, respectively. Among them, a high level of tortuosity indicates a significant crack deflection toughening mechanism, a high level of number of branches indicates significant microcrack bridging and energy dissipation, and a high level of strain concentration indicates strong local plastic deformation or damage capacity at the crack tip, or poor deformation capacity of the material itself leading to high strain concentration.

[0114] By comparing the three parameters at their respective grading levels, the failure mechanism corresponding to the highest-grade characteristic parameter is identified as the dominant cause of the final failure of the sample.

[0115] For example, if the tortuosity is high, while the number of branches and the degree of strain field concentration are medium or low, the main reason is insufficient matrix toughness, but the crack deflection toughening effect is significant.

[0116] If the number of branches is high, while the others are medium or low, the main reason is the outstanding bridging of microcracks in the matrix and the ability to dissipate energy.

[0117] If the strain field concentration is at a high level, while the others are at a medium or low level, then the absolute value of the strain field concentration should be used for judgment: if the value exceeds the yield strain reference value of the soil solidification material, it indicates that the local plastic deformation capacity of the matrix is ​​insufficient, and the high strain concentration leads to brittle fracture; if the value is lower than the yield strain reference value, but is still the highest among all levels, it indicates that the material has the plastic deformation capacity to relax the stress at the crack tip.

[0118] If two or more parameters are both at a high level, it is determined to be a synergistic effect of multiple mechanisms, such as crack deflection and microcrack bridging synergistic toughening.

[0119] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0120] 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.

[0121] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0122] 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.

[0123] Finally, 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 method for testing and analyzing the solidification performance of soil stabilizers, characterized in that: include: The load-displacement curves and microscopic video streams of solidified soil samples during quasi-static continuous loading were obtained, and the load values ​​and corresponding timestamps of the inflection points characterizing crack initiation, stable propagation and unstable propagation were extracted from the load-displacement curves. Based on the timestamp, the image frames corresponding to the time moment are extracted from the microscopic video stream, the pixel coordinates of the crack initiation point are identified, and spatial matching is performed with the position coordinates of pores and interface defects in the background structural feature image of the solidified soil sample to determine the type of defect from which the crack initiation originates. The crack path from crack initiation point to unstable propagation is obtained from the microscopic video stream, and the tortuosity and number of branches of the crack path are obtained. Based on the crack path, the tracking scale of the computational region is adjusted according to the time interval corresponding to the inflection point of the load characteristics to obtain the degree of strain field concentration at the crack tip under different loading stages. Based on the defect type of crack initiation, the tortuosity and number of branches of the crack path, and the degree of strain field concentration at the crack tip, the dominant cause leading to the final failure of the specimen is determined. The identified primary causes leading to the eventual failure of the specimen include: If the crack initiation originates from interface defects, and the tortuosity and number of branches of the crack path are lower than the corresponding thresholds, then the weak interface bonding is determined to be the dominant cause of the failure. If crack initiation originates from pore defects, then separate quantitative standards are established for the tortuosity of the crack path, the number of branches, and the degree of strain field concentration at the crack tip. By comparing the tortuosity of the crack path, the number of branches, and the degree of strain field concentration at the crack tip in their respective grading standards, the failure mechanism corresponding to the characteristic parameter with the highest grading level is identified as the dominant cause of the final failure of the specimen.

2. The method for testing and analyzing the solidification performance of a soil stabilizer according to claim 1, characterized in that: The steps for extracting the load values ​​and corresponding timestamps from the load-displacement curves that characterize the inflection points of crack initiation, stable propagation, and unstable propagation are as follows: The load-displacement curve is filtered, and the first and second derivatives of the load with respect to the displacement are calculated after filtering. The starting point where the first derivative sequence first shows a sustained monotonically decreasing trend is identified as a candidate point for crack initiation. The mean value of the second derivative load before and after the candidate point is calculated. If the mean value is less than zero, the point is confirmed as the first feature inflection point, and its timestamp and load value are recorded. The point where the global maximum value of the load in the load-displacement curve is identified as the second characteristic inflection point, and its timestamp and load value are recorded. In the load decrease segment after the second characteristic inflection point, the curvature sequence is calculated based on the first and second derivatives of the load-displacement curve. The extreme point with the largest absolute value in the curvature sequence is identified as the third characteristic inflection point, and its timestamp and load value are recorded.

3. The method for testing and analyzing the solidification performance of a soil stabilizer according to claim 2, characterized in that: The steps for identifying the pixel coordinates of the crack initiation point are as follows: Extract the image frame corresponding to the timestamp of the first feature inflection point, and the average image of the previous multiple consecutive frames to obtain the current frame and the background frame; Perform image differencing, enhancement, and binarization on the current frame and the background frame to identify all connected regions; The ratio of the principal axis length to the secondary axis length of each connected region is used as the aspect ratio, and the ratio of the area of ​​the connected region to the area of ​​its smallest bounding rectangle is used as the linearity. Sort all connected regions in descending order of aspect ratio and linearity value, and select connected regions that are in the top quartile of both aspect ratio and linearity as candidate regions. Calculate the geometric center coordinates of the earliest appearing region in the candidate region, and use them as the pixel coordinates of the crack initiation point.

4. The method for testing and analyzing the solidification performance of a soil stabilizer according to claim 1, characterized in that: The defect types used to determine the origin of crack initiation include: Extract the geometric center coordinates of each defect from the background structural feature map and calculate its equivalent radius; Using the initial pixel coordinates of the crack initiation point as the query point, calculate the Euclidean distance from it to the coordinates of the geometric center of each defect; The ratio of the square of the Euclidean distance to the square of the equivalent radius of the corresponding defect is denoted as the weight value of each defect. If the base point with the smallest attribution weight is selected, the defect corresponding to that base point is determined to be the origin defect associated with the crack initiation space.

5. The method for testing and analyzing the solidification performance of a soil stabilizer according to claim 2, characterized in that: The process of obtaining the crack path from the crack initiation point to unstable propagation includes: Starting from the image frame corresponding to the crack initiation point's timestamp, iteratively execute the following steps until the timestamp corresponding to the third feature inflection point is reached: (a) Calculate the dense optical flow field between the current frame and the next frame; (b) Predict the candidate position of the crack tip in the next frame based on the motion vector of the current crack tip position; (c) Perform local image feature matching in the neighborhood of the candidate position in the next frame, correct and determine the new position of the crack tip, and add the new position to the crack path point sequence; During the matching process in step (c), if multiple feature regions that meet the matching conditions are identified in the neighborhood, it is determined that a crack bifurcation has occurred, and a new tracking branch is created, and iterative steps are executed in parallel for each tracking branch. Connect the sequence of crack path points obtained in all iterative steps to form a crack path from the initiation point to the unstable propagation.

6. The method for testing and analyzing the solidification performance of a soil stabilizer according to claim 1, characterized in that: The tortuosity of the obtained crack path includes: The crack path is skeletonized, and the center path line of the crack path skeleton is extracted. The ratio of the total length of the path line at the center of the skeleton to the straight-line distance between its first and last endpoints is used as the tortuosity of the crack.

7. The method for testing and analyzing the solidification performance of a soil stabilizer according to claim 6, characterized in that: The steps for obtaining the number of branches are as follows: The number of pixels belonging to the skeleton centerline within the eight-neighborhood of each pixel in the skeleton centerline image is counted and used as the connection number of that pixel. Pixels with a connection count of 1 are designated as endpoints, and pixels with a connection count of 3 or more are designated as branch points. Tracing from the endpoint of the crack initiation point along the centerline of the skeleton to the crack tip at the moment of unstable propagation, when encountering a bifurcation point, the direction with the longest cumulative path length is taken as the main path; Except for the crack initiation point and the crack tip at the moment of unstable propagation, trace back along the skeleton centerline towards the main path, and the path that first reaches the main pixel is recorded as a branch. The total number of branches is counted as the number of branches in the crack path.

8. The method for testing and analyzing the solidification performance of a soil stabilizer according to claim 7, characterized in that: The step of adjusting the tracking scale of the calculation region according to the time interval corresponding to the inflection point of the load characteristics includes: Within the time interval corresponding to each load characteristic inflection point, extract the current crack tip position and its adjacent path points from the crack path; For the crack tip position corresponding to the crack path at any time, extract several adjacent path points before and after the current time. Calculate the average Euclidean distance between adjacent path points as the local step size for the current crack propagation; Using a fixed multiple of the local step size as the current calculation region window, and taking the current crack tip position as the center, a square calculation region is drawn on the sample surface with the calculation region window as the side length, to obtain the concentration analysis of the strain field at the crack tip at any given time.

9. The method for testing and analyzing the solidification performance of a soil stabilizer according to claim 1, characterized in that: The degree of strain field concentration at the crack tip includes: Digital image correlation analysis was performed on the microscopic video stream image sequence within the defined computational region to obtain the maximum principal strain field within the computational region; Calculate the average strain of all pixels with positive strain values ​​in the maximum principal strain field, and use it as an indicator of the concentration of the strain field at the crack tip.

Images