2D-DIC-based jointed rock mass slip test data dynamic retrieval analysis method

Abstract

The invention provides a jointed rock mass slip test data dynamic retrieval analysis method based on 2D-DIC, and belongs to the field of jointed rock mass ultra-low friction effect test data processing analysis, displacement field data completed through Ncorr calculation is read, the displacement field data is converted into a structured data form, and batch processing of multiple image sequences is achieved; based on the displacement field data, an effective displacement area in the block displacement field of each time step is automatically identified, geometric center coordinates of the effective displacement areas are calculated, and horizontal displacement data at the geometric center coordinates are extracted to serve as representative displacement of the block; based on displacement field data and user-defined position coordinates of pixel points of the upper part and the lower part of the joint, the position of a measuring line is set, the axial displacement difference value of the corresponding measuring point is automatically calculated, and the opening-closing amount of the normal joint is obtained.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of jointed rock mass ultra-low friction effect test data processing and analysis, and particularly relates to a digital image correlation calculation optimization method based on Ncorr open source technology, which improves the calculation efficiency and accuracy of block displacement, deformation and joint opening-closing amount through automatic ROI region management and multi-point collaborative analysis. BACKGROUND

[0002] With the development of underground engineering to deep, the instability mechanism of surrounding rock under the coupling effect of high ground stress and dynamic disturbance is increasingly complex. In the strong tectonic stress field, the kilometer deep rock mass accumulates a large amount of potential energy, and when disturbed by stress waves generated by construction disturbance or earthquakes, the jointed rock mass is prone to dynamic damage evolution and non-continuous deformation. Especially when the external dynamic load reaches a certain energy level, the large amount of potential energy accumulated in the rock mass will be released instantaneously through "quasi-resonance" motion. In this process, the dynamic stress redistribution caused by stress waves makes the normal stress between blocks present alternating tension and compression. Under the tensile stress, the blocks slip due to the weakening of friction, and under the normal stress, the blocks stop due to the increase of friction. The periodic "slip-stick" behavior of the joint surface will cause the surrounding rock blocks to produce cumulative displacement, and then cause roof falling, rib spalling and other disasters, which seriously threatens the safety of deep underground engineering.

[0003] At present, for the instability process of jointed surrounding rock under the coupling effect of dynamic and static, it is urgent to develop more accurate deformation monitoring and analysis methods to provide technical support for disaster warning and prevention. Two-dimensional digital image correlation technology (2D-DIC) as a non-contact optical measurement method, by comparing the position changes of feature points before and after deformation, obtains global displacement and deformation information, and has been widely used in the field of geotechnical engineering. However, there are still obvious defects in using Ncorr technology to process ultra-low friction effect test results, such as only supporting interactive processing of single image, unable to realize batch automatic analysis of multi-image sequence, low data processing efficiency under high sampling rate conditions, and difficult to meet the special needs of rock mass joint non-continuous deformation monitoring, unable to automatically track the relative displacement changes of key points (such as the two sides of the joint). Therefore, developing a high-efficiency DIC data processing method which can realize batch processing of multi-image sequence and automatically track the displacement of key measuring points has important research value for improving the efficiency of ultra-low friction effect test research and accurately revealing the instability mechanism of jointed rock mass. SUMMARY

[0004] In order to solve the above technical problems, the application provides a jointed rock mass slip test data dynamic retrieval analysis method based on 2D-DIC, which comprises: Step S1: read the displacement field data calculated by Ncorr and convert the displacement field data into a structured data form to realize batch processing of multiple image sequences; Step S2: based on the displacement field data, automatically identify the effective displacement area in each time step block displacement field, calculate the geometric center coordinates of the effective displacement area, and extract the horizontal displacement data at the geometric center coordinates as the representative displacement of the block; Step S3: based on the displacement field data and the pixel position coordinates of the upper and lower parts of the joint defined by the user, set the position of the measuring line, and automatically calculate the axial displacement difference of the corresponding measuring points to obtain the normal joint opening-closing amount.

[0005] Further, in step S2, an adaptive threshold algorithm is used to identify the effective data matrix in the displacement field, to suppress noise for the original picture area in Ncorr analysis, to exclude low signal-to-noise ratio areas in the displacement field that do not reach the threshold, to record areas that reach the threshold as effective displacement areas, and to retrieve the geometric center point coordinates of the effective displacement areas as horizontal displacement representation points.

[0006] Further, step S3 includes: S3.1: In the ROI segmentation line setting unit, input the number of longitudinal segmentation lines n, the system verifies the validity of the input, and maps the segmentation line position from the local coordinate system to the global coordinate system; S3.2: Perform upper joint opening-closing amount analysis, specify the x-coordinate of the upper joint boundary feature point by the user, the system automatically extracts the displacement along each segmentation line, and calculates the displacement difference of adjacent boundary points to generate the instantaneous opening-closing amount data set of the upper joint; S3.3: Perform lower joint opening-closing amount analysis, use the same architecture as step S3.2, and synchronously generate the instantaneous opening-closing amount data set of the lower joint; S3.4: Based on the data sets generated in steps S3.2 and S3.3, draw a curve graph of the normal joint opening-closing amount changing with time.

[0007] Further, based on the horizontal displacement data and the joint opening-closing amount data, perform super-low friction dynamic characteristic evaluation and data screening to identify data that conforms to the stepped displacement and periodic opening-closing law.

[0008] Further, the super-low friction dynamic characteristic evaluation and data screening method includes: Apply a double-periodic sinusoidal wave disturbance to the block system, and use a super-high-speed camera to synchronously observe the horizontal displacement and normal joint opening-closing data of the block; establish a horizontal displacement idealization model, use a unit transition function to construct a stepped change model for the horizontal displacement data, and use a Fourier series model to simulate the periodic change of the joint opening-closing.

[0009] Further, the horizontal displacement idealization model is established as follows: ; where d ideal (t) represents the idealized horizontal displacement of the key block at time t; k represents the two instantaneous slip behaviors under the action of two periodic sinusoidal stress waves; t k is the time when the kth instantaneous slip occurs; H(t) is a unit jump function; is the jump amplitude.

[0010] Further, an idealized model of joint opening and closing is established: ; where, represents the joint opening amount at time t, represents the initial opening amount of the joint, A m represents the amplitude of the mth harmonic wave, m represents the order of the harmonic, represents the angular frequency of the input sinusoidal wave, and t represents time.

[0011] Further, a CNN network is constructed, the preprocessed test data is input into the CNN network, the dynamic characteristics of the data are extracted, and the probability y that the data conforms to the expected friction dynamic characteristic law is output; if y is greater than a threshold value, it is determined that the data conforms to the law and is stored; otherwise, it is marked as abnormal and is rejected.

[0012] Compared with the prior art, the present application has the following beneficial effects: (1) The present application belongs to the field of experimental data analysis, and specifically relates to a system for analyzing the dynamic behavior of ultra-low friction effect experiments, especially suitable for analyzing the displacement and joint opening-closing amount results of joint rock mass low friction experiments. By developing a batch processing algorithm for multiple image sequences based on Ncorr, the technical bottleneck of traditional DIC technology that cannot realize automatic tracking of multiple measurement points of joint rock mass discontinuous deformation is overcome. An adaptive threshold algorithm is used to identify the effective area of the displacement field, effectively improving the positioning accuracy of ROI.

[0013] (2) A test-analysis closed-loop system is constructed by using a multi-measurement point system analysis architecture based on Matlab GUI, a key block horizontal displacement and joint opening-closing amount double-panel interactive algorithm, and a CNN-driven dynamic characteristic evaluation model. The horizontal displacement step feature and joint waveform phase difference are analyzed while searching for displacement data, overcoming the limitations of traditional analysis methods such as long data processing time, high interference, and poor predictability of experimental results. The system can flexibly process different joint rock mass sliding displacement characteristics and joint opening-closing laws, providing controllable analysis conditions for studying the instability law of key blocks in underground space under stress wave disturbance.

[0014] (3) The dynamic evaluation model is trained by using a convolutional neural network. The evaluation model is based on actual test parameters, and realizes the layer-by-layer extraction of local waveform features, medium-term trends and global laws by establishing a time-history correlation analysis model, defining a stair-shaped displacement idealization criterion, constructing a joint fluctuation amplitude and phase evaluation index, and using a 1x32, 1x16 and 1x8 multi-scale convolution kernel cascade design. The model innovatively defines the stair-shaped displacement idealization criterion and the joint fluctuation amplitude and phase evaluation index, and is integrated with a Bayesian optimization algorithm, so that the model parameters can be modified and optimized according to actual test conditions such as sampling rate and stress conditions, thereby providing an intelligent analysis and evaluation tool for deep rock mass dynamic response research. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 Flowchart for DIC to obtain displacement field data; Figure 2 Flowchart for unit interlocking; Figure 3 Schematic diagram of CNN network structure; Figure 4 Time-history curve of horizontal slip-joint deformation of key block. DETAILED DESCRIPTION

[0016] The application will be described in detail below with specific embodiments. The following examples will help those skilled in the art to further understand the application, but do not limit the application in any form. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the application. These all belong to the protection scope of the application.

[0017] Example 1 The example provides a joint rock mass multi-point data batch retrieval and processing system based on Ncorr, which adopts the following technical scheme: A joint rock mass multi-point data batch retrieval and processing system based on Ncorr includes a data input module, a batch analysis and processing module, and a data output module. The displacement field data output by Ncorr is batch processed by an automatic Matlab program to realize the rapid extraction and structured storage of ROI region measurement point data of multi-image sequences. The measurement point coordinates matrix is used to accurately locate each monitoring position, especially to accurately match the corresponding measurement point positions in the upper and lower regions of the joint, so as to realize the batch acquisition of horizontal displacement data. Then, the axial displacement of the measurement points near the upper and lower joints is subjected to real-time difference operation to accurately calculate the joint opening-closing amount, including: First, the displacement field data calculated by Ncorr is read and converted into a matrix form for subsequent processing; For the characterization of block horizontal displacement data, the algorithm automatically identifies the non-zero area of the displacement matrix at each time step, calculates its geometric center coordinates, and extracts the horizontal displacement value of the point as the representative displacement of the entire block; For the characterization of block normal joint opening-closing data, the system allows users to define the pixel position coordinates of the upper and lower parts of the joint according to the delineated ROI area, and flexibly adjusts the position of the measuring line between the joints by modifying the x and y parameters. Then the axial displacement difference of the corresponding position measuring point is automatically calculated.

[0018] Example 2 The present embodiment provides a multi-measuring point collaborative analysis method for analyzing ultra-low friction effect horizontal displacement and normal joint opening-closing, mainly by constructing a Matlab human-computer interaction interface (GUI), realizing automatic retrieval of test data and generation of preliminary horizontal displacement time history graph and normal joint opening-closing time history graph. The GUI is mainly composed of three parts, namely time module, key block horizontal displacement analysis module and normal joint opening-closing analysis module.

[0019] The system obtains the preliminary analysis results of Ncorr data through the preloaded handles_ncorr.data_dic; The system automatically obtains the total number of frames, inputs the sampling rate of the high-speed camera in the time module interface, divides the total number of frames by the sampling rate to obtain the analysis time, and stores the result to the system variable; Preferably, by calling the API interface of Ncorr, the horizontal displacement field data matrix generated by DIC analysis is automatically loaded. Enter the horizontal displacement analysis module, represent the horizontal displacement of the test piece by automatically extracting the ROI center point coordinates and corresponding displacement data, and the operation steps are as shown in Figure 1 .

[0020] The process of using the horizontal displacement analysis module is as follows: 1. Adaptive threshold algorithm is used to identify the effective data matrix (ROI) in the displacement field. The user manually draws the ROI area through Ncorr or loads the ROI area drawn by PS software; the high-speed camera in Ncorr will be loaded in the form of data matrix. Among them, the ROI area is considered as the area of interest (i.e. data to be processed), and the data outside the ROI area is shown as 0, i.e. not interested in the area (data to be processed). The core of the algorithm is to dynamically determine the binary threshold value through local pixel intensity analysis: ; wherein, μ(x, y) and σ(x, y) represent the mean and standard deviation of the displacement values in the local neighborhood centered at pixel point (x, y), respectively, and C is an empirical constant (preferably, C = 1.5). For the original picture area in the Ncorr analysis, first, noise suppression is performed, low signal-to-noise ratio areas in which |d(t)| < T(x, y) are excluded, areas that satisfy continuous d(t) ≥ T(x, y) are marked as effective displacement areas, the image is segmented to find the best ROI area, and the geometric center point coordinates of the effective area are retrieved as the horizontal displacement representation points. The above operations are realized by clicking the “Search” button in the horizontal displacement analysis module. At the same time, the system outputs all the pixel point displacement matrices in the x direction and the y direction in the ROI area in the work area; 2. On the basis of the obtained time data and the displacement matrix in the obtained ROI area, the “Plot” button in the module is clicked, and the system will automatically draw the horizontal displacement time history curve of the key block; 3. The system automatically extracts and stores the horizontal displacement data to the global variable area, providing data support for subsequent fine processing of the time history curve; Preferably, by calling the API interface of Ncorr, the axial displacement field data matrix generated by DIC analysis, that is, the full ROI data displacement field matrix in the y direction, is automatically loaded. The normal joint opening-closing analysis module is entered; the system adopts a double-panel interactive design, with the left panel being a parameter configuration area and the right panel being a data display area. The left panel adopts a three-level progressive layout, with the ROI segmentation parameter setting unit, the upper joint analysis unit, and the lower joint analysis unit in sequence, and the right panel adopts a two-level progressive layout, with the calculation result display unit and the time history curve display unit in sequence. A logical interlocking mechanism is set between the units to ensure the operation sequence. The specific operation process is as shown in Figure 1 In the ultra-low friction effect test, there is great randomness in analyzing only the axial displacement difference data of a single point, and it is necessary to retrieve the data of multiple points at the same x coordinate and different y coordinates, subtract the data, and then average the calculation to obtain the normal opening-closing amount of the joint of the entire block; Preferably, the process of using the normal joint opening-closing analysis module is as follows: The ROI segmentation line setting unit of the left panel needs to manually input the number n of longitudinal segmentation lines (n is a positive integer), and the system verifies the validity of the input through the built-in boundary condition detection module. When the coordinate search function is triggered, the system will perform the following processing: (1) Based on the preset spatial transformation parameters, the position of the segmentation line in the ROI local coordinate system is mapped to the global coordinate system; (2) A coordinate distribution graph with a scale is generated on the right panel, and the accurate position coordinates of each segmentation line in the global coordinate system are displayed in a list; (3) automatically store the mapping parameters to the system register for subsequent joint analysis calls; Upper joint opening-closing amount analysis. The user needs to specify the x-coordinate of the upper joint boundary feature point, and the y-coordinate has been obtained. Along each partition line, the displacement of the specified feature point is extracted, and the displacement difference between the adjacent boundary points of the upper joint is calculated as the instantaneous opening-closing amount. At the same time, the calculation results are saved as a structured dataset with spatial identifiers, where each data unit includes a time series, corresponding partition line global coordinates, and opening amount amplitude.

[0021] Lower joint opening-closing amount analysis. The same architecture as the upper joint analysis is used, and the system synchronously generates the lower joint opening amount dataset and automatically performs data consistency verification to ensure that the time base of the upper and lower joint analysis is aligned.

[0022] Click the plot button in the joint opening-closing analysis interface to draw the normal joint opening-closing graph. The main view displays the joint opening-closing amount curve over time, and the auxiliary view superimposes the dispersion distribution of each partition line position. This interface will display a total of n main and auxiliary views.

[0023] Preferably, the ROI is divided into n-1 regions by n vertical straight lines. Due to the limitation that the block surface contact is not completely balanced, the joint opening-closing amount corresponding to a single straight line cannot reasonably represent the joint opening-closing amount, therefore, it is necessary to select regular data for averaging calculation, that is, to average the joint opening-closing amount data corresponding to the n straight lines to represent the normal opening-closing amount of the upper and lower joints.

[0024] It should be noted that after setting the parameters of each module, the Finish button must be clicked to perform the next operation.

[0025] Example 2 The present embodiment provides a machine learning-based ultra-low friction effect dynamic characteristic evaluation method.

[0026] Taking the ultra-low friction effect experiment under double-periodic sinusoidal disturbance as an example.

[0027] ; where P(t) represents the double-periodic sinusoidal stress waveform function. The sinusoidal stress amplitude A is set to 160 N, the frequency f is set to 5 Hz, the period number T is set to 2, and t represents the time interval.

[0028] There is a clear rule between the horizontal displacement of the block and the dynamic response of the joint opening-closing under the above stress wave: the horizontal displacement changes in a double-step shape, and the joint opening-closing changes in a double-periodic sinusoidal wave. However, due to environmental disturbances, sensor noise, and other factors, a large amount of actual collected data contains abnormal samples that do not conform to the rules.

[0029] CNN (Convolutional Neural Network), with its excellent local feature extraction capabilities and strong robustness to noise, can accurately identify complex spatiotemporal coupling patterns such as stepped displacement and periodic character law opening, and realize automated, high-precision intelligent analysis of massive, high-noise DIC data in jointed rock mass tests, solving the problems of low efficiency and strong subjectivity of traditional methods.

[0030] Preferably, the ultra-low friction dynamic characteristic evaluation and data screening method based on convolutional neural networks (CNN) is suitable for the analysis of block friction response under dual-cycle stress wave disturbance experiments. This method automatically screens experimental data that conform to expected patterns through waveform matching degree evaluation and dynamic feature extraction, and quantitatively evaluates the dynamic characteristics (stability, response speed, and anti-interference ability) of the friction system.

[0031] Two cycles of stress wave disturbance were applied to the block system, and the horizontal displacement and normal joint opening-closing data of the key blocks were observed using an ultra-high-speed camera. Establish an idealized model for horizontal displacement: ; Where, d ideal (t) represents the idealized horizontal displacement of the critical block within time t; k represents the two instantaneous slip behaviors occurring under the action of two-period sinusoidal stress waves; t k H(t) represents the moment of the k-th instantaneous slip; H(t) is the unit transition function used to simulate instantaneous slip behavior and describe the displacement at time t. k A d that occurs at any moment k The leap; t k =(2k-1)T0 / 4 (step time), where T0 represents the basic period of the input stress wave; Meets 5mm < <25mm.

[0032] In underground engineering, under dual-cycle sinusoidal wave perturbation, critical blocks will exhibit alternating slip-stagnation behavior. At the end of each stress wave half-cycle, the normal stress on the joint surface drops to its minimum, and the frictional force decreases sharply, causing instantaneous slippage of the block, with the displacement exhibiting a step-like jump. High-speed cameras capture the entire process of stress wave perturbation, showing that the slippage event occurs at a specific phase of each stress wave cycle (usually after a trough or crest), and the transition amplitude... Within the range of 0.15~0.25mm, it is related to parameters such as stress wave amplitude and frequency. The above formula is used to predict the motion law of the critical block under dual-cycle sinusoidal stress wave disturbance, that is, the ideal horizontal displacement of the critical block is that instantaneous slip occurs at two different times, and the distance of each slip is within a reasonable range. First, CNN is used to determine whether the actual sliding displacement time history curve is close to the ideal "stepped" shape, and... Fall within a specified range.

[0033] Based on Fourier transform, the input sinusoidal wave is equivalent to the form of fundamental wave and harmonic wave, the unit transition function H(t) simulates the transient slip behavior, which is more suitable for the discretization of joint slip and nonlinear characteristics.

[0034] Establish the idealized model of joint opening and closing: m ; Among them, The joint opening amount at time t is represented by A The initial opening amount of the joint is represented by A m The amplitude of the mth harmonic wave is represented by A The angular frequency of the input sinusoidal wave is represented by ω, and t represents time. Under the condition of double-period stress wave, the joint normal stress will appear as a double-sine wave, and the joint deformation (opening-closing) will appear as a periodic motion.

[0035] The key to the key block super-low friction experiment is to use a super-high-speed camera to obtain the interaction between the deformation (opening-closing) and the horizontal slip of the joint between the blocks. Therefore, the purpose of using CNN to create two idealized models is to obtain the idealized shape and approximate range of the block slip curve and the opening-closing curve in the actual experimental data. For slip, it is a step type, and for the joint deformation curve, it is a double-sine wave type. The purpose is to use this idealized model to remove the working conditions that do not conform to the shape and the data range is not within the idealized range after obtaining multiple joint deformation curves, thereby improving the efficiency of data processing. The relationship between the horizontal displacement idealized model and the joint opening and closing idealized model is parallel, and is used to obtain two different results.

[0036] Convolutional neural network CNN construction: The network input layer receives 10×10×5 three-dimensional data (500 data points are reshaped into 10×10 grid×5 channels), first extracts local spatial features through 16 3×3 convolution kernels (step 1, ReLU activation), and reduces dimension to 8×8×16 through 2×2 maximum pooling; Then use 32 3×3 convolution kernels (step 1, ReLU activation) to capture the middle layer spatial pattern, and 2×2 maximum pooling again to get 4×4×32 feature maps; Then flatten it into a 512-dimensional vector, pass it through a fully connected layer (ReLU activation) with 32 neurons for feature integration, and finally generate prediction probabilities through the output layer activated by sigmoid. This architecture effectively processes grid-based time series data through spatial convolution, with a global parameter size of about 5.8k, suitable for small-scale classification tasks. If you need to process spatiotemporal correlation, you can replace the 2D convolution with a 3×3×3 3D convolution kernel.

[0037] Figure 3 The CNN network structure is a schematic diagram, and the network extracts spatio-temporal features layer by layer through two levels of 3x3 convolution-pooling layers: the first layer of convolution (16 kernels) captures the local association of displacement step mutation and joint sinusoidal wave, and the second layer of convolution (32 kernels) identifies the global pattern of double-periodic cooperative change; after the integration of the fully connected layer, the sigmoid output layer quantifies the waveform matching probability, realizing the filtering of abnormal data.

[0038] The original signal is gridded into a multi-channel input, and the translational invariance of the convolution kernel is used to automatically learn the friction response law under stress wave disturbance (such as the step-sine correspondence). Compared with the traditional threshold method, the robustness to noise interference is significantly improved.

[0039] 500 groups of artificially labeled experimental data (250 groups of "conform" and 250 groups of "non-conform") are collected and mixed with enhanced samples to form a training set; and mixed with enhanced samples generated by slight perturbation (amplitude ±5%, phase ±10°); the model training uses binary cross-entropy loss function, combined with Adam optimizer (learning rate set to 0.001, batch size to 32, and iteration number to 50). The final verification process uses Accuracy ( ), Precision ( ) and Recall ( ) as evaluation indexes, requiring the accuracy of the verification set to be not less than 80%, the precision to be not less than 80% (where TP is true positive, TN is true negative, FP is false positive, and FN is false negative), to ensure the reliability of the model screening.

[0040] The retrieval results in the ultra-low friction effect test-analysis closed loop method are preprocessed, and the processed results are stored as a data matrix to generate dual-channel input data; the data are input into the trained CNN model to obtain the conformity probability y. The screening threshold is set to 0.8, if y>0.8, it is determined as "conform to the rule data", and automatically stored in the database with the confidence recorded; if y≤0.8, it is determined as "not conform to the rule data", and the abnormal type (such as "ladder feature missing" and "sinusoidal wave period abnormality") is marked and eliminated.

[0041] Figure 4 The key block horizontal slip-joint opening and closing time history curve, Figure 4 The key block horizontal slip-joint opening and closing time history curve, Figure 4 The key block horizontal slip-joint opening and closing time history curve, Figure 4 The key block horizontal slip-joint opening and closing time history curve, Figure 4The middle (d) shows the time history curve of the upper and lower joint opening and closing under the stress wave disturbance with a frequency of 5 Hz and an amplitude of 200 N. Under the condition that the test is good (without rotation and offset), the sliding displacement curve selected by the system can reflect the sliding law of the block under the two-period stress wave disturbance, and the normal joint opening and closing amount can correspond to the sliding displacement characteristics.

[0042] The specific embodiments of the present application are described above. It needs to be understood that the present application is not limited to the specific embodiments described above, and various changes or modifications can be made by those skilled in the art within the scope of the claims, which does not affect the essential content of the present application. The embodiments of the present application and the features in the embodiments can be arbitrarily combined with each other without conflict.

Claims

1. A dynamic retrieval and analysis method for jointed rock mass sliding test data based on 2D-DIC, characterized in that, include: Step S1: Read the displacement field data calculated by Ncorr and convert it into structured data to achieve batch processing of multiple image sequences; Step S2: Based on the displacement field data, automatically identify the effective displacement region in the block displacement field at each time step, calculate the geometric center coordinates of the effective displacement region, and extract the horizontal displacement data at the geometric center coordinates as the representative displacement of the block; Step S3: Based on the displacement field data and the user-defined pixel position coordinates of the upper and lower parts of the joint, set the measurement line position and automatically calculate the axial displacement difference of the corresponding measurement points to obtain the normal joint opening-closing amount.

2. The method for dynamic retrieval and analysis of jointed rock mass sliding test data based on 2D-DIC according to claim 1, characterized in that, In step S2, an adaptive threshold algorithm is used to identify the effective data matrix in the displacement field. Noise suppression is performed on the original image region in the Ncorr analysis to exclude low signal-to-noise ratio regions in the displacement field that do not reach the threshold. Regions that reach the threshold are recorded as effective displacement regions, and the coordinates of the geometric center point of the effective displacement region are retrieved as horizontal displacement characterization points.

3. The method for dynamic retrieval and analysis of jointed rock mass sliding test data based on 2D-DIC according to claim 1, characterized in that, Step S3 includes: S3.1: In the ROI segmentation line setting unit, input the number of vertical segmentation lines n. The system verifies the validity of the input and maps the segmentation line positions from the local coordinate system to the global coordinate system. S3.2: Perform upper joint opening-closing analysis. The user specifies the x-coordinate of the upper joint boundary feature points. The system automatically extracts the displacement along each segmentation line and calculates the displacement difference between adjacent boundary points, generating an instantaneous opening-closing dataset of the upper joint. S3.3: Perform lower joint opening-closing analysis, using the same architecture as step S3.2, synchronously generating an instantaneous opening-closing dataset of the lower joint. S3.4: Based on the datasets generated in steps S3.2 and S3.3, plot a curve showing the change of normal joint opening-closing amount over time.

4. The method for dynamic retrieval and analysis of jointed rock mass sliding test data based on 2D-DIC according to claim 3, characterized in that, Based on horizontal displacement data and joint opening-closing data, we conduct dynamic characteristic evaluation and data screening for ultra-low friction, and identify data that conform to stepped displacement and periodic opening-closing laws.

5. The method for dynamic retrieval and analysis of jointed rock mass sliding test data based on 2D-DIC according to claim 4, characterized in that, Methods for evaluating and screening dynamic characteristics of ultra-low friction include: A dual-period sinusoidal perturbation was applied to the block system, and the horizontal displacement and normal joint opening-closing data of the block were simultaneously observed using an ultra-high-speed camera. An idealized model of horizontal displacement was established, and a step change model was constructed using a unit transition function for the horizontal displacement data. For the joint opening and closing, a Fourier series model was used to simulate its periodic changes.

6. The method for dynamic retrieval and analysis of jointed rock mass sliding test data based on 2D-DIC according to claim 5, characterized in that, Establish an idealized model for horizontal displacement: ； Where, d ideal (t) represents the idealized horizontal displacement of the critical block within time t; k represents the two instantaneous slip behaviors occurring under the action of two-period sinusoidal stress waves; t k Let H(t) be the moment of the k-th instantaneous slip; H(t) is the unit transition function. This represents the transition range.

7. The method for dynamic retrieval and analysis of jointed rock mass sliding test data based on 2D-DIC according to claim 6, characterized in that, Establish an idealized model for joint opening and closing: ； in, This represents the joint opening at time t. A represents the initial opening of the joint. m This represents the amplitude of the m-th harmonic, where m represents the order of the harmonic. t represents the angular frequency of the input sine wave, and t represents time.

8. The method for dynamic retrieval and analysis of jointed rock mass sliding test data based on 2D-DIC according to claim 7, characterized in that, A CNN network is constructed, and the preprocessed experimental data is input into the CNN network. The dynamic features of the data are extracted, and the probability y that the data conforms to the expected friction dynamic characteristics is output. If y > threshold, it is determined to be data that conforms to the pattern and stored; otherwise, it is marked as abnormal and removed.

Images