A method and system for monitoring ultra-wide cracks in landslides

Abstract

This invention relates to the field of geological disaster monitoring and discloses a method and system for monitoring ultra-wide cracks in landslides. The method utilizes microwave ranging, MEMS, and visual sensors to collaboratively collect and preprocess crack status and external environmental data. The width, morphology, and mechanical characteristics of the preprocessed cracks are then extracted. Based on these characteristics, crack changes are analyzed within a preset time window to derive an expansion index for anomaly identification. Further analysis is triggered based on the anomaly results, inputting the characteristics into a scenario reasoning model to determine the current evolution stage and whether a sudden change has occurred. If a sudden change occurs, a correlation analysis is performed between the evolution status and the external environment to identify factors accelerating expansion and generate a causal correlation list. This list is then combined with historical data to predict crack evolution trends, obtaining prediction coefficients. Finally, based on the prediction results, a warning alert is triggered. This method achieves precise monitoring of crack status, anomaly identification, scientific prediction of evolution, and effective prevention and control of safety risks.

Description

Technical Field

[0001] This invention relates to the field of geological disaster monitoring technology, and more specifically to a method and system for monitoring ultra-wide cracks in landslides. Background Technology

[0002] Monitoring ultra-wide cracks in landslides is a crucial aspect of geological disaster prevention and control. Traditional techniques mainly rely on manual inspections and simple measuring tools, such as the stake method, nail method, painting method, and patch method. While these methods are simple to operate and inexpensive, they have limitations such as insufficient accuracy, poor real-time performance, and high labor intensity, making them unsuitable for monitoring needs under complex geological conditions. Existing technologies, on the other hand, utilize sensors, remote sensing, GPS, and automated systems to achieve real-time, high-precision acquisition and dynamic analysis of parameters such as crack width, depth, and tilt angle. For example, the application of equipment such as laser rangefinders, total stations, InSAR technology, and crack measurement devices has significantly improved monitoring efficiency and data reliability. Furthermore, by combining environmental data such as groundwater and rainfall, more comprehensive scientific evidence is provided for predicting crack evolution trends and issuing landslide early warnings.

[0003] However, existing technologies still have the following drawbacks:

[0004] Firstly, existing technologies rely on a single sensor (such as a laser rangefinder or manual measurement), resulting in limited data dimensions and difficulty in comprehensively reflecting the crack state; although some methods use multiple sensors, they lack a coordination mechanism, leading to poor data synchronization.

[0005] Secondly, existing technologies only extract single features such as crack width or length, ignoring morphological features (such as crack branching and propagation direction) and mechanical features (such as stress changes and vibration modes), resulting in insufficient understanding of crack evolution mechanisms. Furthermore, crack evolution judgment mainly relies on human experience or simple rules to determine the stage of crack evolution, lacking quantitative models, which leads to subjective and lagging judgments.

[0006] Third, existing technologies simply correlate crack propagation with single environmental factors such as rainfall or earthquakes, ignoring the synergistic effects of multiple factors and time delays. The analysis results are one-sided, and the evolution prediction relies on fitting curves to historical data. The lack of dynamic correction of the current crack state during evolution prediction leads to a large deviation between the prediction results and the actual situation. Summary of the Invention

[0007] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method and system for monitoring ultra-wide cracks in landslides, so as to solve the problems existing in the background art.

[0008] This invention provides the following technical solution: a method for monitoring ultra-wide cracks in landslides, comprising:

[0009] S1: Crack status data is collected collaboratively by microwave ranging sensor, MEMS sensor and vision sensor, and external environment data is acquired at the same time. The crack status data and external environment data are preprocessed.

[0010] S2: Extract features from the preprocessed crack state data to obtain crack width features, crack morphology features, and crack mechanical features;

[0011] S3: Based on the crack width characteristics, crack morphology characteristics, and crack mechanical characteristics, the crack change state is analyzed within a preset time window to obtain the crack propagation index, thereby identifying the abnormal state of the crack.

[0012] S4: Based on the identification results of the crack abnormal state, trigger further analysis of crack evolution, input crack width characteristics, crack morphology characteristics and crack mechanical characteristics into the preset scenario reasoning model, obtain the current evolution stage of the crack, and determine whether the current evolution stage of the crack has undergone a sudden change.

[0013] S5: Based on the judgment results of abrupt changes in the evolutionary stage, combine the current evolutionary state of the crack with external environmental data, conduct environmental correlation analysis on the crack state, identify factors that accelerate crack propagation, and obtain a causal correlation list.

[0014] S6: Combine the causal relationship list with historical crack evolution data to predict crack evolution trends and obtain evolution prediction coefficients;

[0015] S7: Based on the predicted results of the crack evolution state, determine whether to trigger an early warning prompt for the crack evolution state.

[0016] Preferably, the process of acquiring the S1 crack state data includes:

[0017] Simultaneously activate the microwave ranging sensor and the vision sensor;

[0018] The distance to the other side of the crack is obtained by performing multiple distance measurements using a microwave sensor, and microwave distance measurement data is obtained. Multiple high-resolution visual images of the crack are captured by a visual sensor, and MEMS data is continuously recorded by a MEMS sensor during the acquisition process.

[0019] The acquisition of the external environment data includes: synchronously acquiring environmental data around the crack through an external interface;

[0020] Preprocessing of crack status data and external environment data includes: data synchronization, data alignment, data noise reduction, and data correction.

[0021] Preferably, step S2, which involves feature extraction from the preprocessed crack state data, includes:

[0022] The crack width, the average crack width within a preset time window, the crack width change rate within a preset time window, the maximum crack width, and the minimum crack width are extracted from the preprocessed microwave ranging data as crack width features.

[0023] The crack length, maximum crack length, minimum crack length, crack length change rate within a preset time window, crack area, and crack trunk propagation direction are extracted from the preprocessed crack visual image as crack morphology features.

[0024] The cumulative tilt angle, equivalent stress intensity factor, rate of change of equivalent stress intensity factor, tilt angle rate, and vibration spectrum characteristics are extracted from the preprocessed MEMS data as crack mechanical characteristics. The equivalent stress intensity factor is estimated by fusing and analyzing the cumulative tilt angle, tilt angle rate, and vibration spectrum characteristics in the MEMS data and based on a preset geomechanical model or data-driven model.

[0025] Preferably, the analysis of the crack change state in step S3 includes:

[0026] Based on the extracted crack width change rate and crack length change rate, and construct their time series respectively, calculate their first time derivatives based on the crack width change rate time series and crack length change rate time series, construct dynamic acceleration factors, and obtain width acceleration factor and length acceleration factor. The dynamic acceleration factor is used to characterize the degree of change rate surge.

[0027] The crack propagation index is calculated by weighting and fusing the crack width, crack length, crack width change rate, crack length change rate, crack area, equivalent stress intensity factor, equivalent stress intensity factor change rate, width acceleration factor, and length acceleration factor using preset weighting coefficients.

[0028] The crack propagation index is compared with a preset multi-level safety threshold to determine whether the crack is in an abnormal state. If the crack propagation index is less than or equal to the preset safety threshold, the crack is determined to be in a normal state, and the crack state continues to be monitored and identified for abnormalities. If the crack propagation index is greater than the preset safety threshold, the crack is determined to be in an abnormal state, and the crack evolution analysis process is triggered and executed.

[0029] Preferably, step S4 inputs the real-time collected crack width features, crack morphology features, and crack mechanical features into a preset scenario reasoning model to determine the current evolution stage of the crack. The scenario reasoning model has multiple preset discrete evolution stages corresponding to the landslide evolution process. The discrete evolution stages include the initial micro-expansion stage, the stable expansion stage, the accelerated expansion stage, and the near-landslide instability stage. Each evolution stage is defined by a multi-dimensional feature vector space. The multi-dimensional feature vector space consists of the feature thresholds and probability distributions of crack width, crack length, crack width change rate, crack length change rate, and equivalent stress intensity factor.

[0030] The scenario reasoning model calculates the matching probability between the current input features and the feature vector space of each evolution stage, and uses Bayesian networks or evidence theory algorithms to update and output the confidence level of the crack in each evolution stage.

[0031] By comparing the matching probabilities of each evolution stage, the evolution stage with the highest matching probability is identified as the current evolution stage of the crack, and the confidence level of that evolution stage is output.

[0032] The system continuously records the currently identified evolutionary stage and compares it with the evolutionary stage at the previous moment. If the current evolutionary stage is different from the previous evolutionary stage and the confidence level of the current evolutionary stage is greater than the preset mutation threshold, it is determined that a mutation has occurred in the crack evolutionary stage, and a mutation signal is generated to trigger the environmental correlation analysis of S5.

[0033] Preferably, when the mutation signal is received, step S5 triggers and executes a time-series lead-lag correlation analysis algorithm to identify the main environmental factors inducing crack evolution and obtain the causal strength coefficient. Specific analysis steps include:

[0034] Step 1: Taking the mutation point of the evolution stage as the center, extract a preset analysis time window that includes the time series of crack propagation index and the time series of various external environmental data;

[0035] Step 2: For each external environment data time series, calculate its cross-correlation function with the crack propagation index time series to analyze the correlation between the two under different time delays;

[0036] Step 3: For each cross-correlation function, locate the peak of its cross-correlation function and extract the maximum correlation coefficient and time delay corresponding to the peak;

[0037] Step 4: For each external environmental factor, determine whether the time delay corresponding to the external environmental factor is positive and whether the value of the time delay is within the preset physical reasonable time interval. At the same time, determine whether the maximum correlation coefficient corresponding to the external environmental factor is greater than the preset significance threshold.

[0038] Step 5: If an environmental factor satisfies both the time rationality judgment and the correlation significance judgment, then the environmental factor is identified as the cause of this crack evolution, and the maximum correlation coefficient is used as the causal strength coefficient of the cause for quantitative output.

[0039] Step 6: Output a causal relationship list of all environmental factors identified as triggers. The causal relationship list includes all identified triggers, the causal strength coefficient of each trigger, and the time delay.

[0040] If no confirmed cause exists, the sudden change in the crack evolution is determined to be an event with unknown external cause, triggering an internal instability assessment process. This involves increasing the frequency of internal state data collection and conducting in-depth analysis of the crack's vibration spectrum, stress intensity factor, and acceleration factor to identify any precursory features of internal instability. If a clear precursory feature of internal instability is identified, a high-level warning is generated and reported. If no clear precursory feature is identified, a low-level warning with an unknown cause requiring manual intervention is generated and reported.

[0041] Preferably, in step S6, the causal association list output by step S5 is used as a retrieval condition to match and retrieve historical crack evolution cases from the historical database. The matching process quantifies similarity by calculating the similarity between the current causal association list and the causal association lists of each historical case. The similarity is calculated based on the comprehensive difference of the cause type, causal intensity coefficient and time delay, and historical cases with similarity higher than a preset similarity threshold are selected as the effective case set.

[0042] Extract the evolution trajectory data of each case in the effective case set within a preset time period after the mutation, and calculate the statistical distribution characteristics of crack evolution rate, crack evolution acceleration and final instability time.

[0043] The statistical distribution characteristics are weighted and fused with the real-time evolution data of the current crack after the mutation. Then, the evolution trajectory of the current crack is corrected by using the fused data through a preset prediction model. Finally, the evolution prediction coefficient is output. The evolution prediction coefficient is used to predict the probability that the crack will enter the next more dangerous evolution stage within a preset time window in the future.

[0044] Preferably, step S7 compares the output evolution prediction coefficient with a preset evolution threshold. If the evolution prediction coefficient exceeds the preset evolution threshold, it is determined that there is a safety risk in the crack evolution state, and an early warning feedback mechanism is triggered to issue an early warning to the monitoring terminal. At the same time, an early warning monitoring report is automatically generated and uploaded to the monitoring terminal. Otherwise, the crack state data continues to be monitored.

[0045] To achieve the above objectives, the present invention provides the following technical solution: a monitoring system for ultra-wide cracks in landslides, comprising the implementation of the aforementioned method for monitoring ultra-wide cracks in landslides, including:

[0046] Data acquisition module: Collects crack status data in collaboration with microwave ranging sensor, MEMS sensor and vision sensor, and simultaneously acquires external environment data, and preprocesses crack status data and external environment data;

[0047] Feature extraction module: Extracts features from the preprocessed crack state data to obtain crack width features, crack morphology features, and crack mechanical features;

[0048] Crack anomaly identification module: Based on crack width characteristics, crack morphology characteristics, and crack mechanical characteristics, the module analyzes the crack change state within a preset time window to obtain the crack propagation index, thereby identifying the abnormal state of the crack.

[0049] Crack Evolution Analysis Module: Based on the identification results of crack anomaly states, it triggers further analysis of crack evolution, inputs crack width characteristics, crack morphology characteristics, and crack mechanical characteristics into a preset scenario reasoning model, obtains the current evolution stage of the crack, and determines whether a sudden change has occurred in the current evolution stage of the crack.

[0050] Crack Environment Correlation Analysis Module: Based on the judgment results of abrupt changes in the evolution stage, the current evolutionary state of the crack is combined with external environmental data to perform environmental correlation analysis on the crack state, identify factors that accelerate crack propagation, and obtain a causal correlation list.

[0051] Evolution Prediction Module: Combines the causal relationship list with historical crack evolution data to predict crack evolution trends and obtain evolution prediction coefficients;

[0052] Early warning feedback module: Based on the prediction results of crack evolution status, determine whether to trigger an early warning prompt for crack evolution status.

[0053] The technical effects and advantages of this invention are as follows:

[0054] (1) By using microwave ranging sensors, MEMS sensors and vision sensors to collect data in a coordinated manner, the crack width, morphology and mechanical characteristics are acquired synchronously, and external environmental data (such as rainfall and seismic waves) are fused together. In the data preprocessing stage, synchronization, alignment and noise reduction are used to ensure data quality, solve the problem of multi-source data fusion, and realize multi-dimensional and high-precision monitoring of crack status.

[0055] (2) By extracting the width and its dynamic changes from microwave data, extracting morphological features from visual images, and extracting mechanical features from MEMS data, the crack state is more comprehensively characterized through multi-dimensional feature comprehensive analysis, deepening the understanding of the crack evolution mechanism. Then, by using a pre-set scenario reasoning model, the crack evolution is divided into four stages: initial micro-expansion, stable expansion, accelerated expansion, and near-slip instability. Each stage is defined by a multi-dimensional feature vector space. By using Bayesian networks or evidence theory algorithms to calculate the matching probability between the current feature and each stage, the stage with the highest confidence is output, and the changes before and after the stages are continuously compared to realize the real-time identification of abrupt events, solving the problem of subjective and lagging judgment in traditional methods.

[0056] (3) By triggering the time-series lead-lag correlation analysis algorithm when mutations occur during the evolutionary stage, the cross-correlation function between the crack propagation index and various environmental data is calculated, and the peak correlation coefficient and time delay are located. Through dual judgment of physical rationality and significance, the main causes are identified and the causal intensity coefficient is quantified, generating a causal association list. This solves the problem of one-sided analysis results of traditional methods. Then, the current causal association list is matched with historical cases, and cases with high similarity are selected. Their post-mutation evolution trajectory data is extracted and weighted and fused with the current real-time data. The evolution trajectory is corrected through the prediction model, and the evolution prediction coefficient is output. Combining historical patterns with current dynamics improves the predictive adaptability and solves the problem of large deviations between traditional prediction methods and reality. Attached Figure Description

[0057] Figure 1 This is a diagram illustrating the method steps of the present invention.

[0058] Figure 2 This is a system structure block diagram of the present invention. Detailed Implementation

[0059] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. In addition, the forms of the various structures described in the following embodiments are merely illustrative. The method and system for monitoring ultra-wide cracks in landslides involved in the present invention are not limited to the structures described in the following embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] like Figure 1 The embodiment shown provides a method for monitoring ultra-wide cracks in landslides, including:

[0061] S1: Crack status data is collected collaboratively by microwave ranging sensors, MEMS sensors and vision sensors, and external environment data is acquired at the same time. The crack status data and external environment data are preprocessed.

[0062] In this embodiment, the process of acquiring the S1 crack state data includes:

[0063] Simultaneously activate the microwave ranging sensor and the vision sensor;

[0064] The distance to the other side of the crack is obtained by performing multiple distance measurements using a microwave sensor, and microwave distance measurement data is obtained. Multiple high-resolution visual images of the crack are captured by a visual sensor, and MEMS data is continuously recorded by a MEMS sensor during the acquisition process.

[0065] The acquisition of the external environment data includes: synchronously acquiring environmental data around the crack through an external interface;

[0066] Preprocessing of crack status data and external environment data includes: data synchronization, data alignment, data noise reduction, and data correction.

[0067] It should be noted that the specific data collection process is as follows:

[0068] First, a central control unit (such as an embedded processor) sends a synchronization trigger signal that simultaneously activates both the microwave ranging sensor and the vision sensor, ensuring strict consistency in their time references. For example, all sensor data acquisition is synchronized using a GPS pulse-per-second (PPS) signal or a high-precision crystal oscillator clock. During the data acquisition phase, the microwave ranging sensor continuously transmits and receives microwave signals at a frequency of 10Hz, performing multiple ranging measurements, for example, acquiring 10 distance measurements within 1 second to calculate the instantaneous change in crack width, forming a sequence of microwave ranging data. Simultaneously, the vision sensor captures high-resolution (e.g., 20 megapixels) images of the crack at a rate of 1 frame per second, synchronously recording the timestamp of each image. The device's built-in MEMS sensors (including a three-axis accelerometer and a three-axis gyroscope) continuously record the device's attitude and vibration data at a frequency of 100Hz, used for subsequent motion compensation of the visual images and microwave ranging data. External environmental data is obtained through the device's integrated temperature and humidity sensors, barometer, and real-time rainfall, wind speed, and other data obtained from meteorological services via an API interface. After data acquisition, data preprocessing is performed: First, all data is synchronized and aligned based on a unified timestamp. For example, 10Hz microwave data and 100Hz MEMS data are aligned to a 1Hz visual image timestamp using an interpolation algorithm to form a multi-dimensional data vector corresponding to each image frame. Second, data denoising is performed. Kalman filtering is used to remove random noise from the microwave ranging data, and median filtering or Gaussian filtering is used to remove image noise from the visual images. Finally, data correction is performed. Perspective transformation and geometric correction are performed on the visual images using the attitude data recorded by the MEMS sensor to eliminate image distortion caused by device tilt. Temperature drift compensation is performed on the microwave ranging results using temperature sensor data, thereby ensuring that the data used in subsequent analysis has high accuracy and consistency.

[0069] In this embodiment, ultra-wide cracks refer to cracks that occur when a landslide (mountain) is affected by disaster-causing factors, disrupting its original mechanical equilibrium under basic stability, and undergoing a gradual increase in displacement and severe deformation. This is a common phenomenon that occurs before a landslide causes disaster, and "ultra-wide cracks" are relative to the small micro-cracks that occur when the landslide (mountain) is in the creep stage.

[0070] S2: Extract features from the preprocessed crack state data to obtain crack width features, crack morphology features, and crack mechanical features.

[0071] In this embodiment, the feature extraction of the preprocessed crack state data by step S2 includes:

[0072] The crack width, the average crack width within a preset time window, the crack width change rate within a preset time window, the maximum crack width, and the minimum crack width are extracted from the preprocessed microwave ranging data as crack width features.

[0073] The crack length, maximum crack length, minimum crack length, crack length change rate within a preset time window, crack area, and crack trunk propagation direction are extracted from the preprocessed crack visual image as crack morphology features.

[0074] The cumulative tilt angle, equivalent stress intensity factor, rate of change of equivalent stress intensity factor, tilt angle rate, and vibration spectrum characteristics are extracted from the preprocessed MEMS data as crack mechanical characteristics. The equivalent stress intensity factor is estimated by fusing and analyzing the cumulative tilt angle, tilt angle rate, and vibration spectrum characteristics in the MEMS data and based on a preset geomechanical model or data-driven model.

[0075] It should be specifically noted that the extraction of crack width features is based on preprocessed and synchronously aligned microwave ranging data. For example, a preset time window of 24 hours is set, and the average value of all microwave ranging data within this window is calculated as the "average crack width". The "crack width change rate" is obtained by calculating the width difference between adjacent time points (e.g., hourly). At the same time, the maximum and minimum values ​​within this time window are selected as the "maximum crack width" and "minimum crack width", respectively. For the extraction of crack morphology features, the system uses computer vision algorithms to process the preprocessed crack visual image. Specifically, firstly, a deep learning-based semantic segmentation model is used to accurately segment the crack region from the image background. Then, the crack skeleton is calculated through pixel connectivity analysis, thereby quantifying the "crack length". By comparing the crack skeleton in images from multiple consecutive days, the "crack length change rate" can be calculated and the "crack main extension direction" (e.g., represented by the angle with due north) can be identified. At the same time, the crack area can be obtained by counting the total number of pixels in the segmented crack region and multiplying it by the actual physical area of ​​each pixel. The core of extracting the mechanical features of cracks is estimating the "equivalent stress intensity factor." Specifically, this involves: first, directly extracting features from MEMS data, such as calculating the "cumulative tilt angle" of the device relative to its initial state by integrating data from triaxial accelerometers and gyroscopes; obtaining the "tilt angular rate" by differentiating the tilt angle data; and extracting "vibration spectrum features" such as the dominant frequency and amplitude by performing a fast Fourier transform on the vibration data. The extracted mechanical features are then input into a pre-trained data-driven model, such as a support vector machine or neural network model, trained on a large amount of historical data (including MEMS data and actual stress intensity factors measured through finite element simulation or field stress gauges). This model learns the complex nonlinear relationship between the cumulative tilt angle, tilt angular rate, vibration spectrum features, and the actual stress intensity factor, ultimately outputting an estimated value, the "equivalent stress intensity factor." Simultaneously, the "rate of change of the equivalent stress intensity factor" can be obtained by subtracting the equivalent stress intensity factors at consecutive time points.

[0076] S3: Based on the crack width characteristics, crack morphology characteristics, and crack mechanical characteristics, the crack change state is analyzed within a preset time window to obtain the crack propagation index, thereby identifying the abnormal state of the crack.

[0077] In this embodiment, the analysis of the crack change state by S3 includes:

[0078] Based on the extracted crack width change rate and crack length change rate, and construct their time series respectively, calculate their first time derivatives based on the crack width change rate time series and crack length change rate time series, construct dynamic acceleration factors, and obtain width acceleration factor and length acceleration factor. The dynamic acceleration factor is used to characterize the degree of change rate surge.

[0079] The crack propagation index is calculated by weighting and fusing the crack width, crack length, crack width change rate, crack length change rate, crack area, equivalent stress intensity factor, equivalent stress intensity factor change rate, width acceleration factor, and length acceleration factor using preset weighting coefficients.

[0080] The crack propagation index is compared with a preset multi-level safety threshold to determine whether the crack is in an abnormal state. If the crack propagation index is less than or equal to the preset safety threshold, the crack is determined to be in a normal state, and the crack state continues to be monitored and identified for abnormalities. If the crack propagation index is greater than the preset safety threshold, the crack is determined to be in an abnormal state, and the crack evolution analysis process is triggered and executed.

[0081] It should be specifically noted that time series are constructed based on the "crack width change rate" and "crack length change rate" extracted from S2. For example, by setting a time window in hours, a set of width change rate values ​​arranged in chronological order is obtained. and the rate of change of length To characterize the abrupt increase in the rate of change, the first-order time derivatives of these two time series are calculated, known as the "dynamic acceleration factor." Specifically, the central difference method can be used for calculation, for example, the width acceleration factor. ,in The time interval is used. Then, the crack propagation index is calculated using weighted fusion, with the specific calculation formula being: Where W is the current crack width, L is the current crack length, A is the crack area, and K is the equivalent stress intensity factor. The rate of change of the equivalent stress intensity factor. and These are the width and length acceleration factors, respectively. These are weighting coefficients, the specific values ​​of which are determined based on expert experience, historical data statistical analysis, or machine learning algorithms (such as the analytic hierarchy process or random forest feature importance assessment). Finally, the calculated crack propagation index is compared with a preset safety threshold to determine the status.

[0082] S4: Based on the identification results of the crack abnormal state, trigger further analysis of crack evolution, input crack width characteristics, crack morphology characteristics and crack mechanical characteristics into the preset scenario reasoning model, obtain the current evolution stage of the crack, and determine whether the current evolution stage of the crack has undergone a sudden change.

[0083] In this embodiment, step S4 inputs the real-time collected crack width features, crack morphology features, and crack mechanical features into a preset scenario reasoning model to determine the current evolution stage of the crack. The scenario reasoning model has multiple preset discrete evolution stages corresponding to the landslide evolution process. The discrete evolution stages include the initial micro-expansion stage, the stable expansion stage, the accelerated expansion stage, and the near-landslide instability stage. Each evolution stage is defined by a multi-dimensional feature vector space. The multi-dimensional feature vector space consists of the feature thresholds and probability distributions of crack width, crack length, crack width change rate, crack length change rate, and equivalent stress intensity factor.

[0084] The scenario reasoning model calculates the matching probability between the current input features and the feature vector space of each evolution stage, and uses Bayesian networks or evidence theory algorithms to update and output the confidence level of the crack in each evolution stage.

[0085] By comparing the matching probabilities of each evolution stage, the evolution stage with the highest matching probability is identified as the current evolution stage of the crack, and the confidence level of that evolution stage is output.

[0086] The system continuously records the currently identified evolutionary stage and compares it with the evolutionary stage at the previous moment. If the current evolutionary stage is different from the previous evolutionary stage and the confidence level of the current evolutionary stage is greater than the preset mutation threshold, it is determined that a mutation has occurred in the crack evolutionary stage, and a mutation signal is generated to trigger the environmental correlation analysis of S5.

[0087] It should be specifically explained that a pre-defined scenario reasoning model transforms the identification results of abnormal states into precise localization of the crack evolution process. This scenario reasoning model internally predefines four discrete evolution stages corresponding to the evolution process of geological hazards such as landslides: initial micro-expansion stage, stable expansion stage, accelerated expansion stage, and near-landslide instability stage. Each evolution stage is not defined by a single numerical value, but rather by a multi-dimensional feature vector space, which consists of crack width (W), crack length (L), and crack width change rate (W / L). ), crack length change rate ( The probability distribution and threshold range of key features such as the equivalent stress intensity factor (K) are used to construct the feature vector space. For example, the feature vector space of the "stable expansion stage" might be defined as: W between 5-20 mm, L between 0.5-2 m. and All follow a normal distribution with a mean of 0.1 mm / day and a standard deviation of 0.05 mm / day, and the K value is at a low and stable level. The feature vector space of the "accelerated expansion phase" can be defined as: and The mean value increased significantly to 1 mm / day, and the standard deviation of its distribution also increased accordingly, with the K value showing a continuous upward trend.

[0088] The core reasoning process of this scenario-based reasoning model employs a Bayesian network algorithm. Specifically, the model first receives the real-time feature vector extracted in S2. The model then calculates which feature vector X belongs to each preset evolution stage. posterior probability According to Bayes' theorem, ,in It is a stage The probability of observing X in the eigenvector space (which can be calculated using the probability density function of a multidimensional Gaussian distribution). It is a stage The prior probability (which can be set based on historical statistical data) is calculated. The model will calculate the posterior probability of all four stages, normalize it, and finally output the confidence level of the crack's current stage.

[0089] The system continuously records the identified evolutionary stage and its confidence level at each time point. For example, at time t-1, the crack is identified as being in a stable propagation stage (confidence level 80%); at time t, based on new input data, the system identifies the crack as being in an accelerated propagation stage (confidence level 75%). Comparing the evolutionary stages at these two times reveals their differences. Further, the system checks whether the confidence level of the current stage (accelerated propagation stage) is greater than a preset mutation threshold. If the confidence level is greater than the preset mutation threshold, it determines that a mutation has occurred in the crack evolutionary stage from "stable propagation" to "accelerated propagation," and immediately generates a mutation signal. Conversely, if the conditions are not met, such as no stage change or insufficient confidence, it determines that no effective mutation has occurred, no mutation signal is generated, and the system continues to identify the evolutionary stage for the next cycle according to the standard procedure of S4.

[0090] S5: Based on the judgment of abrupt changes in the evolutionary stage, combine the current evolutionary state of the crack with external environmental data to conduct environmental correlation analysis on the crack state, identify factors that accelerate crack propagation, and obtain a causal correlation list.

[0091] In this embodiment, when S5 receives a mutation signal, it triggers and executes a time-leading-lag correlation analysis algorithm to identify the main environmental factors that induce crack evolution and obtain the causal intensity coefficient. The specific analysis steps include:

[0092] Step 1: Taking the mutation point of the evolution stage as the center, extract a preset analysis time window that includes the time series of crack propagation index and the time series of various external environmental data;

[0093] Step 2: For each external environment data time series, calculate its cross-correlation function with the crack propagation index time series to analyze the correlation between the two under different time delays;

[0094] Step 3: For each cross-correlation function, locate the peak of its cross-correlation function and extract the maximum correlation coefficient and time delay corresponding to the peak;

[0095] Step 4: For each external environmental factor, determine whether the time delay corresponding to the external environmental factor is positive and whether the value of the time delay is within the preset physical reasonable time interval. At the same time, determine whether the maximum correlation coefficient corresponding to the external environmental factor is greater than the preset significance threshold.

[0096] Step 5: If an environmental factor satisfies both the time rationality judgment and the correlation significance judgment, then the environmental factor is identified as the cause of this crack evolution, and the maximum correlation coefficient is used as the causal strength coefficient of the cause for quantitative output.

[0097] Step 6: Output a causal relationship list of all environmental factors identified as triggers. The causal relationship list includes all identified triggers, the causal strength coefficient of each trigger, and the time delay.

[0098] If no confirmed cause exists, the sudden change in the crack evolution is determined to be an event with unknown external cause, triggering an internal instability assessment process. This involves increasing the frequency of internal state data collection and conducting in-depth analysis of the crack's vibration spectrum, stress intensity factor, and acceleration factor to identify any precursory features of internal instability. If a clear precursory feature of internal instability is identified, a high-level warning is generated and reported. If no clear precursory feature is identified, a low-level warning with an unknown cause requiring manual intervention is generated and reported.

[0099] It should be specifically explained that when a mutation signal is received, the system first executes step 1, extracting a preset analysis time window centered on the mutation time t, such as 72 hours before and 24 hours after the mutation point, totaling 96 hours of data. This window includes the synchronously aligned crack propagation index time series, as well as time series of multiple external environmental data such as rainfall, soil moisture content, temperature, and wind speed. Next, in step 2, for each external environmental data, its cross-correlation function with the crack propagation index time series is calculated. For example, the system calculates the cross-correlation function between the "rainfall" time series and the crack propagation index time series, which shows the correlation between rainfall and the crack propagation index at different time delays. In step 3, the peak value of the cross-correlation function is located, and two key parameters are extracted: the maximum correlation coefficient and the time delay corresponding to the peak value. A positive time delay indicates that the change in rainfall precedes the change in the crack propagation index by 6 hours. Step 4 is executed, performing a dual judgment on each environmental factor. First, a time reasonableness judgment is performed, checking whether the time delay is positive and whether its value is within the preset physical reasonableness time interval. For example, for a rainfall-induced landslide, a reasonable delay time might be between 2 and 48 hours. If the calculated delay is 6 hours, it is considered reasonable; if it is -2 hours (crack changes precede rainfall) or 100 hours (too long, possibly irrelevant), it is considered unreasonable. Next, a correlation significance assessment is performed, checking if the maximum correlation coefficient is greater than a preset significance threshold (e.g., 0.7). If the calculated coefficient is 0.85, it is considered significant. In step 5, only environmental factors that simultaneously meet both of the above criteria are confirmed as the trigger for this crack evolution. For example, if the delay for "rainfall" is 6 hours and the maximum correlation coefficient is 0.85, then rainfall is confirmed as the trigger, and its causal strength coefficient is quantified as 0.85. Finally, in step 6, a causal association list is output: {Trigger 1: Rainfall, Causal Strength Coefficient: 0.85, Time Delay: 6 hours}. If, after analysis, no environmental factor meets the criteria, this mutation is determined to be an "unknown external trigger event," and the internal instability assessment process is triggered. At this point, the system will automatically increase the acquisition frequency of the MEMS and microwave sensors from 1Hz to 10Hz, and perform in-depth analysis of the vibration spectrum, equivalent stress intensity factor, and acceleration factor acquired at higher frequencies. For example, by analyzing whether new low-frequency components appear in the vibration spectrum or whether the stress intensity factor shows a step increase, it can identify whether there are signs of instability precursors such as internal soil and rock structure deterioration or shear zone formation. If a clear sign of internal instability is identified, a high-level warning of emergency instability risk is generated and reported; if no clear sign is identified, a low-level warning of "unknown cause, manual intervention required" is generated and reported, reminding management personnel to conduct on-site investigation.

[0100] S6: Combine the causal relationship list with historical crack evolution data to predict crack evolution trends and obtain evolution prediction coefficients.

[0101] In this embodiment, S6 uses the causal association list output by S5 as a retrieval condition to match and retrieve historical crack evolution cases from the historical database. The matching process quantifies similarity by calculating the similarity between the current causal association list and the causal association lists of each historical case. The similarity is calculated based on the comprehensive difference of the cause type, causal intensity coefficient and time delay, and historical cases with similarity higher than the preset similarity threshold are selected as the effective case set.

[0102] Extract the evolution trajectory data of each case in the effective case set within a preset time period after the mutation, and calculate the statistical distribution characteristics of crack evolution rate, crack evolution acceleration and final instability time.

[0103] The statistical distribution characteristics are weighted and fused with the real-time evolution data of the current crack after the mutation. Then, the evolution trajectory of the current crack is corrected by using the fused data through a preset prediction model. Finally, the evolution prediction coefficient is output. The evolution prediction coefficient is used to predict the probability that the crack will enter the next more dangerous evolution stage within a preset time window in the future.

[0104] It should be specifically explained that the causal association list output by S5 is used as the search criteria to find similar cases from the historical crack evolution database. For example, if the current causal association list is {Induced Factor 1: Rainfall, Causal Strength Coefficient: 0.85, Time Delay: 6 hours}, the similarity between this list and the causal association lists of each historical case in the database is calculated. This similarity is a comprehensive score calculated based on the differences in inducement type, causal strength coefficient, and time delay. For example, the similarity calculation formula is: Where I represents the trigger type, and Because the triggering factors are of the same type, For different types of triggers, R is the current causal strength coefficient. The historical causal strength coefficient. This is a delay of the current time. Due to historical time delay, The preset maximum reasonable delay, The trigger type weighting coefficient measures the importance of trigger type in similarity calculation. It is determined through expert experience or statistical methods (such as questionnaires) and reflects the subjective or objective weight of the trigger type's influence on crack evolution. For example, in crack evolution analysis, rainfall may be considered a key trigger, so B1 may be assigned a higher value (e.g., 0.5), while other triggers (such as temperature) may have a lower weight. The causal intensity coefficient weighting coefficient is used to measure the importance of the causal intensity coefficient in similarity calculation. It is determined based on expert experience or data-driven methods (such as regression analysis) and reflects the quantitative weight of the influence of causal intensity on crack evolution. The time delay weighting coefficient measures the importance of time delay in similarity calculation. It is determined based on sensitivity analysis of the impact of time delay on crack evolution, and is obtained through historical data fitting or expert judgment. If a historical case is caused by "rainfall" and has a causality strength coefficient of 0.80, with a time delay of 8 hours, its similarity to the current event will be very high. The system will filter out all historical cases with similarity higher than a preset threshold (e.g., 0.8) to form a "valid case set." Statistical analysis will then be performed on the valid case set to extract historical evolution patterns. For example, if the valid case set contains 10 cases, the evolution trajectory data of these 10 cases within 72 hours after the mutation will be extracted. Then, the statistical distribution characteristics of these trajectory data will be calculated, such as: calculating the average crack evolution rate and its standard deviation for all cases within 24 hours after the mutation; calculating the average evolution acceleration; and statistically analyzing the distribution of the time taken for these cases from the mutation point to final instability, such as an average of 48 hours and a standard deviation of 12 hours. If no valid case is found, the system directly extrapolates the trend based on the current real-time evolution data using a pre-defined pure data-driven model, and outputs an evolution prediction coefficient with low confidence. Finally, the statistical distribution features extracted from historical cases are weighted and fused with the real-time evolution data of the current crack after its mutation. For example, if the current crack has been monitored for 6 hours after the mutation, its real-time average evolution rate is 1.5 mm / day. The average rate of the historical case set is 1.2 mm / day. Based on the similarity and quantity of historical cases, historical statistical features are assigned certain weights, while the current real-time data is given higher weights. A corrected evolution rate prediction value is obtained through fusion. The fused data, along with the statistical distribution of historical instability times, is input into a pre-defined prediction model (such as a time series prediction model). The model uses this data to correct the evolution trajectory of the current crack and ultimately outputs an evolution prediction coefficient.

[0105] S7: Based on the predicted results of the crack evolution state, determine whether to trigger an early warning prompt for the crack evolution state.

[0106] In this embodiment, S7 compares the output evolution prediction coefficient with a preset evolution threshold. When the evolution prediction coefficient exceeds the preset evolution threshold, it is determined that there is a safety risk in the crack evolution state, and an early warning feedback mechanism is triggered to issue an early warning prompt to the monitoring terminal. At the same time, an early warning monitoring report is automatically generated and uploaded to the monitoring terminal. Otherwise, the crack state data continues to be monitored.

[0107] like Figure 2The embodiment shown provides an implementation system corresponding to a method for monitoring ultra-wide cracks in landslides, including a data acquisition module, a feature extraction module, a crack anomaly identification module, a crack evolution analysis module, a crack environment correlation analysis module, an evolution prediction module, and an early warning feedback module. The data acquisition module is connected to the feature extraction module, the feature extraction module is connected to the crack anomaly identification module, the feature extraction module is connected to the crack evolution analysis module, the crack evolution analysis module is connected to the crack environment correlation analysis module, the crack environment correlation analysis module is connected to the evolution prediction module, and the evolution prediction module is connected to the early warning feedback module.

[0108] The data acquisition module collects crack status data in collaboration with microwave ranging sensors, MEMS sensors and vision sensors, and simultaneously acquires external environment data, and preprocesses the crack status data and external environment data.

[0109] The feature extraction module extracts features from the preprocessed crack state data to obtain crack width features, crack morphology features, and crack mechanical features.

[0110] The crack anomaly identification module analyzes the crack change state within a preset time window based on crack width characteristics, crack morphology characteristics, and crack mechanical characteristics to obtain a crack propagation index, thereby identifying the abnormal state of the crack.

[0111] The crack evolution analysis module triggers further analysis of crack evolution based on the identification results of crack abnormal states. It inputs crack width characteristics, crack morphology characteristics, and crack mechanical characteristics into a preset scenario reasoning model to obtain the current evolution stage of the crack and determine whether a sudden change has occurred in the current evolution stage of the crack.

[0112] The crack environment correlation analysis module combines the current evolutionary state of the crack with external environmental data based on the judgment result of the abrupt change in the evolutionary stage, performs environmental correlation analysis on the crack state, identifies the factors that accelerate crack expansion, and obtains a causal correlation list.

[0113] The evolution prediction module combines the causal relationship list with historical crack evolution data to predict the crack evolution trend and obtain the evolution prediction coefficient.

[0114] The early warning feedback module determines whether to trigger an early warning prompt regarding the crack evolution state based on the predicted crack evolution state.

[0115] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0116] 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.

Claims

1. A method for monitoring ultra-wide cracks in landslides, characterized in that, include: S1: Crack status data is collected collaboratively by microwave ranging sensor, MEMS sensor and vision sensor, and external environment data is acquired at the same time. The crack status data and external environment data are preprocessed. S2: Extract features from the preprocessed crack state data to obtain crack width features, crack morphology features, and crack mechanical features; S3: Based on the crack width characteristics, crack morphology characteristics, and crack mechanical characteristics, the crack change state is analyzed within a preset time window to obtain the crack propagation index, thereby identifying the abnormal state of the crack. S4: Based on the identification results of the crack abnormal state, trigger further analysis of crack evolution, input crack width characteristics, crack morphology characteristics and crack mechanical characteristics into the preset scenario reasoning model, obtain the current evolution stage of the crack, and determine whether the current evolution stage of the crack has undergone a sudden change. S5: Based on the judgment results of abrupt changes in the evolutionary stage, combine the current evolutionary state of the crack with external environmental data, conduct environmental correlation analysis on the crack state, identify factors that accelerate crack propagation, and obtain a causal correlation list. Upon receiving a mutation signal, S5 triggers and executes a time-leading-lag correlation analysis algorithm to identify the main environmental factors inducing crack evolution and obtain the causal intensity coefficient. Specific analysis steps include: Step 1: Taking the mutation point of the evolution stage as the center, extract a preset analysis time window that includes the time series of crack propagation index and the time series of various external environmental data; Step 2: For each external environment data time series, calculate its cross-correlation function with the crack propagation index time series to analyze the correlation between the two under different time delays; Step 3: For each cross-correlation function, locate the peak of its cross-correlation function and extract the maximum correlation coefficient and time delay corresponding to the peak; Step 4: For each external environmental factor, determine whether the time delay corresponding to the external environmental factor is positive and whether the value of the time delay is within the preset physical reasonable time interval. At the same time, determine whether the maximum correlation coefficient corresponding to the external environmental factor is greater than the preset significance threshold. Step 5: If an environmental factor satisfies both the time rationality judgment and the correlation significance judgment, then the environmental factor is identified as the cause of this crack evolution, and the maximum correlation coefficient is used as the causal strength coefficient of the cause for quantitative output. Step 6: Output a causal relationship list of all environmental factors identified as triggers. The causal relationship list includes all identified triggers, the causal strength coefficient of each trigger, and the time delay. If no confirmed cause exists, the sudden change in the evolution of the crack is determined to be an event with an unknown external cause, triggering an internal instability assessment process. This involves increasing the frequency of internal state data collection and conducting in-depth analysis of the crack's vibration spectrum, stress intensity factor, and acceleration factor to identify any precursory features of internal instability. If a clear precursory feature of internal instability is identified, a high-level warning is generated and reported. If no clear precursory feature is identified, a low-level warning with an unknown cause requiring manual intervention is generated and reported. S6: Combine the causal relationship list with historical crack evolution data to predict crack evolution trends and obtain evolution prediction coefficients; S7: Based on the predicted results of the crack evolution state, determine whether to trigger an early warning prompt for the crack evolution state.

2. The method for monitoring ultra-wide cracks in landslides according to claim 1, characterized in that, The process of acquiring the S1 crack condition data includes: Simultaneously activate the microwave ranging sensor and the vision sensor; The distance to the other side of the crack is obtained by performing multiple distance measurements using a microwave sensor, and microwave distance measurement data is obtained. Multiple high-resolution visual images of the crack are captured by a visual sensor, and MEMS data is continuously recorded by a MEMS sensor during the acquisition process. The acquisition of the external environment data includes: synchronously acquiring environmental data around the crack through an external interface; Preprocessing of crack status data and external environment data includes: data synchronization, data alignment, data noise reduction, and data correction.

3. The method for monitoring ultra-wide cracks in landslides according to claim 2, characterized in that, The S2 step of feature extraction from the preprocessed crack state data includes: The crack width, the average crack width within a preset time window, the crack width change rate within a preset time window, the maximum crack width, and the minimum crack width are extracted from the preprocessed microwave ranging data as crack width features. The crack length, maximum crack length, minimum crack length, crack length change rate within a preset time window, crack area, and crack trunk propagation direction are extracted from the preprocessed crack visual image as crack morphology features. The cumulative tilt angle, equivalent stress intensity factor, rate of change of equivalent stress intensity factor, tilt angle rate, and vibration spectrum characteristics are extracted from the preprocessed MEMS data as crack mechanical characteristics. The equivalent stress intensity factor is estimated by fusing and analyzing the cumulative tilt angle, tilt angle rate, and vibration spectrum characteristics in the MEMS data and based on a preset geomechanical model or data-driven model.

4. The method for monitoring ultra-wide cracks in landslides according to claim 3, characterized in that, The S3 analysis of crack change state includes: Based on the extracted crack width change rate and crack length change rate, and construct their time series respectively, calculate their first time derivatives based on the crack width change rate time series and crack length change rate time series, construct dynamic acceleration factors, and obtain width acceleration factor and length acceleration factor. The dynamic acceleration factor is used to characterize the degree of change rate surge. The crack propagation index is calculated by weighting and fusing the crack width, crack length, crack width change rate, crack length change rate, crack area, equivalent stress intensity factor, equivalent stress intensity factor change rate, width acceleration factor, and length acceleration factor using preset weighting coefficients. The crack propagation index is compared with a preset multi-level safety threshold to determine whether the crack is in an abnormal state. If the crack propagation index is less than or equal to the preset safety threshold, the crack is determined to be in a normal state, and the crack state continues to be monitored and identified for abnormalities. If the crack propagation index is greater than the preset safety threshold, the crack is determined to be in an abnormal state, and the crack evolution analysis process is triggered and executed.

5. The method for monitoring ultra-wide cracks in landslides according to claim 4, characterized in that, The S4 inputs the real-time collected crack width features, crack morphology features, and crack mechanical features into a preset scenario reasoning model to determine the current evolution stage of the crack. The scenario reasoning model has multiple preset discrete evolution stages corresponding to the landslide evolution process. The discrete evolution stages include the initial micro-expansion stage, the stable expansion stage, the accelerated expansion stage, and the pre-landslide instability stage. Each evolution stage is defined by a multi-dimensional feature vector space. The multi-dimensional feature vector space consists of the feature thresholds and probability distributions of crack width, crack length, crack width change rate, crack length change rate, and equivalent stress intensity factor. The scenario reasoning model calculates the matching probability between the current input features and the feature vector space of each evolution stage, and uses Bayesian networks or evidence theory algorithms to update and output the confidence level of the crack in each evolution stage. By comparing the matching probabilities of each evolution stage, the evolution stage with the highest matching probability is identified as the current evolution stage of the crack, and the confidence level of that evolution stage is output. The system continuously records the currently identified evolutionary stage and compares it with the evolutionary stage at the previous moment. If the current evolutionary stage is different from the previous evolutionary stage and the confidence level of the current evolutionary stage is greater than the preset mutation threshold, it is determined that a mutation has occurred in the crack evolutionary stage, and a mutation signal is generated to trigger the environmental correlation analysis of S5.

6. The method for monitoring ultra-wide cracks in landslides according to claim 1, characterized in that, S6 uses the causal association list output by S5 as a retrieval condition to match and retrieve historical crack evolution cases from the historical database. The matching process quantifies similarity by calculating the similarity between the current causal association list and the causal association lists of each historical case. The similarity is calculated based on the comprehensive difference of the cause type, causal intensity coefficient and time delay, and historical cases with similarity higher than the preset similarity threshold are selected as the effective case set. Extract the evolution trajectory data of each case in the effective case set within a preset time period after the mutation, and calculate the statistical distribution characteristics of crack evolution rate, crack evolution acceleration and final instability time. The statistical distribution characteristics are weighted and fused with the real-time evolution data of the current crack after the mutation. Then, the evolution trajectory of the current crack is corrected by using the fused data through a preset prediction model. Finally, the evolution prediction coefficient is output. The evolution prediction coefficient is used to predict the probability that the crack will enter the next more dangerous evolution stage within a preset time window in the future.

7. The method for monitoring ultra-wide cracks in landslides according to claim 1, characterized in that, The S7 compares the output evolution prediction coefficient with the preset evolution threshold. When the evolution prediction coefficient exceeds the preset evolution threshold, it is determined that there is a safety risk in the crack evolution state, and an early warning feedback mechanism is triggered to issue an early warning prompt to the monitoring terminal. At the same time, an early warning monitoring report is automatically generated and uploaded to the monitoring terminal. Otherwise, the crack state data continues to be monitored.

8. A monitoring system for ultra-wide cracks in landslides, implementing the method for monitoring ultra-wide cracks in landslides as described in any one of claims 1-7, characterized in that, include: Data acquisition module: Collects crack status data in collaboration with microwave ranging sensor, MEMS sensor and vision sensor, and simultaneously acquires external environment data, and preprocesses crack status data and external environment data; Feature extraction module: Extracts features from the preprocessed crack state data to obtain crack width features, crack morphology features, and crack mechanical features; Crack anomaly identification module: Based on crack width characteristics, crack morphology characteristics, and crack mechanical characteristics, the module analyzes the crack change state within a preset time window to obtain the crack propagation index, thereby identifying the abnormal state of the crack. Crack Evolution Analysis Module: Based on the identification results of crack anomaly states, it triggers further analysis of crack evolution, inputs crack width characteristics, crack morphology characteristics, and crack mechanical characteristics into a preset scenario reasoning model, obtains the current evolution stage of the crack, and determines whether a sudden change has occurred in the current evolution stage of the crack. Crack Environment Correlation Analysis Module: Based on the judgment results of abrupt changes in the evolution stage, the current evolutionary state of the crack is combined with external environmental data to perform environmental correlation analysis on the crack state, identify factors that accelerate crack propagation, and obtain a causal correlation list. Evolution Prediction Module: Combines the causal relationship list with historical crack evolution data to predict crack evolution trends and obtain evolution prediction coefficients; Early warning feedback module: Based on the prediction results of crack evolution status, determine whether to trigger an early warning prompt for crack evolution status.

Images