Tunnel surrounding rock deformation dynamic monitoring system

Abstract

The present application belongs to the technical field of tunnel surrounding rock profile measurement monitoring, and provides a tunnel surrounding rock deformation dynamic monitoring system, which comprises: synchronously collecting surrounding rock displacement data and probe micro-region local humidity data, triggering a crystallization interference diagnosis time window according to the humidity drop rate; extracting a displacement step mutation waveform in the time window, calculating the rising edge slope and time domain and frequency domain characteristic parameters; automatically identifying crystallization disturbance false data through a salt crystallization push stress feature recognition model; after removing the abnormal data section, performing qualifiedness verification on the remaining effective data, and through curve fitting reconstruction data, outputting the real surrounding rock mechanical deformation data after removing the crystallization interference. The present application can accurately strip the salt crystallization interference error, effectively improving the accuracy and reliability of the dynamic deformation monitoring of the salt-containing stratum tunnel surrounding rock.

Description

Technical Field

[0001] This invention belongs to the field of tunnel surrounding rock contour measurement and monitoring technology, specifically a tunnel surrounding rock deformation dynamic monitoring system. Background Technology

[0002] High-salinity groundwater is present in water-rich salt rock strata and in the fissures of the surrounding rock of submarine tunnels. Saltwater easily seeps into the micro-cracks where the monitoring probe meets the rock wall via capillary channels. The continuous airflow from the tunnel ventilation carries away moisture from these cracks, causing the salt solution to become supersaturated, precipitate salt crystals, and expand, creating a mechanical pushing effect on the probe. Existing methods for monitoring surrounding rock deformation only provide corrosion-resistant sealing to the sensors, failing to consider the false displacement signals caused by salt crystallization in the micro-cracks. Measured monitoring data is mixed with crystallization disturbance values ​​and the actual deformation of the surrounding rock, making it difficult to accurately achieve dynamic monitoring and safety early warning of surrounding rock deformation. Therefore, this invention provides a dynamic monitoring system for tunnel surrounding rock deformation. Summary of the Invention

[0003] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0004] The technical solution adopted by this invention to solve its technical problem is: This invention provides a dynamic monitoring system for deformation of surrounding rock in tunnels, comprising: Data acquisition module: Real-time acquisition of overall rock displacement data output by displacement sensor, as well as local humidity data of the micro-area in contact between displacement sensor probe and rock surface; Diagnostic window trigger module: Real-time comparison of the rate of decrease of local humidity data with the evaporation critical threshold. When the rate of decrease of local humidity data exceeds the evaporation critical threshold, a dedicated diagnostic time window for crystallization interference is opened. Waveform feature extraction module: Within the diagnostic time window, acquire the step change waveform in the surrounding rock displacement data, and calculate the rising slope and time-domain and frequency-domain feature parameters of the step change waveform respectively. Crystallization Interference Identification Module: Construct a salt crystallization pushing stress characteristic identification model, input the rising edge slope and frequency domain characteristic parameters into the salt crystallization pushing stress characteristic identification model, and determine whether the current step signal belongs to false displacement data caused by salt crystallization pushing based on the output results of the salt crystallization pushing stress characteristic identification model; Data correction output module: If the data is determined to be false displacement data, the data segment corresponding to the step change in the false displacement data is removed to obtain valid normal monitoring data. The valid normal monitoring data is analyzed for compliance to determine whether it is qualified. If it is qualified, curve fitting processing is performed on the valid normal monitoring data to output the true mechanical deformation data of the surrounding rock after removing crystallization disturbance.

[0005] As a further improvement of the present invention, the specific acquisition process of the overall displacement data of the surrounding rock output by the real-time acquisition displacement sensor, and the local humidity data arranged in the micro-area of ​​contact between the displacement sensor probe and the rock surface is as follows: Displacement sensors are installed at the monitored section of the tunnel surrounding rock, with the sensor probes tightly attached to the rock surface. The overall displacement data of the surrounding rock is collected in real time by the displacement sensors at a first sampling frequency. In the micro-area where the displacement sensor probes contact the rock surface, a miniature humidity sensor is attached. The local humidity data is collected in real time by the miniature humidity sensor at a second sampling frequency. The collected overall displacement data and local humidity data of the surrounding rock are timestamped to obtain overall displacement data and local humidity data of the surrounding rock with synchronized timestamps.

[0006] As a further improvement of the present invention, the specific process of comparing the rate of decrease of local humidity data with the evaporation critical threshold in real time is as follows: Acquire local humidity data from multiple sampling periods and construct a sliding calculation window for humidity changes; Within the sliding calculation window for humidity change, the difference between the local humidity data at the current moment and the local humidity data at the beginning of the window is calculated and divided by the time span of the sliding calculation window for humidity change to obtain the rate of decrease of the local humidity data. An evaporation critical threshold is pre-calibrated, and the rate of decrease of local humidity data calculated in real time is compared with the pre-calibrated evaporation critical threshold in real time.

[0007] As a further improvement of the present invention, the specific process of opening the dedicated diagnostic time window for crystallization interference is as follows: When the comparison result shows that the rate of decrease of local humidity data is greater than the critical threshold for evaporation, it is determined that water evaporation and salt crystallization have occurred. A time trigger signal is generated using the timestamp of the current determination time as the starting time. Based on the time-triggered signal, a dedicated diagnostic time window for crystallization interference is opened, and the overall displacement data of the surrounding rock is acquired during the opening of the dedicated diagnostic time window for crystallization interference.

[0008] As a further improvement of the present invention, the specific process of obtaining the step abrupt change waveform in the surrounding rock displacement data is as follows: Obtain the overall displacement data of the surrounding rock within the specific diagnostic time window for crystallization interference, and construct a displacement time series; The displacement time series is subjected to first-order difference calculation to obtain the displacement change rate between adjacent sampling points; Set a mutation rate threshold, obtain the displacement change rate of all displacement time series, and when the number of sampling points that continuously exceed the mutation rate threshold reaches the mutation determination point threshold, mark the first sampling point that exceeds the mutation rate threshold as the mutation start point, and mark the sampling points whose displacement change rate falls back to below the mutation rate threshold as the mutation end point. Obtain the overall displacement data segment of the surrounding rock between the start and end points of the mutation, as the step mutation waveform.

[0009] As a further improvement of the present invention, the specific process of calculating the rising edge slope of the step change waveform and the two types of characteristic parameters in the time domain and frequency domain is as follows: Calculate the rising edge slope: Obtain the displacement value corresponding to the end point of the mutation and the displacement value corresponding to the start point of the mutation, calculate the displacement difference between the two, and divide the displacement difference by the time span from the start point to the end point of the mutation to obtain the rising edge slope. Calculate time-domain characteristic parameters: Perform statistical analysis on the step change waveform, obtain the displacement difference as the change amplitude, obtain the time span as the duration, and calculate the variance of the displacement values ​​of all sampling points in the step change waveform. Combine the change amplitude, duration, and variance to form the time-domain characteristic parameters. Calculate frequency domain characteristic parameters: Perform a fast Fourier transform on the step change waveform to convert the time domain signal into a frequency domain signal; obtain the dominant frequency, the spectral energy ratio of the high-frequency band, and the peak power spectral density in the frequency domain signal, and combine the three to form the frequency domain characteristic parameters.

[0010] As a further improvement of the present invention, the specific process of constructing the salt crystallization pushing stress characteristic identification model is as follows: Obtain various step-change waveform samples from historical monitoring data of water-rich salt rock strata, including salt crystal jacking interference samples verified by on-site geological exploration, and real samples of sudden deformation of surrounding rock; Obtain the sample rising edge slope, sample time domain feature parameters, and sample frequency domain feature parameters corresponding to each sample, and normalize the obtained feature data. The samples were labeled with classification labels. Salt crystallization pushing interference samples were labeled as positive class labels, and real surrounding rock sudden deformation samples were labeled as negative class labels. The samples were combined to form a training dataset. An initial salt crystallization pushing stress feature identification model was constructed based on the support vector machine algorithm; The training dataset is input into the initial salt crystallization pushing stress feature recognition model for supervised learning training. The sample features are mapped to a high-dimensional space through a kernel function. The model is iteratively optimized and the optimal classification hyperplane that can maximize the distinction between positive and negative samples is found. After the training converges, the network weights and hyperplane parameters are saved to obtain the completed salt crystallization pushing stress feature recognition model.

[0011] As a further improvement of the present invention, the specific process for determining whether the current step signal belongs to false displacement data caused by salt crystallization push is as follows: The rising edge slope, time-domain feature parameters and frequency-domain feature parameters obtained in real time are normalized and concatenated into a feature vector to be identified. The feature vector to be identified is input into the salt crystallization pushing stress feature identification model; The salt crystallization pushing stress feature recognition model is based on the internal optimal classification hyperplane. It calculates the spatial distance between the feature vector to be identified and the optimal classification hyperplane, and outputs the corresponding classification prediction results. The determination is made based on the output of the salt crystallization jacking stress characteristic identification model: if the output is a positive label, the current step change waveform is determined to be false displacement data caused by salt crystallization jacking; if the output is a negative label, the current step change waveform is determined to be not crystallization interference, but a real sudden deformation of the surrounding rock.

[0012] As a further improvement of the present invention, the specific process for determining whether valid and normal monitoring data is qualified is as follows: Based on the marked mutation start and end points, locate the data interval where the step mutation waveform is located in the overall displacement data of the surrounding rock, and obtain the calculated mutation amplitude. The baseline offset compensation is performed by uniformly subtracting the mutation amplitude from all overall displacement data of the surrounding rock after the mutation endpoint. Delete the data segment between the mutation initiation point and the mutation endpoint, and stitch the subsequent data after baseline compensation with the data before the mutation initiation point on the time axis to obtain effective normal monitoring data; Set a minimum effective data length threshold and a baseline variance threshold; Obtain the total number of remaining sampling points contained in the crystallization interference-specific diagnostic time window after splicing of valid normal monitoring data, and calculate the overall variance of valid normal monitoring data; Determine whether the total number of remaining sampling points is greater than the minimum effective data length threshold and whether the overall variance is less than the baseline variance threshold; If both of the above conditions are met, the data is considered valid and normal; if either condition is not met, the data is considered unqualified.

[0013] As a further improvement of the present invention, the specific process of outputting the true mechanical deformation data of the surrounding rock after removing crystallization disturbance is as follows: A two-dimensional coordinate system is constructed with timestamp as the independent variable and displacement value in valid and normal monitoring data as the dependent variable; The nonlinear fitting method is used to fit the effective normal monitoring data to obtain the optimal fitting parameters and generate a continuous displacement change fitting curve. By using displacement variation fitting curves, numerical interpolation is performed on the time period between the start and end of the eliminated mutation to fill in the missing data points. The interpolated and filled data segments are then re-integrated with the valid normal monitoring data to serve as the true mechanical deformation data of the surrounding rock after removing crystallization disturbances.

[0014] The beneficial effects of this invention are as follows: Existing methods for monitoring surrounding rock deformation only address the sensor with routine anti-corrosion treatment, failing to consider the interference from spurious step displacements caused by salt evaporation and crystallization in micro-regions of the rock surface and crystal jacking. This makes it impossible to distinguish between crystallization disturbances and actual surrounding rock deformation. This invention innovatively introduces a humidity decrease rate threshold-triggered diagnostic mechanism, precisely pinpointing the specific time periods of salt analysis crystallization and crystal growth, achieving accurate location of interference scenarios, and filling the technological gap in monitoring surrounding rock deformation in water-rich, high-salt tunnels.

[0015] This invention combines the rising edge slope of the step waveform with multi-dimensional time and frequency domain features, and relies on a salt crystallization jacking identification model trained by SVM. It can accurately distinguish between false mutations caused by crystallization expansion and real mechanical deformation mutations of the surrounding rock, overcoming the shortcomings of traditional single threshold identification and conventional filtering algorithms that cannot distinguish between the two types of mutation signals, and greatly improving the accuracy of disturbance identification.

[0016] After removing abnormal segments that interfere with crystallization, this invention adds a valid data qualification verification mechanism, combining data length and baseline variance to ensure the validity and reliability of residual data; then, it uses nonlinear curve fitting interpolation to complete missing data, avoiding data discontinuity and monitoring curve breakage caused by direct data deletion, and restoring the true dynamic deformation law of the surrounding rock to the greatest extent. Attached Figure Description

[0017] The invention will now be further described with reference to the accompanying drawings.

[0018] Figure 1 This is a system flowchart of the tunnel surrounding rock deformation dynamic monitoring system of the present invention; Figure 2 This is a schematic diagram of the step abrupt change waveform and feature extraction of surrounding rock displacement data in an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the effects of false displacement data removal and curve fitting correction in an embodiment of the present invention. Detailed Implementation

[0019] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0020] Example 1 like Figure 1As shown in the embodiment of the present invention, the tunnel surrounding rock deformation dynamic monitoring system includes: The data acquisition module collects the overall displacement data of the surrounding rock output by the displacement sensor in real time, as well as the local humidity data of the micro-area in contact between the displacement sensor probe and the rock surface. In some specific embodiments, the data acquisition module performs the following process for real-time acquisition of overall rock displacement data output by the displacement sensor, as well as local humidity data located in the micro-area of ​​contact between the displacement sensor probe and the rock surface: The displacement sensor is installed at the section of the surrounding rock to be monitored, and the displacement sensor probe is tightly attached to the surface of the surrounding rock. The displacement sensor collects the overall displacement data of the surrounding rock in real time according to a preset first sampling frequency. A miniature humidity sensor is attached to the micro-area in contact with the rock surface of the displacement sensor probe. Specifically, the micro-area in contact with the rock surface is defined as the annular space extending outward from the edge of the displacement sensor probe by a predetermined distance, and the tiny gap between the displacement sensor probe and the surrounding rock surface. This area is a key location in water-rich salt rock strata where seepage water is most likely to accumulate and is most susceptible to evaporation due to tunnel ventilation. The local humidity data is collected in real time by the miniature humidity sensor at a preset second sampling frequency; To ensure accurate alignment of subsequent data analysis in the time dimension, a unified system clock is used to timestamp the collected overall displacement data of the surrounding rock and the local humidity data, and the overall displacement data of the surrounding rock and the local humidity data carrying the synchronized timestamp are transmitted to the data processing terminal of the monitoring system in real time.

[0021] It should be noted that the first sampling frequency and the second sampling frequency are as follows: the first sampling frequency is a low-frequency sampling frequency set for the slow creep characteristics of the surrounding rock, used to capture macroscopic mechanical deformation; the second sampling frequency is a high-frequency sampling frequency set for the rapid evaporation characteristics of moisture in the microenvironment, and the second sampling frequency is greater than the first sampling frequency, so as to ensure that the instantaneous sharp drop in humidity that triggers salt crystallization can be keenly captured.

[0022] The diagnostic window trigger module compares the rate of decrease of local humidity data with the evaporation critical threshold in real time. When the rate of decrease of local humidity data exceeds the evaporation critical threshold, a dedicated diagnostic time window for crystallization interference is opened. In some specific embodiments, the process of comparing the rate of decrease of local humidity data with the evaporation critical threshold in the diagnostic window trigger module in real time is as follows: Acquire the local humidity data for multiple consecutive sampling periods with synchronization timestamps, and construct a sliding calculation window for humidity changes (e.g., 10 sampling periods constitute the sliding calculation window for humidity changes). Within the humidity change sliding calculation window, the difference between the local humidity data at the current moment and the local humidity data at the beginning of the window is calculated, and the difference is divided by the time span of the window to obtain the rate of decrease of the local humidity data. Based on the physical characteristics of salt crystallization in water-rich salt rock formations and the average ventilation speed in the tunnel, the evaporation critical threshold is pre-calibrated. The evaporation critical threshold represents the minimum rate of humidity decrease that is sufficient to cause rapid evaporation of water and trigger a large amount of salt crystallization. The rate of decrease of the local humidity data calculated in real time is compared with the pre-calibrated evaporation critical threshold in real time.

[0023] In some specific embodiments, the process of opening a dedicated diagnostic time window for crystallization interference in the diagnostic window trigger module when the rate of decrease in local humidity data exceeds the critical threshold for evaporation is as follows: When the comparison result shows that the rate of decrease of the local humidity data is greater than the evaporation critical threshold, it is determined that a violent water evaporation and salt crystallization phenomenon is occurring in the micro-area of ​​the rock surface contact. Using the timestamp of the current determination time as the starting time, a time trigger signal is generated in the data processing terminal of the monitoring system; Based on the time-triggered signal, the dedicated diagnostic time window for crystallization interference is activated and opened; wherein, the duration of the dedicated diagnostic time window for crystallization interference is set according to the typical physical action cycle of the pushing stress generated by the volume expansion of salt crystals, so as to ensure that the window can fully cover the entire process from the rapid evaporation of water to the pushing interference generated by the expansion of salt crystals on the displacement sensor probe. During the dedicated diagnostic time window for crystallization interference, the system synchronously locks the overall displacement data of the surrounding rock collected within that time period as the target data segment for subsequently acquiring step change waveforms.

[0024] The waveform feature extraction module acquires the step change waveform in the surrounding rock displacement data within the diagnostic time window, and calculates the rising slope of the step change waveform and two types of feature parameters in the time domain and frequency domain respectively. In some specific embodiments, the specific process of obtaining the step abrupt change waveform in the surrounding rock displacement data within the diagnostic time window in the waveform feature extraction module is as follows: Obtain the overall displacement data of the surrounding rock locked within the dedicated diagnostic time window for crystallization interference, and construct a displacement time series; The displacement time series is subjected to first-order difference calculation to obtain the displacement change rate between adjacent sampling points; A mutation rate threshold is set, which is calibrated based on the distinction between the extreme rate of normal surrounding rock physical creep and the empirical rate of instantaneous expansion and jacking of salt crystallization, so as to filter out normal slow surrounding rock deformation. Obtain the displacement change rate of all the displacement time series. When the number of sampling points that continuously exceed the mutation rate threshold reaches the mutation determination point threshold (e.g., 3 sampling points), mark the first sampling point that exceeds the mutation rate threshold as the mutation start point, and mark the sampling points whose displacement change rate subsequently falls back below the mutation rate threshold as the mutation end point. Obtain the overall displacement data segment of the surrounding rock between the mutation start point and the mutation end point, as the step mutation waveform.

[0025] In some specific embodiments, the specific process of calculating the rising edge slope of the step change waveform and the time-domain and frequency-domain feature parameters in the waveform feature extraction module is as follows: Calculate the rising edge slope: Obtain the displacement value corresponding to the abrupt end point and the displacement value corresponding to the abrupt start point, calculate the displacement difference between the two, and divide the displacement difference by the time span from the abrupt start point to the abrupt end point to obtain the rising edge slope, which characterizes the bursting speed of the salt crystallization pushing stress. Calculate the time-domain characteristic parameters: Perform statistical analysis on the step change waveform, obtain the displacement difference as the change amplitude, obtain the time span as the duration, and calculate the variance of the displacement values ​​of all sampling points in the step change waveform. Combine the change amplitude, the duration, and the variance to form the time-domain characteristic parameters, which are used to characterize the macroscopic energy and fluctuation degree of the push-up interference in the time dimension. Calculate the frequency domain characteristic parameters: Perform a fast Fourier transform on the step change waveform to convert the time domain signal into a frequency domain signal; since the process of salt crystal expansion pushing the displacement sensor probe is inevitably accompanied by high-frequency micro-vibrations of micro-crystal friction and fracture, obtain the main frequency, the spectral energy ratio of the high-frequency band, and the peak power spectral density in the frequency domain signal, and combine the three to form the frequency domain characteristic parameters, which are used to characterize the micro-physical features of crystallization stress from the frequency dimension.

[0026] Combination Figure 2 The diagram shows the step abrupt change waveform and feature extraction of surrounding rock displacement data. Within the locked diagnostic time window, normal surrounding rock creep exhibits a slowly rising linear or slightly nonlinear trend; however, when salt crystallization causes instantaneous expansion and pushes against the displacement sensor probe, a significant step abrupt change waveform is superimposed on the displacement curve. By comparing the first-order difference with the abrupt change rate threshold, the abrupt change initiation point is accurately located. Figure 2 (Middle red dot) and mutation endpoint ( Figure 2(Middle blue dot). Based on these two key nodes, the mutation amplitude (characterizing the magnitude of the push energy) and the rising slope connecting the start and end points are further extracted. Figure 2 The black dashed line in the middle represents the burst velocity of the push-out, providing an intuitive and quantifiable temporal physical feature input for the subsequent feature recognition model.

[0027] The crystallization interference identification module constructs a salt crystallization pushing stress characteristic identification model. The rising edge slope and frequency domain characteristic parameters are input into the salt crystallization pushing stress characteristic identification model. Based on the output results of the salt crystallization pushing stress characteristic identification model, it is determined whether the current step signal belongs to false displacement data caused by salt crystallization pushing. In some specific embodiments, the specific process of constructing the salt crystallization pushing stress feature recognition model in the crystallization interference recognition module is as follows: Obtain various step-change waveform samples from historical monitoring data of water-rich salt rock strata, including salt crystal jacking interference samples verified by on-site geological exploration, and real samples of sudden deformation of surrounding rock; The rising slope, time-domain feature parameters, and frequency-domain feature parameters of each sample are obtained, and the obtained feature data are normalized to eliminate dimensional differences. The samples were labeled with classification labels. Salt crystallization pushing interference samples were labeled as positive class labels, and real surrounding rock sudden deformation samples were labeled as negative class labels. The samples were combined to form a training dataset. An initial salt crystallization pushing stress feature identification model was constructed based on the support vector machine algorithm; The training dataset is input into the initial salt crystal pushing stress feature recognition model for supervised learning training. The sample features are mapped to a high-dimensional space through a kernel function. The model is iteratively optimized and the optimal classification hyperplane that can maximize the distinction between positive and negative samples is found. After the training converges, the network weights and hyperplane parameters are saved to obtain the completed salt crystal pushing stress feature recognition model.

[0028] In some specific embodiments, in the crystallization interference identification module, the rising edge slope and frequency domain characteristic parameters are input into the salt crystallization pushing stress characteristic identification model. The specific process of determining whether the current step signal belongs to false displacement data caused by salt crystallization pushing based on the output result of the salt crystallization pushing stress characteristic identification model is as follows: The rising edge slope, the time-domain feature parameters and the frequency-domain feature parameters obtained in real time from the waveform feature extraction module are normalized in the same way as in the model training stage, and then concatenated to construct the feature vector to be identified. The feature vector to be identified is input into the salt crystallization pushing stress feature identification model that has been constructed. The salt crystallization pushing stress feature recognition model is based on the internal optimal classification hyperplane, calculates the spatial distance of the feature vector to be identified relative to the optimal classification hyperplane, and outputs the corresponding classification prediction result. The following judgments are made based on the output of the salt crystallization jacking stress characteristic identification model: if the output is a positive label, it indicates that the burst speed, macroscopic energy and high-frequency microseismic characteristics of the current step mutation waveform are highly consistent with the physical law of instantaneous expansion of salt crystallization, and the current step mutation waveform is determined to be false displacement data brought about by salt crystallization jacking; if the output is a negative label, the current step mutation waveform is determined to be not crystallization interference, but a real sudden deformation of the surrounding rock.

[0029] If the data correction output module determines that the data is false displacement data, it will remove the data segments corresponding to the step change in the false displacement data to obtain valid normal monitoring data. It will then perform a qualification analysis on the valid normal monitoring data to determine whether the valid normal monitoring data is qualified. If qualified, it will perform curve fitting processing on the valid normal monitoring data and output the true mechanical deformation data of the surrounding rock after removing crystallization disturbance.

[0030] In some specific embodiments, in the data correction output module, if the data is determined to be false displacement data, the data segment corresponding to the step change within the false displacement data is removed to obtain valid and normal monitoring data. The specific process is as follows: Based on the mutation start point and mutation end point marked in the waveform feature extraction module, the data interval where the step mutation waveform is located is located in the overall displacement data of the surrounding rock. Because salt crystallization expansion will create a permanent physical rise at the displacement sensor probe, the data baseline will shift overall after the abrupt change, and the amplitude of the abrupt change will be obtained from the waveform feature extraction module. The baseline offset compensation is performed by uniformly subtracting the abrupt change amplitude from all overall displacement data of the surrounding rock after the abrupt change endpoint. Delete the data segment between the mutation start point and the mutation end point, and stitch the subsequent data after baseline compensation with the data before the mutation start point on the time axis to obtain the effective normal monitoring data that eliminates the crystallization elevation error.

[0031] In some specific embodiments, the data correction output module performs a qualification analysis on the valid and normal monitoring data to determine whether the valid and normal monitoring data is qualified. The specific process is as follows: Set a minimum effective data length threshold and a baseline variance threshold; wherein, the minimum effective data length threshold is used to ensure that the remaining data amount is sufficient to support subsequent mathematical fitting, and the baseline variance threshold is used to measure whether the data after removing abrupt changes has recovered to the normal slow creep state of the surrounding rock; Obtain the total number of remaining sampling points contained in the spliced ​​effective normal monitoring data within the dedicated diagnostic time window for crystallization interference, and calculate the overall variance of the effective normal monitoring data; Determine whether the total number of remaining sampling points is greater than the minimum effective data length threshold, and whether the overall variance is less than the baseline variance threshold; If both of the above conditions are met, the valid and normal monitoring data is deemed qualified; if either condition is not met, the data is deemed unqualified, and a manual review alarm is triggered.

[0032] In some specific embodiments, in the data correction output module, if qualified, curve fitting processing is performed on the valid normal monitoring data, and the actual mechanical deformation data of the surrounding rock after removing crystallization disturbances is output. The specific process is as follows: A two-dimensional coordinate system is constructed with timestamp as the independent variable and displacement value in the valid normal monitoring data as the dependent variable; A nonlinear fitting method (such as the Levenberg-Marquardt algorithm or the Gauss-Newton method) is used to fit the effective normal monitoring data to obtain the optimal fitting parameters and generate a continuous displacement change fitting curve. Using the displacement change fitting curve, numerical interpolation is performed on the time period between the removed mutation start point and the mutation end point to fill in the missing data points. The interpolated and completed data segment is then re-integrated with the valid and normal monitoring data to form a complete data sequence that is time-continuous, without abrupt changes, and conforms to the creep law of rock mechanics. This sequence is then output and stored as the true mechanical deformation data of the surrounding rock after removing crystallization disturbances.

[0033] Combination Figure 3 The diagram shown illustrates the effects of spurious displacement data removal and curve fitting correction. Figure 3 The medium gray dashed line represents the original monitoring data including salt crystallization elevation errors. It is evident that a permanent baseline shift occurred after the step abrupt change. This invention first compensates for the baseline shift by uniformly subtracting the abrupt change amplitude from the data after the abrupt change endpoint, and then removes data segments within the abrupt change interval, obtaining... Figure 3 The blue area represents the valid normal monitoring data scatter points; these valid scatter points are fitted using a nonlinear fitting method (such as the Levenberg-Marquardt algorithm) to generate... Figure 3 The red curve represents the continuous displacement variation fitting curve. The red curve not only smoothly connects the fractured data intervals (interpolation completion), but also perfectly reproduces the true mechanical creep trend of the surrounding rock after removing crystallization disturbances, ensuring the continuity and reliability of the final output data.

[0034] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A dynamic monitoring system for deformation of surrounding rock in tunnels, characterized in that: include: Data acquisition module: Real-time acquisition of overall rock displacement data output by displacement sensor, as well as local humidity data of the micro-area in contact between displacement sensor probe and rock surface; Diagnostic window trigger module: Real-time comparison of the rate of decrease of local humidity data with the evaporation critical threshold. When the rate of decrease of local humidity data exceeds the evaporation critical threshold, a dedicated diagnostic time window for crystallization interference is opened. Waveform feature extraction module: Within the diagnostic time window, acquire the step change waveform in the surrounding rock displacement data, and calculate the rising slope and time-domain and frequency-domain feature parameters of the step change waveform respectively. Crystallization Interference Identification Module: Construct a salt crystallization pushing stress characteristic identification model, input the rising edge slope and frequency domain characteristic parameters into the salt crystallization pushing stress characteristic identification model, and determine whether the current step signal belongs to false displacement data caused by salt crystallization pushing based on the output results of the salt crystallization pushing stress characteristic identification model; Data correction output module: If the data is determined to be false displacement data, the data segment corresponding to the step change in the false displacement data is removed to obtain valid normal monitoring data. The valid normal monitoring data is analyzed for compliance to determine whether it is qualified. If it is qualified, curve fitting processing is performed on the valid normal monitoring data to output the true mechanical deformation data of the surrounding rock after removing crystallization disturbance.

2. The tunnel surrounding rock deformation dynamic monitoring system according to claim 1, characterized in that, The specific acquisition process for the real-time acquisition of the overall displacement data of the surrounding rock output by the displacement sensor, and the local humidity data of the micro-area in contact between the displacement sensor probe and the rock surface, is as follows: Displacement sensors are installed at the monitored section of the tunnel surrounding rock, with the sensor probes tightly attached to the rock surface. The overall displacement data of the surrounding rock is collected in real time by the displacement sensors at a first sampling frequency. In the micro-area where the displacement sensor probes contact the rock surface, a miniature humidity sensor is attached. The local humidity data is collected in real time by the miniature humidity sensor at a second sampling frequency. The collected overall displacement data and local humidity data of the surrounding rock are timestamped to obtain overall displacement data and local humidity data of the surrounding rock with synchronized timestamps.

3. The tunnel surrounding rock deformation dynamic monitoring system according to claim 1, characterized in that, The specific process of comparing the rate of decrease in local humidity data with the critical evaporation threshold in real time is as follows: Acquire local humidity data from multiple sampling periods and construct a sliding calculation window for humidity changes; Within the sliding calculation window for humidity change, the difference between the local humidity data at the current moment and the local humidity data at the beginning of the window is calculated and divided by the time span of the sliding calculation window for humidity change to obtain the rate of decrease of the local humidity data. An evaporation critical threshold is pre-calibrated, and the rate of decrease of local humidity data calculated in real time is compared with the pre-calibrated evaporation critical threshold in real time.

4. The tunnel surrounding rock deformation dynamic monitoring system according to claim 1, characterized in that, The specific process for opening the dedicated diagnostic time window for crystallization interference is as follows: When the comparison result shows that the rate of decrease of local humidity data is greater than the critical threshold for evaporation, it is determined that water evaporation and salt crystallization have occurred. A time trigger signal is generated using the timestamp of the current determination time as the starting time. Based on the time-triggered signal, a dedicated diagnostic time window for crystallization interference is opened, and the overall displacement data of the surrounding rock is acquired during the opening of the dedicated diagnostic time window for crystallization interference.

5. The tunnel surrounding rock deformation dynamic monitoring system according to claim 1, characterized in that, The specific process for obtaining the step abrupt change waveform in the surrounding rock displacement data is as follows: Obtain the overall displacement data of the surrounding rock within the specific diagnostic time window for crystallization interference, and construct a displacement time series; The displacement time series is subjected to first-order difference calculation to obtain the displacement change rate between adjacent sampling points; Set a mutation rate threshold, obtain the displacement change rate of all displacement time series, and when the number of sampling points that continuously exceed the mutation rate threshold reaches the mutation determination point threshold, mark the first sampling point that exceeds the mutation rate threshold as the mutation start point, and mark the sampling points whose displacement change rate falls back to below the mutation rate threshold as the mutation end point. Obtain the overall displacement data segment of the surrounding rock between the start and end points of the mutation, as the step mutation waveform.

6. The tunnel surrounding rock deformation dynamic monitoring system according to claim 1, characterized in that, The specific process for calculating the rising edge slope and the time-domain and frequency-domain characteristic parameters of the step change waveform is as follows: Calculate the rising edge slope: Obtain the displacement value corresponding to the end point of the mutation and the displacement value corresponding to the start point of the mutation, calculate the displacement difference between the two, and divide the displacement difference by the time span from the start point to the end point of the mutation to obtain the rising edge slope. Calculate time-domain characteristic parameters: Perform statistical analysis on the step change waveform, obtain the displacement difference as the change amplitude, obtain the time span as the duration, and calculate the variance of the displacement values ​​of all sampling points in the step change waveform. Combine the change amplitude, duration, and variance to form the time-domain characteristic parameters. Calculate frequency domain characteristic parameters: Perform a fast Fourier transform on the step change waveform to convert the time domain signal into a frequency domain signal; obtain the dominant frequency, the spectral energy ratio of the high-frequency band, and the peak power spectral density in the frequency domain signal, and combine the three to form the frequency domain characteristic parameters.

7. The tunnel surrounding rock deformation dynamic monitoring system according to claim 1, characterized in that, The specific process for constructing the salt crystallization pushing stress characteristic identification model is as follows: Obtain various step-change waveform samples from historical monitoring data of water-rich salt rock strata, including salt crystal jacking interference samples verified by on-site geological exploration, and real samples of sudden deformation of surrounding rock; Obtain the sample rising edge slope, sample time domain feature parameters, and sample frequency domain feature parameters corresponding to each sample, and normalize the obtained feature data. The samples were labeled with classification labels. Salt crystallization pushing interference samples were labeled as positive class labels, and real surrounding rock sudden deformation samples were labeled as negative class labels. The samples were combined to form a training dataset. An initial salt crystallization pushing stress feature identification model was constructed based on the support vector machine algorithm; The training dataset is input into the initial salt crystallization pushing stress feature recognition model for supervised learning training. The sample features are mapped to a high-dimensional space through a kernel function. The model is iteratively optimized and the optimal classification hyperplane that can maximize the distinction between positive and negative samples is found. After the training converges, the network weights and hyperplane parameters are saved to obtain the completed salt crystallization pushing stress feature recognition model.

8. The tunnel surrounding rock deformation dynamic monitoring system according to claim 1, characterized in that, The specific process for determining whether the current step signal is false displacement data caused by salt crystallization push is as follows: The rising edge slope, time-domain feature parameters and frequency-domain feature parameters obtained in real time are normalized and concatenated into a feature vector to be identified. The feature vector to be identified is input into the salt crystallization pushing stress feature identification model; The salt crystallization pushing stress feature recognition model is based on the internal optimal classification hyperplane. It calculates the spatial distance between the feature vector to be identified and the optimal classification hyperplane, and outputs the corresponding classification prediction results. The judgment is made based on the output of the salt crystallization pushing stress characteristic identification model: if the output is a positive label, then the current step change waveform is determined to be false displacement data caused by salt crystallization pushing. If the output is a negative label, it is determined that the current step change waveform does not belong to crystallization interference, but is a real sudden deformation of the surrounding rock.

9. The tunnel surrounding rock deformation dynamic monitoring system according to claim 1, characterized in that, The specific process for determining whether valid and normal monitoring data is qualified is as follows: Based on the marked mutation start and end points, locate the data interval where the step mutation waveform is located in the overall displacement data of the surrounding rock, and obtain the calculated mutation amplitude. The baseline offset compensation is performed by uniformly subtracting the mutation amplitude from all overall displacement data of the surrounding rock after the mutation endpoint. Delete the data segment between the mutation initiation point and the mutation endpoint, and stitch the subsequent data after baseline compensation with the data before the mutation initiation point on the time axis to obtain effective normal monitoring data; Set a minimum effective data length threshold and a baseline variance threshold; Obtain the total number of remaining sampling points contained in the crystallization interference-specific diagnostic time window after splicing of valid normal monitoring data, and calculate the overall variance of valid normal monitoring data; Determine whether the total number of remaining sampling points is greater than the minimum effective data length threshold and whether the overall variance is less than the baseline variance threshold; If both of the above conditions are met, the data is considered valid and normal; if either condition is not met, the data is considered unqualified.

10. The tunnel surrounding rock deformation dynamic monitoring system according to claim 1, characterized in that, The specific process for outputting the true mechanical deformation data of the surrounding rock after removing crystallization disturbances is as follows: A two-dimensional coordinate system is constructed with timestamp as the independent variable and displacement value in valid and normal monitoring data as the dependent variable; The nonlinear fitting method is used to fit the effective normal monitoring data to obtain the optimal fitting parameters and generate a continuous displacement change fitting curve. By using displacement variation fitting curves, numerical interpolation is performed on the time period between the start and end of the eliminated mutation to fill in the missing data points. The interpolated and filled data segments are then re-integrated with the valid normal monitoring data to serve as the true mechanical deformation data of the surrounding rock after removing crystallization disturbances.

Images