Intelligent real-time deformation amount prediction and early warning system for surrounding rock of large deformation tunnel

Abstract

This invention relates to the field of deep neural network technology, specifically to an intelligent real-time deformation prediction and early warning system for surrounding rock in tunnels with large deformation. The system includes a spatial segmentation acquisition module, a directional feature construction module, a segment offset evaluation module, a propagation ranking and deduction module, and an early warning path generation module. In this invention, discrete monitoring points are organized according to three-dimensional coordinates and mileage segments as a thermal basis sequence to construct a spatial segmentation structure. The system calculates the distance between segments and the gradient normalization direction, quantifies the directional differences using vector angles to generate coupling weights, characterizes the deformation propagation characteristics, applies weight correction to time-varying differences to obtain dynamic features, determines the level based on offset thresholds, and identifies the dominant direction by sorting the trend vector and the inner product of the segment centers. Early warning links are divided according to position differences and distance thresholds, constructing a multi-level path network to achieve precise location of risk areas and visualization of propagation paths, and supports rapid deployment of reinforcement measures.

Description

Technical Field

[0001] This invention relates to the field of deep neural network technology, and in particular to an intelligent real-time deformation prediction and early warning system for surrounding rock of tunnels with large deformation. Background Technology

[0002] The field of deep neural networks encompasses multi-layered neural network structures used to process high-dimensional data and recognize complex patterns. The core of this field lies in achieving feature extraction and nonlinear mapping through cascaded neurons in input, hidden, and output layers. The number of hidden layers typically exceeds two to enhance learning capabilities. A systematic overview of the entire technology field is as follows: It originates from an extension of artificial neural networks, emphasizing the backpropagation algorithm to optimize weights and minimize the loss function; the training process involves supervised learning with large amounts of labeled data, or unsupervised learning to extract latent representations; its applications are widespread in image classification, speech recognition, and predictive modeling, using convolutional layers to process spatial data, recurrent layers to process sequential data, or attention mechanisms to improve model efficiency; its development incorporates regularization techniques to prevent overfitting and utilizes graphics processing units (GPUs) to accelerate computation.

[0003] Among them, the intelligent real-time deformation prediction and early warning system for tunnels with large deformation refers to the technology of predicting and issuing warnings about the deformation of the surrounding rock of tunnels using sensor data and neural network models. The technical issues addressed by this patent cover surrounding rock displacement monitoring, pressure distribution analysis, strain change tracking, temperature fluctuation recording, and dynamic water pressure detection. Specifically, it solves the problem by using displacement sensors to collect the movement of the surrounding rock, pressure sensors to obtain the stress value of the rock strata, strain gauges to measure the degree of deformation, temperature probes to record environmental thermal changes, and water pressure gauges to monitor seepage pressure. The system then inputs this data into a pre-trained neural network to calculate the deformation and compares the calculation results with a preset threshold to trigger an alarm signal.

[0004] Existing technologies directly input each monitoring point into the neural network as an independent data source, lacking a systematic analysis of the spatial relationship between monitoring points and the deformation propagation path. This makes it impossible to identify the expansion pattern of surrounding rock deformation in the longitudinal or transverse direction of the tunnel. The original monitoring values ​​are not processed by feature engineering, and the spatial directionality and temporal evolution characteristics are not extracted. The model has difficulty learning the spatial propagation pattern of deformation. When a local stress concentration occurs in a section of the surrounding rock, it is impossible to predict the possibility of stress transmission to adjacent sections. It can only make isolated judgments on single-point anomalies, using a fixed threshold and comparing the calculation results to trigger alarm signals. This ignores the dynamic evolution characteristics of the surrounding rock state. When the deformation fluctuates near the threshold, it may generate frequent false alarms. The alarm signals lack spatial organization and path tracking capabilities. When multiple monitoring points trigger alarms simultaneously, it is impossible to determine whether they belong to the same deformation propagation link, making it difficult to quickly locate the core risk area and the direction of expansion. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and propose an intelligent real-time deformation prediction and early warning system for surrounding rock of large deformation tunnels.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A smart real-time deformation prediction and early warning system for surrounding rock of tunnels with large deformation includes: The spatial segment acquisition module acquires the three-dimensional coordinates, mileage segment number, and real-time temperature stress value of the monitoring points in the surrounding rock of the tunnel. It determines the affiliation of the monitoring points based on the coordinates and segment number, calculates the average temperature stress value for each segment, and constructs the thermal basic sequence of the segment. The directional feature construction module extracts the mean temperature stress and center coordinates based on the segment thermodynamic basic sequence, calculates the inter-segment distance and mean gradient, normalizes the temperature stress direction values ​​respectively, performs vector angle calculation to obtain the direction difference, and compares it with the consistency threshold to generate a directional coupling weight sequence containing weight coefficients. The paragraph offset evaluation module extracts the weight coefficients of each paragraph in the directional coupling weight sequence, calculates the time variation value by combining the average temperature and average stress values ​​of the current and previous moments, performs weighted correction on the time variation value by the weight coefficients, compares it with the offset threshold, determines the offset level, generates a paragraph offset level sequence, and transmits the weighted corrected time variation value. The propagation sorting deduction module calls the weighted and corrected time-varying difference to synthesize a trend vector, performs an inner product operation with the coordinates of the paragraph center, sorts according to the size of the inner product result, and obtains the trend projection sorting sequence; The early warning path generation module filters paragraphs that reach a set threshold based on the trend projection sorting sequence and the paragraph offset level sequence, extracts the position of the filtered paragraph in the sorting, calculates the position difference between adjacent paragraphs and compares it with the continuous distance threshold. When the difference is less than the threshold, adjacent paragraphs are included in the same link. When the difference is greater than or equal to the threshold, the current link is cut off and the current paragraph is used as the starting point of the new link. All sub-links and single-point nodes are summarized to generate a multi-level early warning path network.

[0007] As a further embodiment of the present invention, the segment thermal basis sequence includes segment number, average temperature, and average stress; the directional coupling weight sequence includes segment number, directional value, and weight coefficient; the segment offset level sequence includes segment number and offset level; the trend projection sorting sequence includes segment number and sorting position; and the multi-level early warning path network includes link number, sub-link, and single-point early warning node.

[0008] As a further aspect of the present invention, the comparison result between the directional difference and the consistency threshold is as follows: if the threshold is met, a standard weight of 1.0 is assigned to generate a directional coupling weight sequence; if the threshold is not met, a decay coefficient of 0.3-0.8 is assigned according to the deviation degree; if the deviation exceeds twice the threshold, a minimum coefficient of 0.1 is assigned and an anomaly mark is added, and the weight is reduced and included in the directional coupling weight sequence. The process of cutting off the current link includes: sealing the previous sub-link number, segment number and offset level distribution; if there are no subsequent consecutive segments after the current segment, it is marked as a single-point warning node and assigned a subordinate identifier.

[0009] As a further aspect of the present invention, the calculation steps for the time-varying difference value to undergo weighted correction by weighting coefficients are as follows: (1) The time-varying difference of the standard weighted segment is directly compared with the offset threshold; (2) The time-varying difference needs to be multiplied by the corresponding attenuation coefficient before being compared with the offset threshold.

[0010] As a further aspect of the present invention, the inner product operation steps are as follows: (1) The inner product result of the standard weighted paragraphs directly participates in the ranking; (2) The inner product result needs to be multiplied by the corresponding attenuation coefficient before being sorted.

[0011] As a further aspect of the present invention, the spatial segmentation acquisition module includes: The monitoring data acquisition submodule acquires the three-dimensional coordinates, mileage segment number, and real-time temperature and stress values ​​of the tunnel surrounding rock monitoring point. It decomposes the three-dimensional coordinates into position values ​​in the X, Y, and Z axes, extracts the digital identification code of the mileage segment number, records the temperature and stress values ​​and the acquisition time, binds the values, and generates a spatial attribute set of the monitoring point. The mileage segment allocation submodule calls the spatial attribute set of the monitoring points, extracts the three-dimensional position values ​​of each monitoring point, obtains the preset mileage segment boundary values, and performs interval judgment between the three-dimensional position values ​​of the monitoring points and the corresponding directional boundary values. When the position values ​​of the three directions are all within the interval of a certain mileage segment boundary value, the monitoring point is assigned to that mileage segment, and the segment monitoring point distribution table is obtained. The thermal sequence construction submodule extracts the temperature stress values ​​of all monitoring points within each mileage segment based on the segment monitoring point distribution table, statistically analyzes the temperature stress values ​​within the same mileage segment, obtains the average temperature stress of the mileage segment, arranges the average temperature stress of each mileage segment in order of mileage segment number, and generates the segment thermal basic sequence.

[0012] As a further aspect of the present invention, the direction feature construction module includes: The direction vector extraction submodule extracts the mean temperature stress and center coordinate values ​​of each mileage segment based on the thermal basic sequence of the segment, obtains the coordinate difference value and mean difference value of adjacent mileage segments, converts them to the standard numerical range to obtain the normalized distance value and normalized gradient value, and combines them to form a direction vector to generate a set of temperature stress direction vectors. The vector angle evaluation submodule calls the temperature stress direction vector set, extracts the direction vector values ​​of adjacent mileage segments, uses the vector angle formula to obtain the direction difference, sets a direction consistency threshold, compares the direction difference with the threshold, marks it as consistent when the direction difference is within the threshold range, and marks it as deviating when it exceeds the threshold. Records the direction difference and the marking status to obtain the direction deviation evaluation table. The weight coefficient assignment submodule extracts the direction difference and the marking status according to the direction deviation evaluation table. When the marking is consistent, a standard weight coefficient is assigned. When the marking is deviated and the direction difference does not exceed the threshold multiple limit, an attenuation weight coefficient is assigned within a set interval according to the degree of deviation. When the direction difference exceeds the threshold multiple limit, a minimum weight coefficient is assigned and an abnormality mark is added. The weight coefficients are arranged according to the mileage segment number to generate a direction coupling weight sequence.

[0013] As a further aspect of the present invention, the paragraph offset evaluation module includes: The time-varying difference extraction submodule calls the directional coupling weight sequence to extract the weight coefficients of each mileage segment, obtains the average temperature and average stress values ​​of the current and previous moments, performs difference calculations on the average temperature and average stress values ​​by subtracting the values ​​of the previous moment from the values ​​of the current moment, arranges the difference results according to the mileage segment number, and generates a segment time-varying difference set. The weighted correction submodule extracts the time-varying difference value and corresponding weight coefficient of each mileage segment based on the segment time-varying difference value set and the direction coupled weight sequence. In the standard weight type, the original value of the time-varying difference value is kept. In the attenuation weight type, the time-varying difference value and the attenuation coefficient are multiplied. The processed value is recorded according to the mileage segment number to establish a weighted correction difference value table. The grade determination submodule calls the weighted correction difference table, extracts the weighted correction time-varying difference value of each mileage segment, sets the offset threshold, compares the time-varying difference value with the offset threshold value, marks it as normal grade if it is less than the threshold, and marks it as offset grade if it is greater than or equal to the threshold. The grade marks are arranged according to the mileage segment number, a segment offset grade sequence is generated, and the weighted correction time-varying difference value is transmitted.

[0014] As a further aspect of the present invention, the propagation ordering deduction module includes: The trend vector synthesis submodule calls the weighted and corrected time-varying difference value, extracts the temperature time-varying difference value and stress time-varying difference value of each mileage segment, assigns the temperature time-varying difference value as the first dimension component of the vector, assigns the stress time-varying difference value as the second dimension component of the vector, combines the two components according to the mileage segment number to construct a two-dimensional vector, pairs the two-dimensional vector with the segment center coordinates, and generates a trend vector coordinate set. The inner product operation correction submodule extracts the two-dimensional vector components and center coordinate values ​​of each mileage segment based on the trend vector coordinate set and the direction coupled weight sequence. It performs a product operation on the first dimension component and the horizontal coordinate value, and a product operation on the second dimension component and the vertical coordinate value. It performs a sum operation on the two product results to obtain the inner product value. In the standard weight type, the original value of the inner product value is maintained. In the attenuation weight type, the inner product value and the attenuation coefficient are multiplied. The processed values ​​are recorded according to the mileage segment number to establish a corrected inner product value table. The projection sorting generation submodule calls the corrected inner product value table, extracts the corrected inner product values ​​of each mileage segment, performs a descending sorting operation on the inner product values ​​according to their numerical values, arranges the mileage segment identifiers according to their position, and generates a trend projection sorting sequence.

[0015] As a further aspect of the present invention, the early warning path generation module includes: The threshold filtering and positioning submodule extracts the position number and offset level identifier of each mileage segment based on the trend projection sorting sequence and the segment offset level sequence. It compares the offset level identifier with the set offset threshold, filters the mileage segments whose offset level identifier values ​​are greater than or equal to the offset threshold, arranges the filtered mileage segment identifiers in order of position, and generates a threshold-compliant segment position table. The link continuity determination submodule calls the threshold-compliant segment ranking table, extracts the ranking number of each mileage segment, selects adjacent mileage segment ranking numbers and performs a subtraction operation to obtain the ranking difference, and compares it with the continuous distance threshold. When the ranking difference is less than the continuous distance threshold, the adjacent mileage segments are included in the same link and assigned the same link number. When the ranking difference is greater than or equal to the continuous distance threshold, the current link is cut off and the previous sub-link number, segment number and offset level distribution are sealed. The current mileage segment is taken as the starting point of the new link and assigned a new link number. The existence of subsequent continuous mileage segments of the current mileage segment is determined. If they do not exist, they are marked as single-point warning nodes and assigned a subordinate identifier. A hierarchical link node set is established. The network aggregation and construction submodule calls the hierarchical link node set, extracts the link number, link mileage segment identifier and single-point early warning node identifier, arranges the link mileage segment identifier by link number, arranges the single-point early warning node identifier by subordinate identifier, and combines all links and single-point nodes to generate a multi-level early warning path network.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, discrete monitoring points are organized into a thermal basis sequence according to three-dimensional coordinates and mileage segments to construct a spatial segmented structure. The system calculates the distance between segments and the gradient normalization direction, and quantifies the directional differences by vector angle, thereby generating directional coupling weights to characterize deformation propagation characteristics. The time-varying difference is weighted to obtain dynamic evolution characteristics. The level is determined according to the offset threshold, and the dominant direction is identified by sorting the inner product of the trend vector and the segment center. The early warning link is divided according to the position difference and distance threshold, and the discrete signal is constructed into a multi-level path network, thereby realizing the accurate positioning of the risk area and the visualization of the propagation path, supporting the rapid deployment of reinforcement measures. Attached Figure Description

[0017] Figure 1 This is a system flowchart of the present invention; Figure 2 This is a flowchart of the overall modules of the present invention; Figure 3 This is a system block diagram of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] Please see Figures 1-2 The intelligent real-time deformation prediction and early warning system for surrounding rock of tunnels with large deformation includes: The spatial segment acquisition module acquires the three-dimensional coordinates, mileage segment number, and real-time temperature stress value of the monitoring points in the surrounding rock of the tunnel. It determines the affiliation of the monitoring points based on the coordinates and segment number, calculates the average temperature stress value for each segment, and constructs the thermal basic sequence of the segment. The directional feature construction module is based on the segment thermal basic sequence. It extracts the mean temperature stress and center coordinates, calculates the inter-segment distance and mean gradient, normalizes the temperature stress directional values, performs vector angle calculation to obtain the directional difference, and compares it with the consistency threshold to generate a directional coupling weight sequence containing weight coefficients. The comparison results between the directional difference and the consistency threshold are as follows: if the threshold is met, a standard weight of 1.0 is assigned to generate a directional coupling weight sequence; if the threshold is not met, a decay coefficient of 0.3-0.8 is assigned according to the deviation. When the deviation exceeds twice the threshold, a minimum coefficient of 0.1 is assigned and an anomaly label is added, and the weight is reduced and included in the directional coupling weight sequence. The paragraph offset assessment module extracts the weight coefficients of each paragraph in the directional coupling weight sequence, calculates the time variation value by combining the average temperature and average stress values ​​of the current and previous moments, performs weighted correction on the time variation value by the weight coefficients, compares it with the offset threshold, determines the offset level, generates a paragraph offset level sequence, and transmits the weighted corrected time variation value. The calculation steps for weighted correction of time-varying differences are as follows: (1) The time-varying difference of the standard weighted segment is directly compared with the offset threshold; (2) The time-varying difference needs to be multiplied by the corresponding attenuation coefficient before being compared with the offset threshold; The propagation sorting deduction module calls the weighted and corrected time-varying difference to synthesize a trend vector, performs an inner product operation with the coordinates of the paragraph center, sorts according to the size of the inner product result, and obtains the trend projection sorting sequence; The steps for inner product operation are as follows: (1) The inner product result of the standard weighted paragraphs directly participates in the ranking; (2) The inner product result needs to be multiplied by the corresponding attenuation coefficient before being used for sorting; The early warning path generation module filters paragraphs that reach a set threshold based on the trend projection sorting sequence and paragraph offset level sequence, extracts the position of the filtered paragraph in the sorting, calculates the position difference between adjacent paragraphs and compares it with the continuous distance threshold. When the difference is less than the threshold, adjacent paragraphs are included in the same link. When the difference is greater than or equal to the threshold, the current link is cut off and the current paragraph is used as the starting point of the new link. All sub-links and single-point nodes are summarized to generate a multi-level early warning path network. The process of cutting off the current link includes: sealing the previous sub-link number, segment number and offset level distribution; if there are no subsequent consecutive segments after the current segment, it is marked as a single-point warning node and assigned a subordinate identifier. The paragraph thermal foundation sequence includes paragraph number, average temperature, and average stress; the directional coupling weight sequence includes paragraph number, directional value, and weight coefficient; the paragraph offset level sequence includes paragraph number and offset level; the trend projection sorting sequence includes paragraph number and sorting position; and the multi-level early warning path network includes link number, sub-link, and single-point early warning node.

[0020] Please see Figure 3 The spatial segmentation acquisition module includes a monitoring data acquisition submodule, a mileage segment attribution and allocation submodule, and a thermal sequence construction submodule; The monitoring data acquisition submodule acquires the three-dimensional coordinates, mileage segment number, and real-time temperature and stress values ​​of the tunnel surrounding rock monitoring point. It decomposes the three-dimensional coordinates into position values ​​in the X, Y, and Z axes, extracts the digital identification code of the mileage segment number, records the temperature and stress values ​​and the acquisition time, binds the values, and generates a spatial attribute set of the monitoring point. Specifically, the monitoring data acquisition submodule first obtains the three-dimensional coordinates, mileage segment number, and real-time temperature and stress values ​​of the tunnel surrounding rock monitoring points from the sensors. Then, the three-dimensional coordinates are decomposed. For a monitoring point at the left arch waist of the tunnel, its spatial coordinates are divided into three independent position values: the longitudinal position along the tunnel axis (X-axis), the transverse position perpendicular to the tunnel axis (Y-axis), and the vertical elevation position (Z-axis). For the mileage segment number, the numerical identifier is extracted, and the current temperature and stress value and acquisition time are recorded. The X-axis position is bound to the monitoring point number, the Y-axis position is bound to the monitoring point number, the Z-axis elevation is bound to the monitoring point number, and the mileage segment numerical identifier is bound to the monitoring point number. The temperature and stress values ​​are then linked. The force value is bound to the monitoring point number, and the acquisition time is also bound to the monitoring point number to form a complete attribute record for the monitoring point. Similarly, the spatial coordinates of another monitoring point at the right arch waist of the tunnel are divided into the longitudinal position of the X-axis, the lateral position of the Y-axis, and the elevation position of the Z-axis. The numerical identification code in its mileage segment number is extracted, and the temperature stress value and acquisition time are recorded. Each value is bound to the monitoring point number one by one. The coordinates of the monitoring point at the arch crown are then divided into the position values ​​of the three axes, the numerical identification code is extracted, and the temperature stress value and acquisition time are recorded. The binding of each value to the monitoring point number is completed. The longitudinal position, lateral position, elevation position, numerical identification code, temperature stress value, and acquisition time of all monitoring points are combined to form an associated set of monitoring point spatial attributes. Assumption: 3D coordinates of the tunnel monitoring point: ; Monitoring point at the right arch waist of the tunnel: ; Tunnel arch location monitoring point: ; The three-dimensional coordinates of these monitoring points were decomposed into Values ​​in three directions; Disassembly process: : Vertical position (X-axis direction): 0; Lateral position (Y-axis direction): 0; Elevation position (Z-axis direction): 5; Mileage segment number: 1; Temperature stress value: 50; Data collection time: 2025-11-25 10:00:00; Binding records: It is bound to monitoring point number 1; It is bound to monitoring point number 1; It is bound to monitoring point number 1; Mileage segment number 1 is linked to monitoring point number 1; The temperature stress value of 50 is associated with monitoring point number 1. The data collection time is linked to monitoring point number 1; : Vertical position (X-axis direction): 3; Lateral position (Y-axis direction): 4; Elevation position (Z-axis direction): 5; Mileage segment number: 1; Temperature stress value: 55; Data collection time: 2025-11-25 10:05:00; Binding records: It is linked to monitoring point number 2; It is linked to monitoring point number 2; It is linked to monitoring point number 2; Mileage segment number 1 is linked to monitoring point number 2; The temperature stress value of 55 is associated with monitoring point number 2. The data collection time is linked to monitoring point number 2; Dome position monitoring point : Vertical position (X-axis direction): 6; Lateral position (Y-axis direction): 8; Elevation position (Z-axis direction): 5; Mileage segment number: 2; Temperature stress value: 60; Data collection time: 2025-11-25 10:10:00; Binding records: It is linked to monitoring point number 3; It is linked to monitoring point number 3; It is linked to monitoring point number 3; Mileage segment number 2 is linked to monitoring point number 3; The temperature stress value of 60 is associated with monitoring point number 3. The data collection time is linked to monitoring point number 3; The final generated spatial attribute set of monitoring points includes: Monitoring point 1: Mileage segment number 1, temperature stress value 50, data collection time 2025-11-25 10:00:00; Monitoring point 2: Mileage segment number 1, temperature stress value 55, collection time 2025-11-25 10:05:00; Monitoring point 3: Mileage segment number 2, temperature stress value 60, collection time 2025-11-25 10:10:00; The mileage segment allocation submodule calls the spatial attribute set of monitoring points, extracts the three-dimensional position values ​​of each monitoring point, obtains the preset mileage segment boundary values, and performs interval judgment between the three-dimensional position values ​​of the monitoring points and the corresponding directional boundary values. When the position values ​​of the three directions are all within the interval of a certain mileage segment boundary value, the monitoring point is assigned to that mileage segment, and the segment monitoring point distribution table is obtained. Specifically, the system uses the mileage segment allocation submodule to call the spatial attribute set of monitoring points, extract the longitudinal, lateral, and elevation positions of a monitoring point at the tunnel entrance section, and obtain the longitudinal start point, longitudinal end point, lateral left boundary, lateral right boundary, elevation bottom boundary, and elevation top boundary of the first mileage segment. The longitudinal position is compared with the longitudinal start and end points to determine if it falls within the longitudinal range of the first mileage segment. The lateral position is compared with the left and right boundaries to determine if it falls within the lateral range of the first mileage segment. The elevation position is compared with the bottom and top boundaries to determine if it falls within the elevation range of the first mileage segment. This confirms the three dimensions. If all three directions meet the requirements of the first mileage segment, the monitoring point is assigned to the first mileage segment. The three-axis position of another monitoring point in the middle section of the tunnel is extracted and compared with the boundaries of each direction of the first mileage segment. If all three directions are within the range, the monitoring point is assigned to the first mileage segment. The longitudinal position of a monitoring point in the tunnel exit section is extracted and compared with the longitudinal endpoint of the first mileage segment. If it is found to be outside the range, the boundary value of the second mileage segment is called. The three-axis position of the monitoring point is compared with the boundaries of each direction of the second mileage segment. If all three directions are within the range of the second mileage segment, the monitoring point is assigned to the second mileage segment. The monitoring points contained in the first mileage segment and the monitoring points contained in the second mileage segment are summarized to obtain the segment monitoring point distribution table. Assumption: Boundary of the first mileage segment: Vertical starting point: 0; Longitudinal endpoint: 10; Horizontal left boundary: -5; Horizontal right boundary: 5; Bottom elevation boundary: 0; Elevation top boundary: 10; Attribution determination: Monitoring point 1 (location) ): Comparing the longitudinal position 0 with the longitudinal start point 0 and the longitudinal end point 10, it is within the range; The lateral position 0 is within the range when compared with the lateral boundaries -5 and 5. Elevation location 5 is within the range compared to elevation boundaries 0 and 10; Classification: First mileage segment; Monitoring point 2 (location) ): Comparing longitudinal position 3 with longitudinal start point 0 and longitudinal end point 10, it is within the range; Comparing lateral position 4 with lateral boundaries -5 and 5, it falls within the range; Elevation location 5 is within the range compared to elevation boundaries 0 and 10; Attribution: First mileage segment; Monitoring point 3 (location) ): Comparing the longitudinal position 6 with the longitudinal start point 0 and the longitudinal end point 10, it is within the range; Comparing the lateral position 8 with the lateral boundaries -5 and 5, it exceeds the right boundary; Category: Second Mileage Segment

[0021] The final generated table of paragraph monitoring point distribution is as follows: First mileage segment: Monitoring point 1, Monitoring point 2; Second mileage segment: Monitoring point 3; The thermal sequence construction submodule extracts the temperature stress values ​​of all monitoring points in each mileage segment based on the segment monitoring point distribution table, statistically analyzes the temperature stress values ​​in the same mileage segment, obtains the average temperature stress of the mileage segment, arranges the average temperature stress of each mileage segment in order of mileage segment number, and generates the segment thermal basic sequence. Specifically, the thermal sequence construction submodule, based on the segment monitoring point distribution table, extracts the temperature stress values ​​of each monitoring point within the first mileage segment, performs cumulative calculation, counts the total number of monitoring points, and divides the cumulative result by the total number to obtain the average temperature stress of the first mileage segment. It then extracts the temperature stress values ​​of each monitoring point within the second mileage segment, performs cumulative calculation, counts the total number of monitoring points, and divides the cumulative result by the total number to obtain the average temperature stress of the second mileage segment. Similarly, it extracts the temperature stress values ​​of each monitoring point within the third mileage segment, performs cumulative calculation, counts the total number of monitoring points, and divides the cumulative result by the total number to obtain the average temperature stress of the third mileage segment. Finally, it arranges the average temperature stress values ​​of the first, second, and third mileage segments sequentially according to the mileage segment number, forming a numerical sequence of the average temperature stress values ​​of each mileage segment, thus generating the segment thermal basic sequence. Assumption: Temperature stress value for the first mileage section: 50; Cumulative result: 50; Total number of monitoring points: 1; Average temperature stress: ; Temperature stress value for the second mileage section: 55; Cumulative result: 55; Total number of monitoring points: 1; Average temperature stress: ; Temperature stress value for the second mileage section: 60; Cumulative result: 60; Total number of monitoring points: 1; Average temperature stress: ; Generate the basic thermodynamic sequence of the paragraph: Average temperature stress in the first mileage section: ; Average temperature stress in the second mileage section: ; Average temperature stress in the third mile section: ; The final thermodynamic sequence is as follows: ; Final result Average temperature stress in the first mileage section: ; Average temperature stress in the second mileage section: ; Average temperature stress in the third mile section: ; Basic thermodynamic sequence of paragraphs: ; Please see Figure 3 The orientation feature construction module includes an orientation vector extraction submodule, a vector angle evaluation submodule, and a weight coefficient assignment submodule. The direction vector extraction submodule is based on the thermal basic sequence of the segment. It extracts the average temperature stress and center coordinate values ​​of each mileage segment, obtains the coordinate difference and average difference values ​​of adjacent mileage segments, converts them to the standard numerical range to obtain normalized distance values ​​and normalized gradient values, and combines them to form a direction vector to generate a set of temperature stress direction vectors. Specifically, the direction vector extraction submodule, based on the segment thermal fundamental sequence, extracts the average temperature stress of the first mileage segment, the average temperature stress of the second mileage segment, and the average temperature stress of the third mileage segment. It also extracts the center coordinates of each mileage segment, subtracts the center coordinates of adjacent mileage segments to obtain the coordinate difference value, subtracts the average temperature stress of adjacent mileage segments to obtain the mean difference value, calculates the maximum value among all coordinate difference values, divides each coordinate difference value by the maximum value to obtain the normalized distance value, calculates the maximum value among the absolute values ​​of all mean difference values, divides each mean difference value by the maximum value to obtain the normalized gradient value, combines the normalized distance value and the normalized gradient value to form a direction vector, records the direction vector values ​​of each adjacent mileage segment, and generates a set of temperature stress direction vectors. Assumption: 1. Extract the average temperature stress. Average temperature stress in the first mileage section ; Average temperature stress in the second mileage section ; Average temperature stress in the third mile section ; 2. Extract the center coordinates of each mileage segment. Center coordinates of the first mileage segment ; Center coordinates of the second mileage segment ; Center coordinates of the third mile segment ; 3. Calculate the coordinate difference value (1) Coordinate differences between the first and second mileage segments: Coordinate differences between the first and second mileage segments: ; ; (2) Coordinate differences between the second and third mileage segments: ; ; 4. Calculate the mean difference in temperature stress. (1) Difference between the mean values ​​of the first and second mileage segments: ; (2) Difference in mean values ​​between the second and third mileage segments: ; 5. Calculate the maximum coordinate difference and the maximum mean difference. (1) Maximum coordinate difference: , ; (2) Maximum mean difference: , ; 6. Calculate the normalized coordinate difference value (1) Normalized value of the coordinate difference between the first and second mileage segments: ; (2) Normalized values ​​of the coordinate differences between the second and third mileage segments: ; 7. Calculate the difference in normalized mean temperature stress. (1) Normalized value of the difference between the means of the first and second mileage segments: ; (2) Normalized value of the difference between the mean values ​​of the second and third mileage segments: ; 8. Forming direction vectors (1) Direction vectors of the first and second mileage segments: ; (2) Direction vectors for the second and third mileage segments: ; 9. Generate a set of direction vectors The set of vectors representing the direction of temperature stress is: ; The vector angle evaluation submodule calls the temperature stress direction vector set, extracts the direction vector values ​​of adjacent mileage segments, uses the vector angle formula to obtain the direction difference, sets a direction consistency threshold, compares the direction difference with the threshold, marks it as consistent when the direction difference is within the threshold range, and marks it as deviating when it exceeds the threshold. Records the direction difference and the marking status to obtain the direction deviation evaluation table. Specifically, the vector angle evaluation submodule calls the temperature stress direction vector set, extracts the direction vectors of adjacent mileage segments, multiplies the x-coordinates of the two direction vectors, multiplies the y-coordinates of the two direction vectors, adds the two products to obtain the vector dot product, adds the square root of the square of the x-coordinate and y-coordinate of the first direction vector to obtain the modulus, adds the square of the square of the x-coordinate and y-coordinate of the second direction vector and takes the square root to obtain the modulus, multiplies the two vector modulus to obtain the modulus product, divides the vector dot product by the modulus product to obtain the cosine value, performs an inverse cosine operation on the cosine value to obtain the radian value, converts the radian value to an angle value to obtain the direction difference, sets a direction consistency threshold, compares the direction difference with the threshold to determine whether it exceeds the threshold range, marks it as deviation when it exceeds the threshold range, marks it as consistency when it is within the threshold range, records the direction difference and the marking status, and obtains the direction deviation evaluation table; Assumption: 1. Extract the direction vectors of adjacent mileage segments. The set of vectors representing the direction of temperature stress is: ; Direction vectors of two adjacent mileage segments: ; ; 2. Calculate the vector dot product According to the formula: ; Substitute the values: ; 3. Calculate the magnitude of each direction vector. Modulus formula: ; For the first direction vector : ; For the second direction vector : ; 4. Calculate the product of modulo lengths Modulus product: ; 5. Calculate the cosine value Cosine value formula: ; 6. Calculate the radian value Using inverse cosine operation: ; 7. Calculate the angle value Convert radians to degrees: ; 8. Set a directional consistency threshold Assuming the directional consistency threshold is (This value needs to be set as a threshold based on the actual situation); 9. Determine the consistency of direction Compare the directional difference with the threshold: ; 10. Record the direction difference and marker status. Directional difference: ; Status: Consistent; result: Directional difference is The direction consistency is set to consistent, and then the direction difference and the marking status are recorded to obtain the direction deviation evaluation table; The weight coefficient assignment submodule extracts the direction difference and the marking status based on the direction deviation evaluation table. When the marking is consistent, a standard weight coefficient is assigned. When the marking is deviated and the direction difference does not exceed the threshold multiple limit, an attenuation weight coefficient is assigned within a set interval according to the degree of deviation. When the direction difference exceeds the threshold multiple limit, a minimum weight coefficient is assigned and an abnormality mark is added. The weight coefficients are arranged according to the mileage segment number to generate a direction coupling weight sequence. The weight coefficient assignment submodule extracts the direction difference and marking status based on the direction deviation evaluation table. When the marking is consistent, a standard weight coefficient is assigned. When the marking is deviated, the direction consistency threshold is obtained, a threshold multiplier limit is set, the threshold is multiplied by the multiplier limit to obtain the upper limit value, the direction difference is compared with the upper limit value, and if it is determined that it does not exceed the upper limit value, the upper limit value is subtracted from the threshold to obtain the deviation range, the direction difference is subtracted from the threshold to obtain the actual deviation amount, the actual deviation amount is divided by the deviation range to obtain the deviation ratio, the standard weight coefficient is subtracted from the minimum weight coefficient to obtain the weight change range, the weight change range is multiplied by the deviation ratio to obtain the weight attenuation amount, the standard weight coefficient is subtracted from the weight attenuation amount to obtain the attenuation weight coefficient, the attenuation weight coefficient is assigned to the mileage segment, and if it is determined that the direction difference exceeds the upper limit value, the minimum weight coefficient is assigned and an anomaly mark is attached. The weight coefficients are arranged according to the mileage segment number to generate a direction coupling weight sequence. Assumption: Directional difference ; Consistency in direction: consistent; 1. Mark the direction for consistency and set weighting coefficients. When the marking is consistent: Assign standard weighting coefficients ; When marked as a deviation, additional steps are required to calculate the attenuation weighting factor. 2. Obtain the directional consistency threshold and set the threshold multiple limit. Directional consistency threshold: ; Threshold multiple limit: ; Upper limit: Multiplying the directional consistency threshold by the multiplier limit yields the upper limit: ; 3. Determine the relationship between the direction difference and the upper limit value. Directional difference: ; Compare: ; 4. Calculate the deviation range and the actual deviation. Deviation from the range: ; Actual deviation: ; A negative actual deviation indicates good directional consistency. 5. Calculate the deviation ratio Deviation ratio: ; Since the deviation amount is negative, the deviation ratio is also negative. We take the absolute value to ensure that the ratio is positive. ; 6. Calculate the weight decay. : ; Weight variation range: ; Weight decay: ; 7. Calculate the attenuation weighting coefficient. Attenuation weighting coefficient: ; 8. Determine if the direction difference exceeds the upper limit. Because of the direction difference Less than the upper limit It will not be marked as an anomaly, so it is not assigned the minimum weight coefficient, but rather the decayed weight coefficient. ; 9. Arrange the weighting coefficients according to the mileage segment number. Assuming there is only one mileage segment, the weighting coefficient is: ; 10. Generate directional coupling weight sequence The final directional coupling weight sequence is: ; result: Directional difference: ; Mark as consistent and assign a decay weight coefficient. ; The final directional coupling weight sequence is ; Please see Figure 3The paragraph offset evaluation module includes a time-varying difference extraction submodule, a weighted correction submodule, and a grade determination submodule; The time-varying difference extraction submodule calls the directional coupling weight sequence to extract the weight coefficients of each mileage segment, obtains the average temperature and average stress values ​​of the current and previous moments, performs the difference operation on the average temperature and average stress values ​​by subtracting the value of the previous moment from the value of the current moment, arranges the difference results by mileage segment number, and generates a segment time-varying difference set. Specifically, the time-varying difference extraction submodule calls the directional coupling weight sequence to extract the weight coefficients of each mileage segment, calls the current temperature database to extract the current average temperature of each mileage segment, calls the previous temperature database to extract the previous average temperature of each mileage segment, performs temperature mean difference calculation for each mileage segment, using the current average temperature as the minuend and the previous average temperature as the subtrahend, and performs subtraction to obtain the temperature difference, calls the current stress database to extract the current average stress of each mileage segment, calls the previous stress database to extract the previous average stress of each mileage segment, performs stress mean difference calculation for each mileage segment, using the current average stress as the minuend and the previous average stress as the subtrahend, and performs subtraction to obtain the stress difference, arranges the temperature difference and stress difference in order of mileage segment number, and generates a segment time-varying difference set; Assumption: 1. Directional Coupled Weight Sequence ; 2. Average temperature at the current moment The average temperature for each mileage segment was extracted from the temperature database at the current moment: Average temperature at the current moment for the first mileage segment ; Average temperature at the current moment for the second mileage segment ; Average temperature at the current moment in the third mile segment ; 3. Average temperature at the previous moment The average temperature for each mileage segment was extracted from the temperature database of the previous moment: Average temperature at the moment before the first mileage segment ; Average temperature at the moment before the second mile section ; Average temperature at the moment before the third mile segment ; 4. Temperature difference calculation For each mileage segment, calculate the difference between the average temperature at the current time and the average temperature at the previous time: Temperature difference in the first mileage segment: ; Temperature difference in the second mileage segment: ; Temperature difference in the third mile segment: ; 5. Average stress at the current moment The average stress value for each mileage segment was extracted from the stress database at the current moment: Average stress at the current moment for the first mileage segment ; Average stress at the current moment for the second mileage section ; Average stress at the current moment in the third mile segment ; 6. Average stress at the previous moment The average stress value for each mileage segment was extracted from the stress database of the previous moment: Average stress at the previous moment in the first mileage segment ; Average stress at the previous moment in the second mile section ; Average stress at the previous moment in the third mile section ; 7. Stress Difference Calculation For each mileage segment, calculate the difference between the average stress at the current time and the previous time: Stress difference in the first mileage section: ; Stress difference in the second mileage section: ; Stress difference in the third mileage section: ; 8. Generate paragraph time variation set Arrange the temperature difference and stress difference values ​​in order of mileage segment number to obtain the segment time variation value set: ; Substitute parameters: ; Final result: The temperature difference and stress difference in the first mileage section are: ; The temperature difference and stress difference in the second mileage section are... ; The temperature difference and stress difference in the third mile section are: ; The time variation set of the paragraph is: ; The weighted correction submodule extracts the time-varying difference value and corresponding weight coefficient of each mileage segment based on the segment time-varying difference value set and the direction coupled weight sequence. In the standard weight type, the original value of the time-varying difference value is kept. In the attenuation weight type, the time-varying difference value and the attenuation coefficient are multiplied. The processed value is recorded according to the mileage segment number to establish a weighted correction difference value table. Specifically, the weighted correction submodule extracts the time-varying difference value and corresponding weight coefficient of each mileage segment based on the segment time-varying difference value set and the direction-coupled weight sequence. It calls the weight coefficient of each mileage segment and the standard weight value to perform an equality judgment. If the judgment result is equal, it is confirmed as the standard weight type. The original values ​​of temperature difference and stress difference remain unchanged, and the weighted correction value is recorded. If the judgment result is unequal, it is confirmed as the attenuation weight type. For the temperature difference, a weighted correction operation is performed, using the temperature difference as the multiplicand and the attenuation weight coefficient as the multiplier. The weighted correction temperature difference value is obtained by performing a multiplication operation. For the stress difference, a weighted correction operation is performed, using the stress difference as the multiplicand and the attenuation weight coefficient as the multiplier. The weighted correction temperature difference and stress difference value are arranged in order of mileage segment number to establish a weighted correction difference table. Assumption: The time variation set of the paragraph is: ; The directional coupling weight sequence is: ; 1. Extracting time variation values ​​and weighting coefficients Extract the temperature difference and stress difference for each mileage segment from the time variation value set: ; ; ; ; ; ; Simultaneously extract the corresponding weight coefficients ( ): First Mileage Segment Weighting Coefficient ; Second Mileage Segment Weighting Coefficient ; Third Mileage Segment Weighting Coefficient ; 2. Equality judgment between weighting coefficients and standard weight values According to the settings, the standard weighting coefficient is Compare it with the attenuation weighting coefficient, because and The mileage segments are not equal, therefore these mileage segments will be classified as having a depreciation weight type; 3. Weighted Correction Calculation For different attenuation weight types, it is necessary to perform weighted correction on the temperature difference and stress difference. The correction process involves multiplying the difference by the attenuation weighting factor: Weighted temperature difference: ; ; ; Weighted stress difference: ; ; ; 4. Generate a weighted corrected difference table The weighted temperature difference and stress difference are arranged in order of mileage segment number to obtain the weighted correction difference table: ; Substitute parameters: ; Final result The weighted adjusted temperature difference and stress difference for the first mileage segment are: ; The weighted adjusted temperature difference and stress difference for the second mileage segment are: ; The weighted temperature difference and stress difference for the third mileage segment are: ; The weighted adjusted difference table is as follows: ; The grade determination submodule calls the weighted correction difference table, extracts the weighted correction time-varying difference value of each mileage segment, sets the offset threshold, compares the time-varying difference value with the offset threshold value, marks it as normal grade if it is less than the threshold, and marks it as offset grade if it is greater than or equal to the threshold. The grade marks are arranged according to the mileage segment number, a segment offset grade sequence is generated, and the weighted correction time-varying difference value is passed. Specifically, the grade determination submodule calls the weighted correction difference table, extracts the weighted correction temperature difference and stress difference for each mileage segment, sets temperature offset threshold and stress offset threshold, performs grade determination for the temperature difference of each mileage segment, compares the weighted correction temperature difference with the temperature offset threshold, marks the temperature level as normal if it is less than the threshold, and marks the temperature level as offset if it is greater than or equal to the threshold. For the stress difference of each mileage segment, it performs grade determination, compares the weighted correction stress difference with the stress offset threshold, marks the stress level as normal if it is less than the threshold, and marks the stress level as offset if it is greater than or equal to the threshold. The grade labels are arranged in order of mileage segment number, the weighted correction temperature difference and stress difference of each mileage segment are transmitted, a segment offset grade sequence is generated and the weighted correction time-varying difference is transmitted. Assumption: ; Set threshold: ; Stress offset threshold ; 1. Extract the weighted corrected temperature difference and stress difference values. Extract the temperature difference and stress difference for each mileage segment from the weighted correction difference table: ; 2. Determination of Temperature Difference Level The temperature difference for each mileage segment is used to determine the severity level: ; judge: Marked as normal level; Second mileage segment: ; judge: Marked as normal level; Third mileage segment: ; judge: Marked as normal level; 3. Determination of stress difference level The stress difference value for each mileage segment is used to determine the stress level. ; judge: Marked as normal level; Second mileage segment: ; judge: Marked as offset level; Third mileage segment: ; judge: Marked as offset level; 4. Generate paragraph offset level sequence Generate a segment offset level sequence based on the temperature and stress level markings for each mileage segment; Arranged in mileage segment order, the following sequence is obtained: First mileage segment: Temperature level normal, stress level normal. Second mileage segment: Temperature level normal level, stress level deviation level. Third mileage segment: Temperature level normal level, stress level deviation level. The final paragraph offset level sequence is as follows: ; Final result First mileage segment: Temperature level normal, stress level normal. Second mileage segment: Temperature level normal level, stress level deviation level. Third mileage segment: Temperature level normal level, stress level deviation level. Therefore, the paragraph offset level sequence is as follows: ; Please see Figure 3 The propagation sorting deduction module includes a trend vector synthesis submodule, an inner product operation correction submodule, and a projection sorting generation submodule; The trend vector synthesis submodule calls the weighted and corrected time-varying difference value, extracts the temperature time-varying difference value and stress time-varying difference value of each mileage segment, assigns the temperature time-varying difference value as the first dimension component of the vector, assigns the stress time-varying difference value as the second dimension component of the vector, combines the two components according to the mileage segment number to construct a two-dimensional vector, pairs the two-dimensional vector with the segment center coordinates, and generates a trend vector coordinate set. Specifically, the trend vector synthesis submodule calls the weighted and corrected time-varying difference value, extracts the temperature time-varying difference value and stress time-varying difference value for each mileage segment from the weighted and corrected difference value table, assigns the temperature time-varying difference value to the first dimension component of the vector for each mileage segment, assigns the stress time-varying difference value to the second dimension component of the vector, performs a two-dimensional vector combination operation, calls the value of the first dimension component of the vector, calls the value of the second dimension component of the vector, places the first dimension component at the beginning of the vector, places the second dimension component at the second position of the vector, and constructs a two-dimensional vector by combining them in order of position. It extracts the center coordinate value of each mileage segment from the segment thermal basis sequence, performs a pairing operation between the two-dimensional vector and the center coordinate for each mileage segment, calls the two-dimensional vector as the vector parameter, calls the center coordinate value as the coordinate parameter, establishes an association between the vector parameter and the coordinate parameter, records the pairing results, arranges each pairing result in order of mileage segment number, and generates a trend vector coordinate set. Assumption: ; Each tuple represents the weighted and corrected temperature and stress differences for a mileage segment; The center coordinates of each mileage segment in the thermal foundation sequence are set as follows: ; ; Center coordinates of the third mile segment ; 1. Extract the weighted adjusted difference Extract the temperature difference and stress difference for each mileage segment from the weighted correction difference table: ; ; ; 2. Construct a two-dimensional vector For each mileage segment, the temperature difference is assigned to the first dimension of the vector, and the stress difference is assigned to the second dimension of the vector, thus constructing a two-dimensional vector: ; Second mileage segment: ; Third mileage segment: ; 3. Extract center coordinates Extract the center coordinates of each mileage segment from the segment's thermal baseline sequence: ; Center coordinates of the second mileage segment ; Center coordinates of the third mile segment ; 4. Pairing two-dimensional vectors with center coordinates Pair the two-dimensional vector of each mileage segment with its corresponding center coordinates: ; Second mileage segment: ; Third mileage segment: ; 5. Establish relationships and record pairing results. For each mileage segment, establish a relationship between the two-dimensional vector and the corresponding coordinates, and record the pairing results: The pairing result for the first mileage segment is ; The pairing result for the second mileage segment is ; The pairing results for the third mile segment are as follows: ; 6. Generate trend vector coordinate set Arrange the paired results in mileage segment number order to generate a trend vector coordinate set: ; Substitute parameters: ; Final result The trend vector coordinate set is: ; The inner product operation correction submodule is based on the trend vector coordinate set and the direction coupled weight sequence. It extracts the two-dimensional vector components and center coordinate values ​​of each mileage segment, performs a product operation on the first dimension component and the horizontal coordinate value, and performs a product operation on the second dimension component and the vertical coordinate value. It then performs a sum operation on the two product results to obtain the inner product value. In the standard weight type, the original value of the inner product value is maintained. In the attenuation weight type, the inner product value and the attenuation coefficient are multiplied. The processed values ​​are recorded according to the mileage segment number, and a corrected inner product value table is established. Specifically, the inner product operation correction submodule, based on the trend vector coordinate set and the direction-coupled weight sequence, extracts the first dimension component, the second dimension component, and the center coordinate value of the two-dimensional vector for each mileage segment from the trend vector coordinate set. The center coordinate value is then decomposed into an abscissa and a ordinate value. A product operation is performed on the first dimension component and the abscissa value, using the first dimension component as the multiplicand and the abscissa value as the multiplier to obtain the first product result. Similarly, a product operation is performed on the second dimension component and the ordinate value, using the second dimension component as the multiplicand and the ordinate value as the multiplier to obtain the second product result. Finally, the product results are processed... If a summation operation is performed, the first product result is used as the first addend, the second product result is used as the second addend, and the addition operation is performed to obtain the inner product value. The corresponding mileage segment weight coefficient is extracted from the directional coupling weight sequence. The weight coefficient is compared with the standard weight value for equality. If they are equal, it is confirmed as the standard weight type, and the original value of the inner product value is kept and no correction operation is performed. If they are not equal, it is confirmed as the attenuation weight type, and the inner product value is used as the multiplicand, and the attenuation weight coefficient is used as the multiplier. The multiplication operation is performed to obtain the corrected inner product value. The processed inner product values ​​are recorded in the order of mileage segment number, and a corrected inner product value table is established. Assumption: ; The directional coupling weight sequence is: ; coefficients of the directional coupling weight sequence Applicable to all mileage ranges; 1. Extract data from the trend vector coordinate set. For each mileage segment, extract the first dimension component, the second dimension component, and the corresponding center coordinates of the two-dimensional vector from the trend vector coordinate set: ; Second Mileage Segment: Two-Dimensional Vector ; Third mileage segment: ; 2. Disassembly center coordinates The center coordinates of each mileage segment are decomposed into horizontal coordinates. and ordinate : First mileage segment: ; Second mileage segment: ; Third mileage segment: ; 3. Calculate the inner product of each mileage segment. For each mileage segment, perform the following steps: First mileage segment: ; ; ; Second mileage segment: ; ; ; Third mileage segment: ; ; ; 4. Judgment of the equality between the weighting coefficient and the standard weight value The inner product value needs to be compared with the standard weights. Assume the standard weight coefficients are... Then, the attenuation weight coefficients extracted from the directional coupling weight sequence are: ; For all mileage segments, the inner product value and the standard weighting coefficient Since they are not equal, it is determined to be a decay weight type. 5. Correct the inner product value For attenuation weight types, the following formula is used for correction: ; First mileage segment: ; Second mileage segment: ; Third mileage segment: ; 6. Record the processed inner product value. Based on the above calculations, record the corrected inner product value for each mileage segment, and finally generate a table of corrected inner product values: ; Final result Corrected inner product value for the first mileage segment: ; Corrected inner product value for the second mileage segment: ; Corrected inner product value for the third mileage segment: ; The inner product value table is revised as follows: ; The projection sorting generation submodule calls the corrected inner product value table, extracts the corrected inner product values ​​of each mileage segment, performs a descending sort operation on the inner product values ​​according to their numerical values, arranges the mileage segment identifiers according to their position, and generates a trend projection sorting sequence. Specifically, the projection sorting generation submodule calls the corrected inner product value table, extracts the corrected inner product values ​​for each mileage segment from the corrected inner product value table, performs a descending sorting operation on all inner product values, compares each inner product value pairwise to determine the size relationship, places the larger value at the beginning of the sequence and the smaller value at the end of the sequence, and performs the comparison and position adjustment operations one by one to complete the inner product values ​​to descending order from largest to smallest, obtains the sorted value sequence, and calls the mileage segment identifiers according to the position order based on the mileage segment identifiers corresponding to each value in the sorted value sequence, and arranges the mileage segment identifiers corresponding to each position in the sorted value sequence in order to generate the trend projection sorting sequence; Assumption: ; 1. Extract the corrected inner product value Extract the corrected inner product values ​​for each mileage segment from the corrected inner product value table: ; ; Corrected inner product value for the third mileage segment: ; 2. Perform descending sort operation Arrange these modified inner product values ​​in descending order (from largest to smallest): Original inner product value: ; After sorting in descending order, we get: ; 3. Obtain the sorted numerical sequence The sequence of corrected inner product values ​​after arrangement is as follows: ; 4. Obtain the corresponding mileage segment identifier based on the sorted numerical sequence. Arrange the mileage segment identifiers in sequence according to the sorted values; Assuming the mileage segment identifiers are numbered sequentially: The first mileage segment is marked as 1; The second mileage segment is marked as 2; The third mileage segment is marked as 3; Arrange the mileage segment identifiers in rank order based on the inner product values ​​after permutation: ; ; Minimum corrected inner product value Corresponding mileage segment identifier ; Therefore, the sorting sequence of mileage segment identifiers is as follows: ; 5. Generate trend projection sorting sequence Finally, the trend projection sort sequence is: ; Final result ; Please see Figure 3 The early warning path generation module includes a threshold filtering and positioning submodule, a link continuity determination submodule, and a network aggregation and construction submodule. The threshold filtering and positioning submodule extracts the position number and offset level identifier of each mileage segment based on the trend projection sorting sequence and the segment offset level sequence. It compares the offset level identifier with the set offset threshold, filters the mileage segments whose offset level identifier values ​​are greater than or equal to the offset threshold, arranges the filtered mileage segment identifiers in order of position, and generates a threshold-compliant segment position table. Specifically, the threshold screening and positioning submodule extracts the position number and offset level identifier of each mileage segment from the trend projection sorting sequence and the segment offset level sequence. For each mileage segment, the offset level identifier is used as a comparison parameter, and an offset threshold value is set. This offset threshold is set according to the distribution range of offset level identifiers of abnormal fluctuation mileage segments in historical monitoring data. The minimum and median values ​​of the offset level identifiers of historical abnormal mileage segments are statistically analyzed, and the value between the two is selected as the offset threshold. For each mileage segment, the offset level identifier and the offset threshold are compared. If the offset level identifier value is greater than or equal to the offset threshold, the mileage segment is marked as a qualified mileage segment. If the offset level identifier value is less than the offset threshold, it is marked as a non-qualified mileage segment. All mileage segment identifiers marked as qualified are filtered out, and the position number corresponding to the qualified mileage segment is extracted. The filtered mileage segment identifiers and position numbers are arranged in ascending order of position number to generate a threshold qualified segment position table. Assumption: The trend projection sort sequence is: ; The paragraph offset level sequence is as follows: ; Assuming offset level identifier ; 1. Extract data from the trend projection sort sequence and the paragraph offset level sequence. Extract the rank number and offset level identifier for each mileage segment from these two sequences. ; Offset level identifier in paragraph offset level sequence: First mileage segment offset level indicator: Normal level (0); Second mileage segment offset level identifier: offset level (1); Third mileage segment offset level identifier: Offset level (1); 2. Set the offset threshold By using historical monitoring data, the distribution range of offset level indicators for abnormal mileage segments was obtained, and the minimum and median values ​​of offset level indicators for historical abnormal mileage segments were calculated. ; ; Therefore, the offset threshold is set as follows: ; 3. Comparison of Offset Level Indicator and Offset Threshold For each mileage segment, determine whether its offset level indicator is greater than or equal to the offset threshold (0.5): First mileage segment: Offset level indicator is normal (0), 0 < 0.5, marked as non-compliant mileage segment; Second mileage segment: Offset level indicator (1). Marked as the compliant mileage section; Third mileage segment: Offset level indicator (1). Marked as the compliant mileage section; 4. Select compliant mileage segments. Filter out the mileage segments marked as meeting the target mileage: Second mileage segment marker: 2; Third mileage segment marker: 3; 5. Arrange in ascending order of rank. Arrange the selected mileage segment identifiers in ascending order of their rank: ; 6. Generate a table of paragraph rankings that meet the threshold. Generate a table of threshold-compliant segments, where the selected mileage segment identifiers and position numbers are arranged in order of position number: ; Final result The table of paragraph rankings that meet the threshold is as follows: ; The link continuity determination submodule calls the threshold-compliant segment ranking table, extracts the ranking number of each mileage segment, selects adjacent mileage segment ranking numbers and performs a subtraction operation to obtain the ranking difference, and compares it with the continuous distance threshold. When the ranking difference is less than the continuous distance threshold, the adjacent mileage segments are included in the same link and assigned the same link number. When the ranking difference is greater than or equal to the continuous distance threshold, the current link is cut off and the previous sub-link number, segment number and offset level distribution are sealed. The current mileage segment is taken as the starting point of the new link and assigned a new link number. The existence of subsequent continuous mileage segments of the current mileage segment is determined. If they do not exist, they are marked as single-point warning nodes and assigned a subordinate identifier. A hierarchical link node set is established. Specifically, the link continuity determination submodule calls the threshold-compliant segment ranking table, extracts the ranking number of each compliant mileage segment from the table, selects two adjacent mileage segments in the sequence as comparison objects, uses the ranking number of the later mileage segment as the minuend and the ranking number of the earlier mileage segment as the subtrahend, performs a subtraction operation to obtain the ranking difference, sets a continuous distance threshold, which is set according to the actual physical distribution interval of the pipeline mileage segments, calculates the average of the actual interval distance between adjacent mileage segments and converts it into the ranking number difference as the continuous distance threshold, compares the ranking difference with the continuous distance threshold, and determines that the two mileage segments are continuous when the ranking difference is less than the continuous distance threshold, includes the two mileage segments in the same link and assigns the same link number, and records... Record the mileage segment identifiers included in the link. If the difference in position is greater than or equal to the continuous distance threshold, it is confirmed that there is no continuity. Cut off the current link, save the previous link number, count the number of mileage segments included in the previous link and the distribution of the offset level identifiers of each mileage segment. Take the next mileage segment as the starting point of the new link and assign a new link number. Perform position difference calculation and continuity judgment on adjacent mileage segments in the sequence one by one to complete the link affiliation operation of all mileage segments. For the current mileage segment, determine whether there are any continuous mileage segments after it. Call the position number of the subsequent mileage segment to perform position difference calculation and compare it with the continuous distance threshold. If it is determined that there are no continuous mileage segments after it, mark the mileage segment as a single point warning node and assign a subordinate identifier to it. Establish a hierarchical link node set. Assumption: ; This indicates two qualifying mileage segments, where the mileage segment identifier is the same as the rank number; Continuous distance threshold setting: The continuous distance threshold is set based on the actual physical distribution intervals of the pipeline mileage sections. The average value of the actual interval distance is 1. Therefore... ; 1. Extract the ranking number of the qualified mileage segment. Extract the rank number of each qualifying mileage segment from the rank table of qualifying segments: The position number of the first mileage segment: 2; The position number of the second mileage segment: 3; 2. Calculate the difference in rank. For two adjacent mileage segments, calculate their rank difference: First pair of mileage segments (position numbers 2 and 3): ; 3. Comparison with continuous distance threshold The position difference value of 1 is compared with the continuous distance threshold value of 1: ; Since the difference in rank is less than or equal to the continuous distance threshold, the two mileage segments are confirmed to be continuous, included in the same link, and assigned the same link number. 4. Assign link numbers Based on the assessment, these two mileage segments are continuous, so they are included in the same link and assigned link number 1. The identifiers of the mileage segments contained in this link are recorded: Link 1: Includes mileage segment identifiers 2 and 3; 5. Check subsequent continuity Check if there are consecutive mileage segments following: The last mile segment in the current link is identified as 3; Based on the given threshold and conditions, check if there are any consecutive mileage segments. In this example, since there are only two mileage segments, there are no consecutive mileage segments. Mark the last mileage segment as a single-point warning node. 6. Generate a hierarchical set of link nodes Based on the above steps, a link was identified and a single-point early warning node was marked. The final hierarchical link node set is as follows: Link 1: Includes mileage segment identifiers 2 and 3 Single-point early warning node: Mileage segment identifier 3 Final result The hierarchical link node set is as follows: ; The network aggregation and construction submodule calls the hierarchical link node set, extracts each link number, link mileage segment identifier and single-point early warning node identifier, arranges the link mileage segment identifier by link number, arranges the single-point early warning node identifier by subordinate identifier, and combines all links and single-point nodes to generate a multi-level early warning path network. Specifically, the network aggregation and construction submodule calls the hierarchical link node set, extracts each link number and all mileage segment identifiers contained in the link from the set one by one, arranges the mileage segment identifiers in order of position for each link, completes the arrangement of all link mileage segment identifiers, extracts all single-point early warning node identifiers from the set, arranges the single-point early warning node identifiers in order of subordinate identifiers, combines the results of all link arrangement and the results of single-point node arrangement, takes the results of each link arrangement in order of link number as the network level, takes the results of single-point node arrangement as the network independent node level, combines all arrangement results in order of level to generate a multi-level early warning path network; Assumption: ; Mileage segment identifiers included in Link 1: ; Single-point early warning node: ; 1. Extract link number and mileage segment identifier Extract the link number and mileage segment identifier for each link from the hierarchical link node set: Link 1 includes mileage segment identifiers ; 2. Arrange the mileage segment identifiers in the link in chronological order. For each link, the mileage segment identifiers are arranged in rank order: Mileage segment identifiers in Link 1 are arranged in sequence: ; 3. Extract single-point early warning node identifiers Extract the identifier of a single early warning node from the hierarchical link node set: Single-point early warning node identifier is ; 4. Arrange single-point early warning nodes according to their subordinate identifier. Because there is only one single-point early warning node identifier The arrangement result is as follows: Single-point early warning node arrangement results: ; 5. Results of combined link arrangement and results of single node arrangement The link arrangement results and the single-point early warning node arrangement results are combined into a multi-level early warning path network: The permutation result of link 1 is ; The arrangement result of single-point early warning nodes is as follows ; Combine these two in hierarchical order: Network layers: ; Independent node hierarchy: ; 6. Generate a multi-level early warning path network The final multi-level early warning path network is as follows: ; Final result The multi-level early warning path network is as follows: ; By extracting and arranging the mileage segment identifiers of the links, and combining the single-point early warning node with the link information, a multi-level early warning path network is constructed. This network can be used for path analysis and early warning positioning, thereby achieving accurate positioning of risk areas and visualization of propagation paths, and supporting the rapid deployment of reinforcement measures.

[0022] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A real-time intelligent rock deformation prediction and early warning system for tunnels with large deformation, characterized in that: The system includes: The spatial segment acquisition module acquires the three-dimensional coordinates, mileage segment number, and real-time temperature stress value of the monitoring points in the surrounding rock of the tunnel. It determines the affiliation of the monitoring points based on the coordinates and segment number, calculates the average temperature stress value for each segment, and constructs the thermal basic sequence of the segment. The directional feature construction module extracts the mean temperature stress and center coordinates based on the segment thermodynamic basic sequence, calculates the inter-segment distance and mean gradient, normalizes the temperature stress direction values ​​respectively, performs vector angle calculation to obtain the direction difference, and compares it with the consistency threshold to generate a directional coupling weight sequence containing weight coefficients. The paragraph offset evaluation module extracts the weight coefficients of each paragraph in the directional coupling weight sequence, calculates the time variation value by combining the average temperature and average stress values ​​of the current and previous moments, performs weighted correction on the time variation value by the weight coefficients, compares it with the offset threshold, determines the offset level, generates a paragraph offset level sequence, and transmits the weighted corrected time variation value. The propagation sorting deduction module calls the weighted and corrected time-varying difference to synthesize a trend vector, performs an inner product operation with the coordinates of the paragraph center, sorts according to the size of the inner product result, and obtains the trend projection sorting sequence; The early warning path generation module filters paragraphs that reach a set threshold based on the trend projection sorting sequence and the paragraph offset level sequence, extracts the position of the filtered paragraph in the sorting, calculates the position difference between adjacent paragraphs and compares it with the continuous distance threshold. When the difference is less than the threshold, adjacent paragraphs are included in the same link. When the difference is greater than or equal to the threshold, the current link is cut off and the current paragraph is used as the starting point of the new link. All sub-links and single-point nodes are summarized to generate a multi-level early warning path network.

2. The intelligent real-time deformation prediction and early warning system for surrounding rock of large deformation tunnels according to claim 1, characterized in that: The segment thermal foundation sequence includes segment number, average temperature, and average stress; the directional coupling weight sequence includes segment number, direction value, and weight coefficient; the segment offset level sequence includes segment number and offset level; the trend projection sorting sequence includes segment number and sorting position; and the multi-level early warning path network includes link number, sub-link, and single-point early warning node.

3. The intelligent real-time deformation prediction and early warning system for surrounding rock of large deformation tunnels according to claim 1, characterized in that: The comparison result between the directional difference and the consistency threshold is as follows: if the threshold is met, a standard weight of 1.0 is assigned to generate a directional coupling weight sequence; if the threshold is not met, a decay coefficient of 0.3-0.8 is assigned according to the deviation. When the deviation exceeds twice the threshold, a minimum coefficient of 0.1 is assigned and an anomaly mark is added, and the weight is reduced and included in the directional coupling weight sequence. The process of cutting off the current link includes: sealing the previous sub-link number, segment number and offset level distribution; if there are no subsequent consecutive segments after the current segment, it is marked as a single-point warning node and assigned a subordinate identifier.

4. The intelligent real-time deformation prediction and early warning system for surrounding rock of large deformation tunnels according to claim 1, characterized in that: The calculation steps for the time-varying difference value to undergo weighted correction by weighting coefficients are as follows: (1) The time-varying difference of the standard weighted segment is directly compared with the offset threshold; (2) The time-varying difference needs to be multiplied by the corresponding attenuation coefficient before being compared with the offset threshold.

5. The intelligent real-time deformation prediction and early warning system for surrounding rock of large deformation tunnels according to claim 1, characterized in that: The steps for the inner product operation are as follows: (1) The inner product result of the standard weighted paragraphs directly participates in the ranking; (2) The inner product result needs to be multiplied by the corresponding attenuation coefficient before being sorted.

6. The intelligent real-time deformation prediction and early warning system for surrounding rock of large deformation tunnels according to claim 1, characterized in that: The spatial segmentation acquisition module includes: The monitoring data acquisition submodule acquires the three-dimensional coordinates, mileage segment number, and real-time temperature and stress values ​​of the tunnel surrounding rock monitoring point. It decomposes the three-dimensional coordinates into position values ​​in the X, Y, and Z axes, extracts the digital identification code of the mileage segment number, records the temperature and stress values ​​and the acquisition time, binds the values, and generates a spatial attribute set of the monitoring point. The mileage segment allocation submodule calls the spatial attribute set of the monitoring points, extracts the three-dimensional position values ​​of each monitoring point, obtains the preset mileage segment boundary values, and performs interval judgment between the three-dimensional position values ​​of the monitoring points and the corresponding directional boundary values. When the position values ​​of the three directions are all within the interval of a certain mileage segment boundary value, the monitoring point is assigned to that mileage segment, and the segment monitoring point distribution table is obtained. The thermal sequence construction submodule extracts the temperature stress values ​​of all monitoring points within each mileage segment based on the segment monitoring point distribution table, statistically analyzes the temperature stress values ​​within the same mileage segment, obtains the average temperature stress of the mileage segment, arranges the average temperature stress of each mileage segment in order of mileage segment number, and generates the segment thermal basic sequence.

7. The intelligent real-time deformation prediction and early warning system for surrounding rock of large deformation tunnels according to claim 1, characterized in that: The orientation feature construction module includes: The direction vector extraction submodule extracts the mean temperature stress and center coordinate values ​​of each mileage segment based on the thermal basic sequence of the segment, obtains the coordinate difference value and mean difference value of adjacent mileage segments, converts them to the standard numerical range to obtain the normalized distance value and normalized gradient value, and combines them to form a direction vector to generate a set of temperature stress direction vectors. The vector angle evaluation submodule calls the temperature stress direction vector set, extracts the direction vector values ​​of adjacent mileage segments, uses the vector angle formula to obtain the direction difference, sets a direction consistency threshold, compares the direction difference with the threshold, marks it as consistent when the direction difference is within the threshold range, and marks it as deviating when it exceeds the threshold. Records the direction difference and the marking status to obtain the direction deviation evaluation table. The weight coefficient assignment submodule extracts the direction difference and the marking status according to the direction deviation evaluation table. When the marking is consistent, a standard weight coefficient is assigned. When the marking is deviated and the direction difference does not exceed the threshold multiple limit, an attenuation weight coefficient is assigned within a set interval according to the degree of deviation. When the direction difference exceeds the threshold multiple limit, a minimum weight coefficient is assigned and an abnormality mark is added. The weight coefficients are arranged according to the mileage segment number to generate a direction coupling weight sequence.

8. The intelligent real-time deformation prediction and early warning system for surrounding rock of large deformation tunnels according to claim 1, characterized in that: The paragraph offset evaluation module includes: The time-varying difference extraction submodule calls the directional coupling weight sequence to extract the weight coefficients of each mileage segment, obtains the average temperature and average stress values ​​of the current and previous moments, performs difference calculations on the average temperature and average stress values ​​by subtracting the values ​​of the previous moment from the values ​​of the current moment, arranges the difference results according to the mileage segment number, and generates a segment time-varying difference set. The weighted correction submodule extracts the time-varying difference value and corresponding weight coefficient of each mileage segment based on the segment time-varying difference value set and the direction coupled weight sequence. In the standard weight type, the original value of the time-varying difference value is kept. In the attenuation weight type, the time-varying difference value and the attenuation coefficient are multiplied. The processed value is recorded according to the mileage segment number to establish a weighted correction difference value table. The grade determination submodule calls the weighted correction difference table, extracts the weighted correction time-varying difference value of each mileage segment, sets the offset threshold, compares the time-varying difference value with the offset threshold value, marks it as normal grade if it is less than the threshold, and marks it as offset grade if it is greater than or equal to the threshold. The grade marks are arranged according to the mileage segment number, a segment offset grade sequence is generated, and the weighted correction time-varying difference value is transmitted.

9. The intelligent real-time deformation prediction and early warning system for surrounding rock of large deformation tunnels according to claim 1, characterized in that: The propagation ordering deduction module includes: The trend vector synthesis submodule calls the weighted and corrected time-varying difference value, extracts the temperature time-varying difference value and stress time-varying difference value of each mileage segment, assigns the temperature time-varying difference value as the first dimension component of the vector, assigns the stress time-varying difference value as the second dimension component of the vector, combines the two components according to the mileage segment number to construct a two-dimensional vector, pairs the two-dimensional vector with the segment center coordinates, and generates a trend vector coordinate set. The inner product operation correction submodule extracts the two-dimensional vector components and center coordinate values ​​of each mileage segment based on the trend vector coordinate set and the direction coupled weight sequence. It performs a product operation on the first dimension component and the horizontal coordinate value, and a product operation on the second dimension component and the vertical coordinate value. It performs a sum operation on the two product results to obtain the inner product value. In the standard weight type, the original value of the inner product value is maintained. In the attenuation weight type, the inner product value and the attenuation coefficient are multiplied. The processed values ​​are recorded according to the mileage segment number to establish a corrected inner product value table. The projection sorting generation submodule calls the corrected inner product value table, extracts the corrected inner product values ​​of each mileage segment, performs a descending sorting operation on the inner product values ​​according to their numerical values, arranges the mileage segment identifiers according to their position, and generates a trend projection sorting sequence.

10. The intelligent real-time deformation prediction and early warning system for surrounding rock of large deformation tunnels according to claim 1, characterized in that: The early warning path generation module includes: The threshold filtering and positioning submodule extracts the position number and offset level identifier of each mileage segment based on the trend projection sorting sequence and the segment offset level sequence. It compares the offset level identifier with the set offset threshold, filters the mileage segments whose offset level identifier values ​​are greater than or equal to the offset threshold, arranges the filtered mileage segment identifiers in order of position, and generates a threshold-compliant segment position table. The link continuity determination submodule calls the threshold-compliant segment ranking table, extracts the ranking number of each mileage segment, selects adjacent mileage segment ranking numbers and performs a subtraction operation to obtain the ranking difference, and compares it with the continuous distance threshold. When the ranking difference is less than the continuous distance threshold, the adjacent mileage segments are included in the same link and assigned the same link number. When the ranking difference is greater than or equal to the continuous distance threshold, the current link is cut off and the previous sub-link number, segment number and offset level distribution are sealed. The current mileage segment is taken as the starting point of the new link and assigned a new link number. The existence of subsequent continuous mileage segments of the current mileage segment is determined. If they do not exist, they are marked as single-point warning nodes and assigned a subordinate identifier. A hierarchical link node set is established. The network aggregation and construction submodule calls the hierarchical link node set, extracts the link number, link mileage segment identifier and single-point early warning node identifier, arranges the link mileage segment identifier by link number, arranges the single-point early warning node identifier by subordinate identifier, and combines all links and single-point nodes to generate a multi-level early warning path network.

Images