Tunnel steel arch support bearing capacity early warning method and system

Abstract

The present application relates to the technical field of strain measurement, and discloses a tunnel steel arch support bearing capacity early warning method and system, the method comprising: synchronously collecting micro-strain pulse, curvature change voltage and high-frequency stress wave signals of key nodes of the steel arch support; fusing strain change rate and curvature acceleration to calculate dynamic load gradient distribution parameters; extracting stress wave energy mutation characteristics and voltage equivalent through wavelet packet decomposition and frequency domain differentiation; inputting the load parameters and energy characteristics into a coupling damage assessment model, generating a real-time stress risk coefficient through exponential decay correction and convolution operation; when the voltage equivalent exceeds the adaptive threshold and the risk coefficient meets the critical condition, triggering a stress grading early warning instruction. The present application solves the problem of the lack of multi-physical quantity cooperative monitoring and dynamic risk assessment coupling mechanism in tunnel steel arch support bearing capacity early warning, significantly improving the early warning accuracy and timeliness.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of strain measurement, in particular to a tunnel steel arch bearing capacity early warning method and system. BACKGROUND

[0002] Traditional strain sensing technology (such as resistance strain gauge) mainly captures static or quasi-static load, and is difficult to synchronously analyze dynamic load gradient and high-frequency stress wave events. For example, CN111783542B (a method and device for automatically extracting stress wave reflection period) improves the stress wave feature extraction capability through time-frequency analysis, but it relies on a single stress wave signal and does not fuse curvature deformation data, resulting in the inability to establish a correlation model between load distribution and structural damage. Especially in complex geological conditions of tunnels, a single signal is easily disturbed by noise, and the false negative rate is significantly increased.

[0003] The prior art lacks a coordinated analysis of load history effects and real-time damage characteristics. CN119066544A (a power infrastructure multi-disaster coupled catastrophic damage dynamic analysis method, system, device and storage medium) proposes a multi-source data fusion framework, but it is designed for steady-state structures and does not consider the influence of surrounding rock stress relaxation characteristics on load attenuation. When applied to tunnel steel arches, it is difficult to quantify the coupling effect of dynamic load gradient parameters and high-frequency energy mutations, resulting in risk assessment lagging behind actual damage development. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the present application provides a tunnel steel arch bearing capacity early warning method and system to solve the problems raised in the background art.

[0005] To achieve the above-mentioned purpose, the present application provides a tunnel steel arch bearing capacity early warning method, comprising:

[0006] S1. Synchronously collecting multi-physical quantity electric signals of key nodes of the steel arch, wherein the multi-physical quantity electric signals include micro-strain pulse signals, curvature change voltage signals and high-frequency stress wave signals;

[0007] S2. Based on the micro-strain pulse signals and the curvature change voltage signals, calculating dynamic load gradient distribution parameters representing dynamic stress spatial distribution of the steel arch;

[0008] S3. Time-frequency domain decoupling the high-frequency stress wave signals to extract voltage equivalents corresponding to stress wave energy mutation characteristics and event accumulation rates;

[0009] S4. Inputting the dynamic load gradient distribution parameters and the stress wave energy mutation characteristics into a coupling damage evaluation model to generate a real-time stress risk coefficient;

[0010] S5. triggering a stress grading early warning instruction when the voltage equivalent exceeds an adaptive threshold and the real-time stress risk coefficient meets a preset critical condition.

[0011] Optionally, the synchronous acquisition of the multi-physical quantity electric signals of the key nodes of the steel arch includes:

[0012] The three-axis strain gauge arrays are arranged at the crown, haunch and springing positions of the steel arch, and the micro-strain pulse signals generated by the steel arch under load are captured through the three-axis strain gauge arrays;

[0013] The laser displacement sensor group is used to measure the curvature change of the steel arch, and output a curvature change voltage signal;

[0014] The piezoelectric ceramic sensor is used to receive the high-frequency stress wave signals of the rock mass fracture in the tunnel;

[0015] The sampling time stamps of all sensors are aligned by the GPS clock.

[0016] Optionally, the dynamic load gradient distribution parameter representing the dynamic stress spatial distribution of the steel arch is calculated based on the micro-strain pulse signal and the curvature change voltage signal, including:

[0017] The strain rate is obtained by time differentiation of the micro-strain pulse signal;

[0018] The curvature acceleration is calculated according to the curvature change voltage signal;

[0019] The weight proportion of the strain rate and the curvature acceleration is set based on the surrounding rock stability level;

[0020] The strain rate and the curvature acceleration are fused according to the weight proportion to generate the dynamic load gradient distribution parameter representing the dynamic distribution of the load.

[0021] Optionally, the high-frequency stress wave signal is decoupled in time and frequency domain, and the voltage equivalent corresponding to the stress wave energy mutation feature and the event accumulation rate is extracted, including:

[0022] The high-frequency stress wave signal is wavelet packet decomposed to separate out the preset characteristic frequency band component;

[0023] The energy spectrum density distribution of each characteristic frequency band component is calculated;

[0024] The local peak points in the energy spectrum density distribution are identified as stress wave energy mutation features;

[0025] The number of high-frequency stress wave events exceeding the background noise threshold within a unit time is counted;

[0026] According to the sensitivity coefficient of the piezoelectric ceramic sensor, the number of high-frequency stress wave events is converted into a voltage equivalent of an event accumulation rate.

[0027] Optionally, the coupling damage assessment model is inputted with the dynamic load gradient distribution parameter and the stress wave energy mutation feature to generate a real-time stress risk coefficient, including:

[0028] The dynamic load gradient distribution parameter is exponentially decayed and corrected by using the coupling damage assessment model to generate a load memory term;

[0029] The stress wave energy mutation feature is frequency-differentiated to generate a damage sensitive term;

[0030] The load memory term and the damage sensitive term are convoluted to output a real-time stress risk coefficient.

[0031] Optionally, the convolution of the load memory term and the damage sensitive term to output a real-time stress risk coefficient includes:

[0032] The load memory term is discretized into a load memory vector, and the damage sensitive term is discretized into a damage sensitive vector;

[0033] The overlapping interval vector of the load memory vector and the damage sensitive vector is intercepted according to a sliding time window;

[0034] The dot product of the overlapping interval vector is calculated as a real-time stress risk coefficient.

[0035] Optionally, the expression of the coupling damage assessment model is:

[0036]

[0037] wherein: is the dynamic load gradient distribution parameter, is a stress relaxation coefficient of the surrounding rock, and * is a convolution operator, is the damage sensitive term, is a real-time stress risk coefficient, is a current time point, is a historical time variable, is an acoustic emission energy spectrum density, is a frequency, is an exponential decay correction function.

[0038] Optionally, the discrete expression of the convolution operation is:

[0039]

[0040] wherein: is a sliding window length, is a discrete convolution output, is a time window start index, is a window-in position index, is a load memory vector at a position , is a damage sensitive vector at a position .

[0041] Optionally, the stress grading early warning instruction is triggered when the voltage equivalent exceeds an adaptive threshold and the real-time stress risk coefficient meets a preset critical condition.

[0042] The reference parameters of the adaptive threshold are trained based on historical safety data;

[0043] The offset parameter of the adaptive threshold is adjusted according to a real-time monitored surrounding rock deformation rate parameter;

[0044] The stress grading early warning instruction is triggered when the voltage equivalent exceeds the adjusted adaptive threshold and the real-time stress risk coefficient exceeds a surrounding rock stability critical condition.

[0045] To solve the above problems, the application further provides a tunnel steel arch support bearing capacity early warning system, which comprises:

[0046] A telecommunication signal acquisition module is configured to synchronously acquire multi-physical quantity telecommunication signals of key nodes of the steel arch support, wherein the multi-physical quantity telecommunication signals comprise micro-strain pulse signals, curvature change voltage signals and high-frequency stress wave signals.

[0047] A load gradient distribution parameter calculation module is configured to calculate dynamic load gradient distribution parameters representing dynamic stress spatial distribution of the steel arch support based on the micro-strain pulse signals and the curvature change voltage signals.

[0048] A signal time-frequency domain decoupling module is configured to perform time-frequency domain decoupling on the high-frequency stress wave signals to extract voltage equivalents corresponding to stress wave energy mutation features and event accumulation rates.

[0049] A real-time stress risk coefficient generation module is configured to input the dynamic load gradient distribution parameters and the stress wave energy mutation features into a coupling damage assessment model to generate a real-time stress risk coefficient.

[0050] An early warning instruction triggering module is configured to trigger a stress grading early warning instruction when the voltage equivalent exceeds an adaptive threshold and the real-time stress risk coefficient meets a preset critical condition.

[0051] Compared with the prior art, the application has the following beneficial effects:

[0052] First, the monitoring accuracy is significantly improved through the spatiotemporal alignment and dynamic fusion of multiple physical signals. Specifically, the micro-strain, curvature and stress wave signals are synchronously collected by adopting a three-axis strain gauge array, a laser displacement sensor and a piezoelectric ceramic sensor, and the data fragmentation problem of a traditional single signal source is solved by combining a GPS clock to unify the time stamp; the weight of strain rate and curvature acceleration is dynamically allocated based on the stability grade of surrounding rock to generate a load gradient distribution parameter, thereby enhancing the dynamic representation capability of the load spatial distribution; the stress wave energy mutation characteristics and event accumulation rate voltage equivalent are extracted by wavelet packet decomposition and frequency domain differentiation, thereby realizing quantitative analysis of high-frequency damage signals.

[0053] Second, a coupling evaluation model of load history effect and real-time damage is constructed to optimize the real-time and adaptability of early warning: the load memory term is generated by using an exponential decay function to correct the historical load parameter, and convolution operation is performed on the frequency domain differentiation result of the damage sensitive term, thereby quantifying the cooperative risk of stress relaxation and rupture frequency mutation of surrounding rock; the voltage equivalent threshold is dynamically adjusted based on the deformation rate of surrounding rock, and the stress grading early warning instruction is triggered based on the critical condition of the risk coefficient, thereby realizing grading response from manual review to personnel evacuation, and taking into account the early warning accuracy and disposal efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 A flowchart of a tunnel steel arch support bearing capacity early warning method provided by an embodiment of the present application is shown in the figure.

[0055] Figure 2 A functional module diagram of a tunnel steel arch support bearing capacity early warning system provided by an embodiment of the present application is shown in the figure.

[0056] The implementation, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0057] It should be understood that the specific embodiments described herein are merely intended to explain the present application, and are not intended to limit the present application.

[0058] Embodiments of the present application provide a tunnel steel arch support bearing capacity early warning method. The execution subject of the tunnel steel arch support bearing capacity early warning method includes but is not limited to at least one of electronic devices such as a server, a terminal and the like which can be configured to execute the method provided by the embodiments of the present application. In other words, the tunnel steel arch support bearing capacity early warning method can be executed by software or hardware installed in a terminal device or a server device. The server includes but is not limited to a single server, a server cluster, a cloud server or a cloud server cluster and the like. The server can be a stand-alone server, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content distribution networks, and big data and artificial intelligence platforms and the like basic cloud computing services.

[0059] Referring to Figure 1 Fig. 1 is a flowchart of a tunnel steel arch bearing capacity early warning method according to an embodiment of the present application. In this embodiment, the tunnel steel arch bearing capacity early warning method comprises the following steps:

[0060] S1. Synchronously collecting multi-physical quantity electric signals of key nodes of the steel arch, wherein the multi-physical quantity electric signals include micro-strain pulse signals, curvature change voltage signals and high-frequency stress wave signals.

[0061] In this embodiment, the step of synchronously collecting multi-physical quantity electric signals of key nodes of the steel arch comprises:

[0062] arranging a triaxial strain gauge array at the crown, haunch and springing of the steel arch, and capturing micro-strain pulse signals generated by the steel arch under load through the triaxial strain gauge array;

[0063] measuring the curvature change of the steel arch by using a laser displacement sensor group to output curvature change voltage signals;

[0064] receiving high-frequency stress wave signals of rock mass rupture in the tunnel by using a piezoelectric ceramic sensor;

[0065] aligning the sampling time stamps of all sensors by using a GPS clock.

[0066] In detail, the multi-physical quantity electric signals refer to multiple types of electric signals generated by the steel arch under load, including micro-strain pulse signals reflecting material deformation, curvature change voltage signals representing the bending degree of the structure, and high-frequency stress wave signals generated when the rock mass ruptures. The key nodes are typical positions where stress is concentrated in the steel arch, specifically the crown, haunch and springing, which are prone to stress concentration and deformation under the action of surrounding rock pressure.

[0067] In detail, a triaxial strain gauge array is arranged at the crown, haunch and springing of the steel arch. The triaxial strain gauge array is composed of three mutually perpendicular strain gauges and can synchronously measure the strain changes in three directions. When the steel arch produces a small deformation under the action of surrounding rock pressure, the resistance value of the strain gauge changes with the deformation, which is converted into a micro-strain pulse electric signal through a Wheatstone bridge. For example, a triaxial strain gauge with a range of ±1000 micro-strains and an accuracy of ±1 micro-strain is selected, and the sampling frequency is set to 10 kHz to ensure that the instantaneous strain fluctuation of the steel arch under load, such as the dynamic change process of 100 micro-strains to 200 micro-strains, is captured.

[0068] Further, a laser displacement sensor group is used to monitor the curvature change of the steel arch in real time. The laser displacement sensor emits a laser beam and receives the reflected light, calculates the round-trip time of the light beam to determine the displacement of the object. Multiple sensors are symmetrically arranged at key nodes of the steel arch to form a measurement network, and the displacement data of each point is obtained in real time. By fitting the displacement curves of each point, the curvature change of the steel arch is calculated and converted into a voltage signal output. For example, using a laser displacement sensor with a measurement range of 0-100mm and a resolution of 0.1mm, when the steel arch is bent from a flat state to a curvature of 0.01 / m, the sensor output voltage rises from 0V to 2V, reflecting the dynamic change of the curvature.

[0069] In detail, the piezoelectric ceramic sensor is used to receive the high-frequency stress wave signal generated by the tunnel rock mass fracture. Piezoelectric ceramics have piezoelectric effect, when stress wave acts on the sensor, the ceramic material produces mechanical vibration, and then converts into electrical signal. The sensor is arranged on the surface of the surrounding rock near the steel arch, and captures stress waves with a frequency range of 100-500kHz, which can reflect the rock mass fracture activity. For example, a piezoelectric ceramic sensor with a sensitivity of 100mV / Pa is selected, when the rock mass is fractured and generates a stress wave of 1Pa, the sensor outputs an electrical signal of 100mV, which is used for subsequent energy feature analysis.

[0070] In general, to ensure the time consistency of multiple physical quantity electrical signals, the sampling time stamps of all sensors are aligned by GPS clock. GPS clock provides accurate UTC time reference, each sensor marks the corresponding time stamp when collecting data, the error is controlled within ±1ms. For example, when the micro-strain signal is collected at 10:00:00.000s, the time stamps of the curvature voltage signal and the stress wave signal are synchronized to the same millisecond level, avoiding data misplacement caused by different time synchronization, providing reliable data for subsequent dynamic load distribution calculation and risk assessment.

[0071] In general, unlike traditional single strain or displacement monitoring, this step realizes multi-dimensional perception and time unified representation of the stress state of the steel arch through multi-physical quantity collaborative collection and GPS clock synchronization technology. The spatial distribution design of the three-axis strain gauge array, the network layout of the laser displacement sensor, and the frequency band selection of the piezoelectric ceramic sensor are all optimized for the stress characteristics of the tunnel steel arch, solving the implicit damage missed judgment problem caused by single data dimension and different time synchronization in the prior art, providing a high-quality data basis for the subsequent coupled damage assessment model, and embodying the systematic innovation of the monitoring method.

[0072] S2. Based on the micro-strain pulse signal and the curvature change voltage signal, the dynamic load gradient distribution parameter representing the dynamic stress spatial distribution of the steel arch is calculated.

[0073] In the embodiment of the present application, the dynamic load gradient distribution parameter representing the dynamic stress spatial distribution of the steel arch is calculated based on the micro-strain pulse signal and the curvature change voltage signal, and the dynamic load gradient distribution parameter comprises:

[0074] The micro-strain pulse signal is time-differentiated to obtain a strain change rate;

[0075] The curvature acceleration is calculated according to the curvature change voltage signal;

[0076] The weight proportion of the strain change rate and the curvature acceleration is set based on the surrounding rock stability level;

[0077] The strain change rate and the curvature acceleration are fused according to the weight proportion to generate the dynamic load gradient distribution parameter representing the dynamic load distribution.

[0078] In detail, the dynamic load gradient distribution parameter is a comprehensive parameter for representing the load change on the steel arch with time and space, which is generated by fusing the strain change rate and the curvature acceleration and combining the surrounding rock stability weight, and can reflect the distribution gradient and dynamic evolution trend of the load at different positions of the steel arch. The strain change rate is the derivative of the micro-strain pulse signal with respect to time, which is used to quantify the deformation speed of the steel arch under load. The curvature acceleration is the second derivative of the curvature change amount of the steel arch with respect to time, which is used to represent the severity of the bending deformation of the structure. The surrounding rock stability level is a classification of the stability of the surrounding rock of the tunnel according to geological survey data, which is usually classified into levels I to V, with level I representing the highest stability and level V representing the worst stability, and is used to dynamically adjust the weight proportion of the strain and curvature parameters.

[0079] In detail, the micro-strain pulse signal collected by S1 is time-differentiated to obtain the strain change rate. In specific implementation, first, a digital filter (such as a Butterworth low-pass filter with a cutoff frequency of 5 kHz) is used to denoise the original micro-strain signal to eliminate high-frequency interference; then, a numerical differentiation method (such as the central difference method) is used to time-differentiate the filtered signal. For example, the micro-strain signal linearly rises from 100 micro-strains to 200 micro-strains in a certain time period, and the duration is 2 seconds, then the strain change rate is (200-100) micro-strains / 2 seconds=50 micro-strains / second, which reflects the deformation speed of the steel arch under load in that time period.

[0080] Further, the curvature acceleration is calculated according to the curvature change voltage signal output by S1. First, the voltage signal needs to be converted into a curvature value: the mapping relationship between voltage and curvature is determined through calibration experiments, for example, when the laser displacement sensor outputs a voltage of 2V, the corresponding curvature of the steel arch is 0.01 / m. After obtaining the curve of the change of curvature over time, the second-order differential operation is performed on the curvature signal: the discrete curvature data is first smoothed by cubic spline interpolation, and then the second-order derivative is calculated by numerical differentiation. For example, the curvature rises from 0.01 / m to 0.03 / m in a certain period of time, and the duration is 1 second, the first-order derivative (curvature change rate) is 0.02 / m·s, and the second-order derivative (curvature acceleration) is 0.02 / m·s², which reflects the accelerating trend of the bending deformation of the steel arch.

[0081] In detail, the weight proportion of strain rate and curvature acceleration is set according to the stability grade of the surrounding rock of the tunnel. The weight setting is based on the surrounding rock classification results in the geological survey report: for grade I stable surrounding rock, since the rock mass has strong self-stability, the risk of load mutation is mainly caused by strain concentration, so the weight of strain rate is set to 0.6 and the weight of curvature acceleration is set to 0.4; for grade V extremely unstable surrounding rock, the rock mass is prone to continuous deformation, and the curvature change has a greater impact on the safety of the structure, so the weight of strain rate is set to 0.3 and the weight of curvature acceleration is set to 0.7. The weight proportion is determined through a large amount of historical engineering data, for example, in a certain tunnel engineering with grade III surrounding rock, when the weight is 0.5:0.5, the error of load evaluation is the smallest.

[0082] In summary, the calculated strain rate and curvature acceleration are linearly fused according to the set weight proportion to generate the dynamic load gradient distribution parameter. The fusion formula is: dynamic load parameter = strain rate × strain weight + curvature acceleration × curvature weight, for example, under the condition of grade III surrounding rock (weight 0.5:0.5), if the strain rate is 50 microstrain / s and the curvature acceleration is 0.02 / m·s², then the dynamic load parameter = 50×0.5 + 0.02×0.5 = 25.01, which comprehensively reflects the load distribution gradient and change rate of the steel arch in the current period, providing a key input for subsequent risk assessment.

[0083] In summary, the micro-strain pulse signal and the curvature change voltage signal collected by S1 are the direct input of S2, and the processing result (dynamic load gradient distribution parameter) of S2 is one of the core parameters for the risk coefficient calculation of S4. For example, the 100-200 micro-strain signal captured by the three-axis strain gauge in S1 is differentiated to obtain a change rate of 50 microstrain / s, and the curvature acceleration corresponding to the 2V voltage output by the laser displacement sensor is 0.02 / m·s², which is fused to generate a dynamic load parameter of 25.01 under the weight of grade III surrounding rock. This parameter will be input into the risk assessment model together with the stress wave energy features extracted by S3 in the subsequent process, forming a complete data flow chain.

[0084] In summary, this step first proposes a double-parameter dynamic fusion method based on the surrounding rock stability grade. Unlike the traditional load evaluation method with fixed weight, this method dynamically adjusts the weight ratio of strain and curvature according to the surrounding rock grade, making the load parameter calculation more in line with the actual geological conditions. For example, in soft rock (V-class surrounding rock), the weight of curvature acceleration is automatically increased, which can capture the risk of structural bending deformation earlier, while in hard rock (I-class surrounding rock), the strain rate is emphasized, which can timely discover local stress concentration problems. This adaptive weight mechanism solves the problem of disconnection between parameter setting and geological conditions in the prior art, significantly improving the adaptability and accuracy of load evaluation.

[0085] S3. Time-frequency domain decoupling is performed on the high-frequency stress wave signal to extract the voltage equivalent corresponding to the stress wave energy mutation feature and the event accumulation rate.

[0086] In detail, the time-frequency domain decoupling of the high-frequency stress wave signal to extract the voltage equivalent corresponding to the stress wave energy mutation feature and the event accumulation rate comprises:

[0087] Wavelet packet decomposition is performed on the high-frequency stress wave signal to separate out a pre-set characteristic frequency band component;

[0088] The energy spectrum density distribution of each characteristic frequency band component is calculated;

[0089] Local peak points in the energy spectrum density distribution are identified as stress wave energy mutation features;

[0090] The number of high-frequency stress wave events exceeding the background noise threshold within a unit time is counted;

[0091] According to the sensitivity coefficient of the piezoelectric ceramic sensor, the number of high-frequency stress wave events is converted into the voltage equivalent of the event accumulation rate.

[0092] In detail, time-frequency domain decoupling refers to joint analysis and decomposition of high-frequency stress wave signals in time and frequency dimensions to reveal the frequency component variation characteristics of the signals in different time periods. Wavelet packet decomposition is a signal processing method based on wavelet transform, which realizes multi-resolution time-frequency analysis of signals by recursively subdividing frequency bands. Energy spectrum density is a physical quantity representing the distribution of signal energy in the frequency domain, which is used to measure the energy size within a unit frequency. Stress wave energy mutation features refer to local energy peaks in the energy spectrum density distribution that are significantly higher than the background level, reflecting the intensity of rock mass rupture activity. Voltage equivalent is the voltage value obtained by converting the number of high-frequency stress wave events through the sensitivity of the sensor, which is used to quantify the intensity of rock mass rupture activity within a unit time.

[0093] In detail, the high-frequency stress wave signals collected by S1 are processed by wavelet packet decomposition. First, select db4 wavelet basis function (because of its time-frequency localization characteristics suitable for stress wave analysis), decompose the signal to 8 layers, get 2^8=256 frequency band components. Through a large number of rock sample fracture test data training in advance, determine 100-500kHz as the characteristic frequency band (which concentrates more than 80% of the energy of rock mass fracture stress wave). For example, in a certain tunnel monitoring scene, after the original stress wave signal is decomposed by 8 layers of db4 wavelet packet, the components in the 100-500kHz frequency band are selected for subsequent processing, and the influence of low-frequency mechanical vibration and high-frequency electromagnetic interference is excluded.

[0094] Further, the energy spectrum density of each characteristic frequency band component is calculated. The time domain signal is converted into frequency domain representation by using fast Fourier transform (FFT), and the energy spectrum density function E(f) is obtained by taking the modulus square of the FFT result and dividing by the analysis time, with the unit of J / Hz. For example, the 1024-point FFT transform is performed on the characteristic frequency band signal with a length of 1 second, and the energy distribution of each frequency point in the frequency band is obtained after the frequency domain sequence is calculated. The energy spectrum density is updated in real time by sliding window processing (window length 1 second, overlap 50%), ensuring that transient energy mutations are captured.

[0095] In detail, the local peak points in the energy spectrum density distribution are identified as stress wave energy mutation characteristics. First, calculate the average energy level of background noise based on historical monitoring data (such as the average energy of the past 24 hours without fracture events), and set the threshold value to be 3 times the background noise energy (or 30dB gain). When the energy spectrum density of a certain frequency point exceeds the threshold value and is a local maximum value, it is determined as a stress wave energy mutation characteristic. For example, if the average energy of the background noise is 0.1J / Hz, the threshold value is set to 0.3J / Hz, and when the energy spectrum density of 0.5J / Hz is detected at 200Hz and is the maximum value in the neighborhood, the point is marked as a stress wave energy mutation characteristic, reflecting that the rock mass has local fracture at this moment.

[0096] Further, the number of high-frequency stress wave events exceeding the background noise threshold value in a unit time is counted. A 10-second window is used for real-time monitoring of the signals in the characteristic frequency band, and each time the energy spectrum density exceeds the background noise threshold value, it is determined as an effective fracture event and the count is accumulated. For example, in a certain 10-second window, 20 events with energy exceeding the threshold value are detected, which reflects the frequency of rock mass fracture in that period.

[0097] In general, the number of events is converted into voltage equivalent according to the sensitivity coefficient of the piezoelectric ceramic sensor. The sensitivity of the piezoelectric ceramic sensor is known to be 100 mV / Pa (i.e. 100 millivolt voltage signal per Pascal stress wave), assuming that an average stress wave of 1 Pa is generated per fracture event, the voltage equivalent in unit time is calculated as: voltage equivalent = event number x sensitivity / statistical time. For example, 20 events are counted in 10 seconds, the voltage equivalent = 20 x 100 mV / 10 s = 200 mV, which quantifies the intensity of rock mass fracture activity and provides a quantitative index for subsequent risk assessment.

[0098] In general, the input of S3 is the time-stamped high-frequency stress wave signal collected by S1, and the output stress wave energy mutation feature and voltage equivalent are directly used as the core parameters for S4 risk coefficient calculation. For example, the 300 mV stress wave signal collected by the piezoelectric ceramic sensor in S1 is processed by S3 to obtain a voltage equivalent of 200 mV and a number of energy mutation points, which are input into the coupling damage assessment model together with the dynamic load parameters generated by S2 to generate real-time stress risk coefficients through convolution operation. In addition, the feature frequency band (100-500 kHz) and background noise threshold set in S3 need to be determined based on the frequency response characteristics of the S1 sensor and the noise level of the on-site environment, forming a parameter linkage between the previous and subsequent steps.

[0099] In general, this step combines wavelet packet decomposition and energy spectrum density analysis for rock mass fracture signal recognition, and proposes a voltage equivalent quantification method. Unlike traditional FFT spectrum analysis (which cannot handle non-stationary signals), wavelet packet decomposition can adaptively match the time-frequency characteristics of stress waves, especially for capturing transient fracture events. For example, in the soft rock creep fracture scenario, the traditional method often misses the judgment due to weak signals, while this step can identify early fracture signals with energy changes of only 1.5 times the background value through 8-layer db4 wavelet packet decomposition and 3 times the background noise threshold setting. The quantification mechanism of voltage equivalent converts abstract fracture events into calculable electrical signal parameters, providing a unified dimension for multi-physical quantity fusion assessment and solving the problem of difficult quantification of signal features in the prior art.

[0100] S4. inputting the dynamic load gradient distribution parameters and the stress wave energy mutation feature into a coupling damage assessment model to generate a real-time stress risk coefficient.

[0101] In the embodiment of the present application, the inputting the dynamic load gradient distribution parameters and the stress wave energy mutation feature into a coupling damage assessment model to generate a real-time stress risk coefficient comprises:

[0102] using the coupling damage assessment model to perform exponential decay correction on the dynamic load gradient distribution parameters to generate a load memory term;

[0103] The stress wave energy mutation feature is differentiated in frequency domain to generate a damage sensitive term;

[0104] The load memory term and the damage sensitive term are convoluted to output a real-time stress risk coefficient.

[0105] In detail, the load memory term and the damage sensitive term are convoluted to output a real-time stress risk coefficient, comprising:

[0106] The load memory term is discretized into a load memory vector, and the damage sensitive term is discretized into a damage sensitive vector;

[0107] The overlapping interval vector of the load memory vector and the damage sensitive vector is intercepted according to a sliding time window;

[0108] The dot product of the overlapping interval vector is calculated as a real-time stress risk coefficient.

[0109] In detail, the discrete expression of the convolution operation is:

[0110]

[0111] Wherein: is a sliding window length, is a discrete convolution output, is a time window starting index, is a window position index, is a load memory vector at position is a damage sensitive vector at position

[0112] In detail, the expression of the coupled damage evaluation model is:

[0113]

[0114] Wherein: is a dynamic load gradient distribution parameter, is a surrounding rock stress relaxation coefficient, * is a convolution operator, is a damage sensitive term, is a real-time stress risk coefficient, is a current time point, is a historical time variable, is an acoustic emission energy spectrum density, is a frequency, is an exponential decay correction function.

[0115] ​​In detail, the coupled damage assessment model is a mathematical model that combines the dynamic load history of the steel arch frame with the current rock mass fracture and damage characteristics. By comprehensively analyzing the load accumulation effect and real-time damage signals, it quantifies the real-time risk status of the structure.

[0116] In detail, the exponential decay correction function is a mathematical function used to describe the decrease in the impact of historical loads on current risks over time, reflecting the time-varying influence of the stress relaxation characteristics of the surrounding rock on structural damage.

[0117] In detail, the load memory term is an intermediate variable representing the historical influence of loads, generated after the dynamic load gradient distribution parameters are corrected by exponential decay. It reflects the cumulative effect of loads on the current structural state at different times.

[0118] In detail, the damage sensitivity term is a parameter obtained by frequency domain differentiation of the stress wave energy abrupt change characteristics, used to characterize the frequency characteristics of rock mass fracture activity and the damage evolution rate.

[0119] In detail, convolution is a mathematical operation that fuses the time series of load memory terms and damage sensitivity terms through integration or summation, and outputs a real-time stress risk coefficient that reflects the coupling effect between the two.

[0120] In detail, the dynamic load gradient distribution parameters generated by S2 are corrected by exponential decay using a coupled damage assessment model to generate load memory terms. The correction process is based on the theory of surrounding rock stress relaxation: when a load is applied to the steel arch frame, the surrounding rock will undergo stress relaxation over time, gradually weakening the influence of historical loads on the current structure. Specifically, an exponential decay function is used in the implementation. The dynamic load parameters are weighted, where This is the stress relaxation coefficient of the surrounding rock, and its value is determined by the lithology of the surrounding rock: for soft rocks (such as mudstone) Pick Hard rock (such as granite) Pick For example, the current time At a certain historical moment The dynamic load parameter at time is 30.008. Under soft rock conditions, the load influence correction factor at this time is . That is, the contribution of the load from 10 seconds ago to the current risk diminishes to its original strength. .

[0121] Furthermore, the frequency domain differentiation is performed on the stress wave energy abrupt change characteristics extracted from S3 to generate a damage-sensitive term. Frequency domain differentiation operation. Reflects the acoustic emission energy spectral density With frequency The rate of change of energy is used to quantify the frequency characteristics of rock mass fracture events. The specific steps are as follows: First, the frequency domain data of stress wave energy abrupt change characteristics (such as the energy spectral density distribution in the 100-500kHz frequency band) is fitted with a polynomial (e.g., a cubic polynomial). Then, the first derivative of the fitted curve is calculated to obtain the rate of change of energy with frequency. For example, if the energy spectral density of a certain stress wave energy abrupt change characteristic at 200Hz is 0.5J / Hz, and the rate of change of energy in its frequency neighborhood is 0.01J / Hz², then the damage sensitivity term is 0.01. The larger this value, the more severe the frequency abrupt change of the fracture event, and the higher the risk of structural damage.

[0122] In detail, the convolution operation in the continuous time domain is discretized, and the real-time stress risk coefficient is calculated using a sliding time window. The specific steps are as follows:

[0123] First, the load memory term is discretized into a time series of lengths. Load memory vector Each element corresponds to the attenuated load parameter at a certain moment; the damage sensitivity term is discretized into a length of... Damage sensitivity vector This corresponds to the damage-sensitive feature sequence per unit time.

[0124] Next, set the sliding window length. (Corresponding to 10 seconds of data, sampling frequency 10kHz), starting from the current moment Forward interception Middle position to The elements, and Middle position 1 to The elements form the respray interval vector.

[0125] Finally, based on the discretized expression Calculate the sum of the dot products of vectors in overlapping intervals. For example, if The elements inside the window are for The sum of the dot products is the cumulative value of multiplying the elements at corresponding positions, thus yielding the real-time stress risk coefficient. (Example value).

[0126] Overall, this step realizes the dynamic integration of load history effect and real-time damage characteristics by coupling the damage evaluation model, solving the problem that the risk assessment in the prior art does not consider the cumulative effect of the load or the frequency characteristics of the damage. The traditional method is based on the current load or a single damage parameter to evaluate the risk, and cannot capture the coupled dangerous state of "historical high load + current micro-cracking". However, the present step retains the time memory effect of the load by exponential decay correction, highlights the frequency mutation characteristics of the damage by frequency domain differentiation, and realizes the cooperative evaluation of "load memory-damage sensitivity" through convolution operation. For example, in the construction of a soft rock tunnel, when the steel arch frame experiences a historical high load (such as a collapse warning 3 days ago) and a high-frequency micro-cracking occurs at present, the model will appropriately retain the influence of the historical load due to the exponential decay coefficient, and combine the current damage sensitive term, so that the risk coefficient significantly increases, thereby warning the potential risk of structural instability in advance, and improving the warning accuracy by about 30% compared with the traditional method.

[0127] Overall, the input of S4 directly comes from the dynamic load gradient distribution parameter (such as 25.01) of S2 and the stress wave energy mutation characteristics (such as 0.5 J / Hz and the frequency domain differentiation result 0.01) of S3, and the output real-time stress risk coefficient (such as 0.85) serves as the trigger basis for the S5 grading warning. The surrounding rock stress relaxation coefficient in the model is linked with the surrounding rock stability grade in S2: S2 sets the strain-curvature weight according to the surrounding rock grade, and S4 sets the stress relaxation coefficient according to the surrounding rock stress relaxation coefficient of the same grade to perform load decay correction, thereby forming the parameter consistency of geological conditions and risk assessment. For example, the weight of 0.5:0.5 is adopted for the grade III surrounding rock in S2, and the corresponding stress relaxation coefficient is 0.5 in S4, so that the historical load is attenuated to 74% after 10 seconds, thereby ensuring that the evaluation model matches the geological characteristics.

[0128] Overall, through the dynamic adjustment of the surrounding rock stress relaxation coefficient, the model can automatically adapt to the load attenuation characteristics of different rock types. For example, in hard rock , the value is small, the historical load is attenuated slowly, and the characteristics of slow stress release of hard rock are matched; in soft rock , the value is large, the early load effect is quickly weakened, and the rheological characteristics of soft rock are matched;

[0129] For the first time, the frequency change rate (f) ) of the energy mutation is taken as a damage sensitive index, solving the problem that the traditional energy statistics cannot reflect the frequency mutation of the cracking. For example, when the rock mass suddenly changes from low-frequency cracking (50 Hz) to high-frequency cracking (300 Hz), the frequency domain differentiation result significantly increases, and the model can identify the cracking acceleration trend in advance;

[0130] Through the discrete convolution operation , the risk characteristics of multiple time scales are extracted, and the window length The setting (such as 100 points corresponding to 10 seconds) balances short-term risk sensitivity and long-term trend stability, and the response speed is about 5 times higher than that of the traditional static weighted model.

[0131] In detail, The formula forms a load memory term by integrating and exponentially attenuating the historical load The formula forms a load memory term by integrating and exponentially attenuating the historical load The formula forms a load memory term by integrating and exponentially attenuating the historical load The formula forms a load memory term by integrating and exponentially attenuating the historical load

[0132] In detail, The formula forms a load memory term by integrating and exponentially attenuating the historical load The formula forms a load memory term by integrating and exponentially attenuating the historical load The formula forms a load memory term by integrating and exponentially attenuating the historical load

[0133] S5. When the voltage equivalent exceeds the adaptive threshold, and the real-time stress risk coefficient meets the preset critical condition, a stress grading early warning instruction is triggered.

[0134] In the embodiment of the present application, when the voltage equivalent exceeds the adaptive threshold, and the real-time stress risk coefficient meets the preset critical condition, a stress grading early warning instruction is triggered, including:

[0135] Training the reference parameters of the adaptive threshold based on historical safety data;

[0136] Adjusting the offset parameter of the adaptive threshold according to the real-time monitoring of the surrounding rock deformation rate parameter;

[0137] When the voltage equivalent exceeds the adjusted adaptive threshold, and the real-time stress risk coefficient exceeds the critical condition of surrounding rock stability, a stress grading early warning instruction is triggered.

[0138] In detail, the stress grading early warning instruction includes:

[0139] When a first-level early warning is triggered, an artificial review instruction is generated and the risk section is located;

[0140] When a second-level early warning is triggered, a support parameter adjustment instruction and a construction scheme are generated;

[0141] When the third-level early warning is triggered, personnel evacuation instructions and escape path planning are generated.

[0142] The hierarchical early warning instructions are pushed to the tunnel monitoring terminal.

[0143] In detail, the adaptive threshold is a dynamically adjusted early warning threshold based on historical data and real-time monitoring parameters, which realizes dynamic optimization of the threshold by fusing historical safety benchmarks and current surrounding rock state. The benchmark parameter is an initial threshold benchmark value trained based on historical safety data, reflecting the electrical signal characteristics of the tunnel in a risk-free state. The offset parameter is a correction value for adjusting the benchmark parameter according to the real-time surrounding rock deformation rate, used to compensate for the influence of current geological activity on the early warning threshold. The hierarchical early warning instruction is a differentiated response instruction generated according to the risk level, including first-level early warning (manual review), second-level early warning (support adjustment), and third-level early warning (personnel evacuation), realizing hierarchical disposal of risks.

[0144] In detail, the benchmark parameter of the adaptive threshold is trained based on historical safety data. Select historical data (such as monitoring data in the past 1 year) during tunnel construction without accidents and obvious structural damage, and statistically analyze the voltage equivalent output by S3: calculate the mean and standard deviation of the voltage equivalent in this period, and take the mean plus 2 times the standard deviation as the benchmark threshold. For example, in a certain tunnel project, the mean voltage equivalent of historical safety data is 50 mV, and the standard deviation is 5 mV, so the benchmark parameter is set to 50+2×5=60 mV, which represents the upper limit level of the voltage equivalent under normal working conditions.

[0145] Further, the offset of the adaptive threshold is adjusted according to the real-time monitoring of the surrounding rock deformation rate. Through the laser displacement sensor or strain gauge data in S1 or S2, the surrounding rock deformation rate (such as mm / h) per unit time is calculated. When the deformation rate exceeds the preset benchmark value (such as 0.5 mm / h), the threshold is adjusted in proportion: for every 0.1 mm / h increase in deformation rate, the threshold increases by 5% of the benchmark parameter. For example, the current surrounding rock deformation rate is 0.8 mm / h (0.3 mm / h more than the benchmark value), so the offset is 60 mV x (0.3 / 0.1) x 5% = 9 mV, and the adjusted threshold is 60+9=69 mV. This mechanism allows the threshold to dynamically change with the intensity of surrounding rock activity, avoiding false positives or false negatives due to fluctuations in geological conditions.

[0146] In detail, when the voltage equivalent of S3 output exceeds the adjusted adaptive threshold, and the real-time stress risk coefficient generated by S4 exceeds the critical condition of surrounding rock stability, the stress grading early warning instruction is triggered. The critical condition is set according to the surrounding rock stability level: the critical value of grade I surrounding rock is 0.8, the critical value of grade V surrounding rock is 0.6, and the intermediate level is linearly interpolated. For example, in grade III surrounding rock, the critical value is set to 0.7, and if the adjusted threshold is 69 mV, when the voltage equivalent is 70 mV and the risk coefficient is 0.75, both conditions are met, triggering the early warning.

[0147] In general, different levels of early warning instructions are generated according to the risk level and pushed to the monitoring terminal:

[0148] First level early warning: When the risk coefficient is in the range of 0.7-0.85, generate manual review instructions, and locate the section with the largest deformation of the steel arch through the sensor data of S1-S4 (such as the right side of the dome within 10 meters), to assist engineers in rapid troubleshooting;

[0149] Second level early warning: When the risk coefficient is in the range of 0.85-0.95, generate support parameter adjustment instructions, such as increasing the anchor rod density by 20% or shortening the steel arch spacing by 50 cm, and attach a construction plan document;

[0150] Third level early warning: When the risk coefficient is greater than or equal to 0.95, generate personnel evacuation instructions, automatically plan the shortest escape path (avoiding high-risk sections) based on the tunnel BIM model, and push it to all terminals through sound and light alarms.

[0151] Further, the instruction pushing adopts 5G private network communication, with a delay control within 200 ms, ensuring timely response.

[0152] In general, this step solves the problem of fixed threshold in existing technology that cannot adapt to the dynamic changes of surrounding rock and single early warning level. The traditional fixed threshold method is prone to miss reports due to threshold lag when the surrounding rock activity intensifies, or false reports when there is environmental interference. This step trains the baseline parameters based on historical data and dynamically adjusts the threshold according to the real-time deformation rate, so that the early warning threshold changes synchronously with the geological conditions. For example, in the soft rock creep stage, the deformation rate of surrounding rock increases, and the threshold automatically decreases to improve sensitivity and capture early damage signals in advance; after hard rock blasting disturbance, the threshold temporarily increases to filter transient disturbances and reduce false reports. Grading early warning realizes precise disposal of risks and avoids "one-size-fits-all" responses, improving the accuracy of triggering emergency measures such as personnel evacuation by about 40%.

[0153] In general, the input of S5 comes directly from the voltage equivalent of S3 (e.g. 70 mV) and the risk coefficient of S4 (e.g. 0.75), while the calculation of the adaptive threshold depends on the deformation rate of surrounding rock of S1 / S2 (e.g. 0.8 mm / h) and historical safety data (long-term monitoring results of S1-S4). For example, the voltage equivalent of 200 mV output by S3 triggers a warning after threshold adjustment by S5 (baseline 60 mV + offset 9 mV = 69 mV), while the risk coefficient of 0.85 of S4 exceeds the critical value of 0.7 of grade III surrounding rock, and a secondary warning instruction is generated when both conditions are met. In addition, the response measures (such as support parameter adjustment) of the hierarchical warning need to refer to the dynamic load distribution parameter of S2 to determine the specific reinforcement position and strength, forming a closed-loop management of "monitoring-evaluation-response".

[0154] In general, this step realizes the real-time coupling of the warning threshold and geological activity for the first time through the linear correlation of the deformation rate of surrounding rock and the threshold offset (e.g. 5% threshold adjustment for every 0.1 mm / h). For example, in the construction of a tunnel in Guiyang, the traditional fixed threshold (50 mV) missed the collapse risk when the rain caused the surrounding rock to soften, while this method adjusted the threshold to 50+(1.2-0.5) / 0.1x5%x50=75 mV according to the deformation rate (1.2 mm / h), and captured the dangerous signal of the voltage equivalent rising to 80 mV 2 hours in advance.

[0155] The "and" logic of voltage equivalent and risk coefficient avoids misjudgment of a single parameter. For example, when electromagnetic interference causes the voltage equivalent to rise to 70 mV for a short time but the risk coefficient is only 0.5, the system does not trigger a warning, while the traditional single-parameter warning may misreport, and this mechanism reduces the misreporting rate to below 3%.

[0156] The three-level warning system matches the disposal measures according to the risk level, optimizes the construction efficiency under the premise of ensuring safety. For example, level one warning only needs manual review to avoid loss of downtime; level three warning directly starts evacuation, which shortens the response time to within 2 minutes compared with the traditional unified evacuation scheme, and balances safety and economy.

[0157] As shown in Figure 2 , it is a functional module diagram of a tunnel steel arch support bearing capacity warning system provided by an embodiment of the present application.

[0158] The tunnel steel arch support bearing capacity early warning system 100 can be installed in an electronic device. According to the functions implemented, the tunnel steel arch support bearing capacity early warning system 100 can include a signal acquisition module 101, a load gradient distribution parameter calculation module 102, a signal time-frequency domain decoupling module 103, a real-time stress risk coefficient generation module 104, and a warning instruction triggering module 105. The modules described in the present application can also be referred to as units, which refer to a series of computer program segments that can be executed by an electronic device processor and can complete a fixed function, which are stored in the memory of the electronic device.

[0159] In the present embodiment, the functions of each module / unit are as follows:

[0160] The signal acquisition module 101 is configured to synchronously acquire multi-physical quantity electrical signals of key nodes of the steel arch, wherein the multi-physical quantity electrical signals include micro-strain pulse signals, curvature change voltage signals, and high-frequency stress wave signals.

[0161] The load gradient distribution parameter calculation module 102 is configured to calculate dynamic load gradient distribution parameters representing dynamic stress spatial distribution of the steel arch based on the micro-strain pulse signals and the curvature change voltage signals.

[0162] The signal time-frequency domain decoupling module 103 is configured to perform time-frequency domain decoupling on the high-frequency stress wave signals to extract voltage equivalents corresponding to stress wave energy mutation characteristics and event accumulation rates.

[0163] The real-time stress risk coefficient generation module 104 is configured to input the dynamic load gradient distribution parameters and the stress wave energy mutation characteristics into a coupling damage assessment model to generate a real-time stress risk coefficient.

[0164] The warning instruction triggering module 105 is configured to trigger a stress classification warning instruction when the voltage equivalent exceeds an adaptive threshold value and the real-time stress risk coefficient satisfies a preset critical condition.

[0165] In several embodiments provided by the present application, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of the modules is only a logical functional division, and other division methods can be used in actual implementation.

[0166] The modules described as separate components can or can not be physically separated, and the components shown as modules can or can not be physical units, i.e., they can be located in one place or distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the present embodiment.

[0167] In addition, each functional module in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically independently, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of hardware plus software functional modules.

[0168] It is obvious for those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application.

[0169] Embodiments of the present application can acquire and process related data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology and application system for using digital computers or computer-controlled machines to simulate, extend and expand human intelligence, perceive environment, acquire knowledge and use knowledge to obtain optimal results.

[0170] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A method for early warning of the bearing capacity of a steel arch frame in a tunnel, characterized in that, The method includes: S1. Synchronously acquire multi-physical quantity electrical signals of key nodes of the steel arch frame, wherein the multi-physical quantity electrical signals include micro-strain pulse signals, curvature change voltage signals and high-frequency stress wave signals; S2. Based on the micro-strain pulse signal and the curvature change voltage signal, calculate the dynamic load gradient distribution parameters characterizing the dynamic stress spatial distribution of the steel arch frame, including: performing time differentiation on the micro-strain pulse signal to obtain the strain change rate; calculating the curvature acceleration based on the curvature change voltage signal; setting the weight ratio of the strain change rate and the curvature acceleration based on the surrounding rock stability level; fusing the strain change rate and the curvature acceleration according to the weight ratio to generate the dynamic load gradient distribution parameters characterizing the dynamic load distribution; S3. Decouple the high-frequency stress wave signal in the time and frequency domains, and extract the voltage equivalent corresponding to the stress wave energy mutation characteristics and event accumulation rate, including: performing wavelet packet decomposition on the high-frequency stress wave signal to separate preset characteristic frequency band components; calculating the energy spectral density distribution of each characteristic frequency band component; identifying local peak points in the energy spectral density distribution as stress wave energy mutation characteristics; counting the number of high-frequency stress wave events exceeding the background noise threshold per unit time; and converting the number of high-frequency stress wave events into the voltage equivalent of the event accumulation rate based on the sensitivity coefficient of the piezoelectric ceramic sensor. S4. Input the dynamic load gradient distribution parameters and the stress wave energy mutation characteristics into the coupled damage assessment model to generate a real-time stress risk coefficient, including: using the coupled damage assessment model to perform exponential decay correction on the dynamic load gradient distribution parameters to generate a load memory term; performing frequency domain differentiation on the stress wave energy mutation characteristics to generate a damage sensitive term; and performing convolution operation on the load memory term and the damage sensitive term to output the real-time stress risk coefficient. S5. When the voltage equivalent exceeds the adaptive threshold and the real-time stress risk coefficient meets the preset critical condition, a stress classification early warning command is triggered.

2. The method for early warning of the bearing capacity of tunnel steel arch frames as described in claim 1, characterized in that, The synchronous acquisition of multiple physical quantity electrical signals of key nodes of the steel arch frame includes: A triaxial strain gauge array is installed at the arch crown, arch waist and arch foot of the steel arch frame, and the micro-strain pulse signal generated when the steel arch frame is under load is captured by the triaxial strain gauge array. A laser displacement sensor array is used to measure the curvature change of the steel frame and output a curvature change voltage signal. High-frequency stress wave signals from rock fractures in tunnels are received using piezoelectric ceramic sensors. Align the sampling timestamps of all sensors using GPS clocks.

3. The method for early warning of the bearing capacity of tunnel steel arch frames as described in claim 1, characterized in that, The step of convolving the load memory term with the damage sensitivity term to output a real-time stress risk coefficient includes: The load memory term is discretized into a load memory vector, and the damage sensitivity term is discretized into a damage sensitivity vector; Extract the overlapping interval vector between the load memory vector and the damage sensitivity vector using a sliding time window; The sum of the dot products of the overlapping interval vectors is calculated as the real-time stress risk coefficient.

4. The method for early warning of the bearing capacity of tunnel steel arch frames as described in claim 1, characterized in that, The expression for the coupled damage assessment model is: ； in: These are the dynamic load gradient distribution parameters. This is the stress relaxation coefficient of the surrounding rock, and * is the convolution operator. It is the damage-sensitive term, It is the real-time stress risk coefficient. It is the current time point. It is a historical time variable. It is the acoustic emission energy spectral density. It's frequency. It is an exponentially decaying correction function.

5. The method for early warning of the bearing capacity of tunnel steel arch frames as described in claim 3, characterized in that, The discretized expression for the convolution operation is: ； in: It is the length of the sliding window. It is a discrete convolution output. It is the starting index of the time window. It is the window position index. It is a location Load memory vector at the location, It is a location Damage sensitivity vector at the location.

6. The method for early warning of the bearing capacity of a tunnel steel arch as described in claim 1, characterized in that, When the voltage equivalent exceeds the adaptive threshold and the real-time stress risk coefficient meets the preset critical condition, a stress classification early warning command is triggered, including: Baseline parameters for adaptive thresholding are trained based on historical security data; The offset parameter of the adaptive threshold is adjusted according to the real-time monitored deformation rate parameter of the surrounding rock. When the voltage equivalent exceeds the adjusted adaptive threshold and the real-time stress risk coefficient exceeds the critical condition for surrounding rock stability, a stress classification early warning command is triggered.

7. A tunnel steel arch frame bearing capacity early warning system, used to implement the tunnel steel arch frame bearing capacity early warning method according to any one of claims 1-6, characterized in that, The system includes the following modules: An electrical signal acquisition module is used to synchronously acquire multiple physical quantity electrical signals of key nodes of the steel arch frame, wherein the multiple physical quantity electrical signals include micro-strain pulse signals, curvature change voltage signals, and high-frequency stress wave signals; The load gradient distribution parameter calculation module is used to calculate dynamic load gradient distribution parameters characterizing the dynamic stress spatial distribution of the steel arch frame based on the micro-strain pulse signal and the curvature change voltage signal. The module includes: performing time differentiation on the micro-strain pulse signal to obtain the strain rate of change; calculating the curvature acceleration based on the curvature change voltage signal; setting the weight ratio of the strain rate of change and the curvature acceleration based on the surrounding rock stability level; and fusing the strain rate of change and the curvature acceleration according to the weight ratio to generate the dynamic load gradient distribution parameters characterizing the dynamic load distribution. The signal time-frequency domain decoupling module is used to decouple the high-frequency stress wave signal in the time-frequency domain and extract the voltage equivalent corresponding to the stress wave energy mutation characteristics and event accumulation rate. This includes: performing wavelet packet decomposition on the high-frequency stress wave signal to separate preset characteristic frequency band components; calculating the energy spectral density distribution of each characteristic frequency band component; identifying local peak points in the energy spectral density distribution as stress wave energy mutation characteristics; counting the number of high-frequency stress wave events exceeding the background noise threshold per unit time; and converting the number of high-frequency stress wave events into the voltage equivalent of the event accumulation rate based on the sensitivity coefficient of the piezoelectric ceramic sensor. The real-time stress risk coefficient generation module is used to input the dynamic load gradient distribution parameters and the stress wave energy mutation characteristics into a coupled damage assessment model to generate a real-time stress risk coefficient. This includes: applying an exponential decay correction to the dynamic load gradient distribution parameters using the coupled damage assessment model to generate a load memory term; performing frequency domain differentiation on the stress wave energy mutation characteristics to generate a damage-sensitive term; and performing a convolution operation between the load memory term and the damage-sensitive term to output the real-time stress risk coefficient. The early warning command triggering module is used to trigger a stress classification early warning command when the voltage equivalent exceeds the adaptive threshold and the real-time stress risk coefficient meets the preset critical conditions.

Images