Tunnel collapse emergency rescue material supply method and system

Abstract

The invention discloses a tunnel collapse emergency rescue material supply method and system, and relates to the technical field of tunnel collapse emergency rescue, and the method comprises the following steps: in the tunnel construction process, a rescue small-diameter pipeline is arranged from a tunnel face to a secondary lining end along a tunnel primary support surface or an excavation free face; high-precision pressure sensors are installed at the rescue end and the tunnel face end respectively and used for monitoring the pressure change condition in the pipeline in real time, and measurement data of all the pressure sensors are transmitted back to a monitoring center through wireless data transmission or wired optical fibers so as to be analyzed in real time; through the installed pressure sensor, pressure data of the tunnel face end are collected in real time, continuous time sequence data are established, and the collected data are preprocessed. According to the method, the survival probability of trapped persons in collapse accidents can be effectively increased, serious consequences caused by failure of a conveying system are reduced, and the tunnel construction safety and rescue efficiency are remarkably improved.

Description

Technical Field

[0001] The invention relates to the technical field of tunnel collapse emergency rescue, and in particular to a tunnel collapse emergency rescue material supply method and system. Background Art

[0002] The tunnel collapse emergency rescue material supply system is an emergency rescue channel specially used to provide basic survival materials for trapped personnel after a collapse accident occurs during tunnel construction. In the soft surrounding rock tunnel, this system lays one or more small-diameter pipes (about 25mm) from the face to the end of the second lining along the arch, waist or side wall of the initial support to transport oxygen, drinking water, milk or other nutrient solutions to maintain the vital signs of trapped personnel. This pipe is different from the escape pipe (the inner diameter of the escape pipe is not less than 80cm), and its main function is not to evacuate people, but to supply materials. In order to ensure the integrity and reliability of the pipeline when a collapse occurs, the small-diameter pipe uses a whole material without joints, has high tensile strength, extrusion resistance, excellent toughness, and is non-toxic and safe. The pipe wall preferably contains a steel wire reinforcement structure. Compared with the escape pipe, this system is easier to move forward during the continuous excavation of the face, does not affect normal construction, and can effectively make up for the failure of the escape pipe due to dislocation or damage after the collapse. Since small-diameter pipes are difficult to be broken by landslides, even if a landslide occurs, material transportation can still be ensured, which buys precious time for rescue and increases the survival probability of trapped people.

[0003] The prior art has the following deficiencies:

[0004] If one end of the pipeline is buried after a tunnel collapses, local negative pressure may form inside the pipeline, causing the transported liquid or gas to flow backwards or be completely blocked, or even forming a "siphon effect", resulting in the inability to transport rescue supplies to the face normally. When this happens, even if the trapped people are very close to the pipeline exit, they will not be able to get oxygen or drinking water, resulting in severe dehydration, hypoxia or even coma, reducing the probability of survival.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0006] The purpose of the present invention is to provide a method and system for supplying emergency rescue materials for tunnel collapse, which adopts high-precision pressure sensors to establish continuous time series data, and converts key pressure change data into feature vectors that can be processed by machine learning models through feature extraction technology, so as to accurately judge the risk of pipeline obstruction. When local negative pressure or transportation blockage trend is detected, the system can automatically switch to emergency control mode, start a small pressure compensation device (such as a micro air pump) at the face end to release positive pressure air, dynamically adjust the compensation force, prevent gas or liquid from flowing back or being completely blocked, and ensure smooth transportation of rescue materials to solve the problems in the above-mentioned background technology.

[0007] In order to achieve the above object, the present invention provides the following technical solution: a method for supplying emergency rescue materials for tunnel collapse, comprising the following steps:

[0008] Before tunnel construction, a small-diameter rescue pipe is laid along the tunnel interior from the tunnel face to the end of the secondary lining;

[0009] High-precision pressure sensors are installed at the rescue end and the face end to monitor the pressure changes inside the pipeline in real time. All the measurement data of the pressure sensors are transmitted back to the monitoring center via wireless data transmission or wired optical fiber for real-time analysis.

[0010] The pressure sensors installed are used to collect pressure data at the face end in real time, establish continuous time series data, and pre-process the collected data;

[0011] Filter out key data that can reflect the pressure drop from the preprocessed pressure data, and use feature extraction technology to convert the filtered data into feature vectors that can be understood and processed by the machine learning model;

[0012] The feature vector is input into a pre-trained machine learning model to predict pipeline obstruction and determine whether there is a risk of pipeline obstruction.

[0013] When the machine learning model detects that there is a hidden danger of obstruction in the pipeline transportation channel, it automatically switches to emergency control mode. In emergency control mode, the small pressure compensation device installed at the face end is immediately started, automatically releasing positive pressure air into the pipeline, and dynamically adjusting the amount of positive pressure air released according to the pressure changes predicted by the model.

[0014] Preferably, the installation location of the small-diameter rescue pipe is selected on the arch, arch waist or side wall to ensure that it does not affect normal construction and can be extended synchronously with the tunnel excavation.

[0015] Preferably, key data that can reflect the pressure drop are screened out from the preprocessed pressure data, wherein the screened key data include the gradient of the pressure change in space and the curvature of the pressure time series curve. Under the monitoring window, the screened gradient of the pressure change in space and the curvature of the pressure time series curve are converted into a pressure gradient index and a pressure curvature index through feature extraction technology. The pressure gradient index and the pressure curvature index are used as feature vectors that can be understood and processed by the machine learning model to quantify the severity of the pressure change at different positions in the delivery channel and the nonlinearity of the pressure time series change. By evaluating the spatial pressure unevenness and the temporal pressure mutation degree, the potential obstruction risk of the delivery channel is comprehensively reflected.

[0016] Preferably, the characteristic vector pressure gradient index and pressure curvature index are input into a pre-trained machine learning model, and a channel obstruction risk coefficient is generated through the model. The channel obstruction risk coefficient is used to predict pipeline obstruction and determine whether there is a risk of transportation channel obstruction.

[0017] Preferably, the channel obstruction risk coefficient is used to predict pipeline obstruction to determine whether there is a risk of transmission channel obstruction. The specific steps are as follows:

[0018] If the channel obstruction risk coefficient is greater than the channel obstruction risk coefficient threshold, it is determined that the transport channel has a risk of being obstructed;

[0019] If the channel obstruction risk coefficient is less than or equal to the channel obstruction risk coefficient threshold, it is determined that the transportation channel can achieve efficient emergency rescue material supply.

[0020] Preferably, in the emergency control mode, the small pressure compensation device installed at the face end is immediately started to automatically release positive pressure air into the pipeline, and dynamically adjust the amount of positive pressure air released according to the pressure change predicted by the model. The specific steps are as follows:

[0021] When a potential obstruction is detected in the pipeline, the initial release demand of the positive pressure air is calculated first, and the pressure compensation device is started. In the emergency control mode, the current pressure compensation intensity factor is calculated based on the difference between the channel obstruction risk coefficient and the channel obstruction risk coefficient threshold. The pressure compensation intensity factor determines the release amplitude of the positive pressure air. The calculation expression is as follows:

[0022]

[0023] , where: C p is the pressure compensation intensity factor, which is used to measure the positive pressure air compensation intensity that needs to be released at present. c is the currently calculated channel obstruction risk coefficient, R th is the channel obstruction risk coefficient threshold, α pis the system adjustment coefficient, which is used to control the reference value of compensation strength, β p is a nonlinear adjustment parameter used to strengthen or weaken the response curve of the compensation strength. p is the minimum compensation offset.

[0024] Preferably, after the positive pressure air release is triggered, the actual released air volume flow is calculated according to the change of the real-time pipeline pressure, and the pressure compensation strength is dynamically adjusted to ensure that the compensation process accurately matches the pipeline obstruction. The calculation expression of the actual released air volume flow is:

[0025]

[0026] , where: Q a is the actual air release, indicating the air volume flow rate released by the current air compensation system to the pipeline, V p is the effective volume per unit length of the pipeline, which means the volume of air that can be accommodated per meter of pipeline. δP is the current pipeline pressure change, that is, the difference between the most recent pressure measurement and the pressure measurement at the previous moment. ζ v is a nonlinear weight, controlling the impact of the risk factor of channel obstruction during the adjustment process, Q min The basic air release volume ensures that there is still a minimum air compensation volume even when the channel obstruction risk factor just exceeds the channel obstruction risk factor threshold.

[0027] Preferably, under the monitoring window, the specific steps of converting the gradient of pressure change in the screened space into a pressure gradient index by feature extraction technology are as follows:

[0028] The pressure transition factor is defined to measure the degree of sudden change of pressure difference in space. The calculation expression of the pressure transition factor is as follows:

[0029]

[0030] , where Λ PG is the pressure transition factor, which represents the degree of pressure difference propagation between adjacent monitoring points, P i Represents the instantaneous pressure value measured by the i-th pressure sensor, and its exponential transformation The detection sensitivity is enhanced by amplifying the influence of the pressure mutation region by power, where γ>1, d i+1，i is the physical distance between the i-th pressure sensor and the i+1-th pressure sensor, which is used to measure the pressure gradient change on the spatial scale. α is the distance correction coefficient, which is used to adjust the influence of the spatial scale on the gradient calculation to adapt to the distribution of different monitoring equipment. β is the pressure difference adjustment parameter to reduce false alarms. δ is the nonlinear gradient enhancement index, which takes 1<δ<2, and is used to amplify the influence of the pressure sudden change area. N represents the total number of pressure sensors.

[0031] In order to further extract the overall obstruction characteristics of the transmission channel, the final pressure gradient index is constructed by combining the gradient characteristics between large-span measuring points. The constructed expression is as follows:

[0032]

[0033] , where O gx is the overall pressure gradient index, which is used to measure the spatial pressure inhomogeneity of the entire pipeline transmission channel and reflect the risk of obstruction. η is the gradient index enhancement factor, η>1, which is used to further amplify the impact of the local mutation area to enhance the sensitivity to the risk of obstruction. -6 As a stability adjustment factor, it is used to prevent the denominator from approaching zero and improve the stability of the calculation. Γ is used as a high-order gradient correction parameter to introduce the gradient information between large-span measuring points, so that the pressure gradient index not only focuses on the local pressure difference, but also reflects the obstruction characteristics of the transportation channel in a wider range. i+2，i is the physical distance between the i-th pressure sensor and the i+2-th pressure sensor, which is used to characterize the pressure gradient characteristics between measurement points with a larger span. is the spatial span index, It is used to adjust the scale effect of gradient calculation so that the pressure gradient index can adapt to different pressure sensor distribution conditions.

[0034] Preferably, under the monitoring window, the specific steps of converting the curvature degree of the screened gradient and pressure time series curve into the pressure curvature index by feature extraction technology are as follows:

[0035] The pressure gradient change rate tensor is constructed to measure the spatial change trend and mutation degree of the pressure field. The pressure gradient change rate tensor is calculated by the change rate of the pressure gradient, the second-order gradient change, and the third-order gradient change. The constructed expression is as follows:

[0036]

[0037] , where PGVT is the pressure gradient change rate tensor, P represents the instantaneous pressure value, that is, the pressure state at any time within the monitoring window, is the pressure gradient vector, which indicates the rate of change of pressure in the spatial direction. The calculation expression is:

[0038]

[0039] ,in, and They represent the second-order partial derivatives of pressure on the x and y axes, respectively, measuring the local curvature of the pressure field and reflecting the non-uniformity of the local pressure gradient. and Represents the third-order partial derivatives of pressure on the x and y axes, respectively, and measures the degree of pressure mutation, that is, the high-order nonlinear change of the pressure field. ′ It is a nonlinear influencing factor used to adjust the influence weight between the second-order and third-order gradients to ensure that the curvature calculation is suitable for the tunnel transportation environment;

[0040] After calculating the pressure gradient change rate tensor, the pressure curvature index is further calculated to comprehensively evaluate the potential risk of obstruction of the transmission channel. The pressure curvature index combines the maximum curvature eigenvalue of the pressure time series curve and the cumulative effect of the pressure gradient. The calculation expression is as follows:

[0041]

[0042] , where: P ci is the pressure curvature index, λ max It represents the principal curvature eigenvalue, that is, the maximum curvature of the pressure time series curve in the local area, and is defined as follows: Where H(P) is the Hessian matrix of the pressure time series, and its eigenvalue λ max ,λ min Reflects the main direction change trend of the pressure surface, and selects the maximum eigenvalue λ max represents the direction of the most drastic local pressure change, β′ is the curvature nonlinear amplification factor, which controls the contribution of the principal curvature to enhance the sensitivity to the mutation area, and Ω is the spatial area within the monitoring window, which means that the pressure gradient change in the entire area is integrated to obtain the global obstruction risk assessment.

[0043] A tunnel collapse emergency rescue material supply system, comprising a rescue pipeline layout module, a pressure monitoring installation module, a data transmission and monitoring module, a pressure data acquisition and preprocessing module, an obstruction risk prediction module and a dynamic pressure compensation module;

[0044] The rescue pipe laying module is used to lay rescue small-diameter pipes along the tunnel from the tunnel face to the end of the secondary lining before tunnel construction.

[0045] Pressure monitoring installation module: high-precision pressure sensors are installed at the rescue end and the face end respectively to monitor the pressure changes inside the pipeline in real time. The measurement data of all pressure sensors are transmitted back to the monitoring center via wireless data transmission or wired optical fiber for real-time analysis;

[0046] The data transmission and monitoring module collects pressure data at the face end in real time through the installed pressure sensors, establishes continuous time series data, and pre-processes the collected data;

[0047] The pressure data collection and preprocessing module selects key data that can reflect the pressure drop from the preprocessed pressure data, and converts the selected data into feature vectors that can be understood and processed by the machine learning model through feature extraction technology;

[0048] The obstruction risk prediction module inputs the feature vector into a pre-trained machine learning model to predict pipeline obstruction and determine whether there is a risk of obstruction in the transmission channel;

[0049] The dynamic pressure compensation module automatically switches to emergency control mode when the machine learning model detects that the pipeline transportation channel is blocked. In the emergency control mode, the small pressure compensation device installed at the face end starts immediately, automatically releasing positive pressure air into the pipeline, and dynamically adjusts the amount of positive pressure air released according to the pressure changes predicted by the model.

[0050] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0051] The present invention adopts high-precision pressure sensors to establish continuous time series data, and converts key pressure change data into feature vectors that can be processed by machine learning models through feature extraction technology, so as to accurately judge the risk of pipeline obstruction. When local negative pressure or transportation blockage trend is detected, the system can automatically switch to emergency control mode, start the small pressure compensation device (such as a micro air pump) at the face end to release positive pressure air, dynamically adjust the compensation force, prevent gas or liquid from flowing back or being completely blocked, and ensure the smooth transportation of rescue materials. Through this intelligent dynamic control scheme, the present invention can effectively improve the survival probability of trapped people in landslide accidents, reduce the serious consequences caused by failure of the transportation system, and significantly improve the safety of tunnel construction and rescue efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0053] Figure 1 The present invention is a method flow chart of a method for supplying emergency rescue materials for tunnel collapse.

[0054] Figure 2 The present invention is a module schematic diagram of a tunnel collapse emergency rescue material supply system. DETAILED DESCRIPTION

[0055] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0056] The present invention provides Figure 1 A tunnel collapse emergency rescue material supply method is shown, comprising the following steps:

[0057] Before tunnel construction, a small-diameter rescue pipe (with an inner diameter of about 25 mm) should be reasonably laid out along the tunnel interior from the tunnel face to the end of the secondary lining;

[0058] Small-diameter rescue pipes must be made of materials without joints, whole, tensile, extrusion-resistant, highly tough, non-toxic, and with steel wire reinforced structures on the inner wall to ensure that they will not be easily damaged or affected by landslides during construction. The installation location of small-diameter rescue pipes should be preferentially selected at the arch, arch waist or side wall to ensure that they do not affect normal construction and can be extended synchronously with tunnel excavation. This step is the basis of the entire rescue system, ensuring that the rescue transportation channel is preset when a landslide occurs, so that subsequent pressure monitoring, material transportation, and intelligent regulation can be carried out efficiently. The high-strength design and reasonable layout of small-diameter rescue pipes can improve their survivability in extreme situations and avoid damage or deformation of the pipes due to landslides.

[0059] High-precision pressure sensors are installed at the rescue end (near the tunnel entrance) and the face end (deep in the tunnel) to monitor the pressure changes inside the pipeline in real time. The measurement data of all pressure sensors are transmitted back to the monitoring center via wireless data transmission or wired optical fiber for real-time analysis;

[0060] The sensor must have high sensitivity, vibration resistance, waterproof and dustproof, impact resistance and other characteristics to ensure stable operation in complex construction environments with high temperature, high humidity and high dust. The measurement data of all pressure sensors must be transmitted back to the monitoring center via wireless data transmission or wired optical fiber for real-time analysis. This step provides core basic data for subsequent data analysis and machine learning prediction, ensuring that the pressure status of the conveying channel can be mastered in real time, thereby providing early warning of potential conveying obstacles.

[0061] The pressure sensors installed are used to collect pressure data at the face end in real time, establish continuous time series data, and pre-process the collected data;

[0062] Since the collected raw data may have problems such as noise, instantaneous fluctuations, and abnormal data points, data preprocessing is required, including (the following is just an enumeration, not exhaustive):

[0063] Noise filtering (using low-pass filtering, sliding average, etc. to remove meaningless random fluctuations).

[0064] Data debiasing (correcting abnormal drift caused by equipment errors or environmental factors).

[0065] This step ensures that the data input into the machine learning model is reliable and representative through data preprocessing, thereby improving the accuracy and robustness of the prediction model.

[0066] Filter out key data that can reflect the pressure drop from the preprocessed pressure data, and use feature extraction technology to convert the filtered data into feature vectors that can be understood and processed by the machine learning model;

[0067] The key data that can reflect the pressure drop are screened out from the preprocessed pressure data, wherein the screened key data include the gradient of pressure change in space and the curvature of the pressure time series curve. Under the monitoring window, the screened gradient of pressure change in space and the curvature of the pressure time series curve are converted into a pressure gradient index and a pressure curvature index through feature extraction technology. The pressure gradient index and the pressure curvature index are used as feature vectors that can be understood and processed by the machine learning model to quantify the severity of pressure changes at different positions in the transportation channel and the nonlinearity of the pressure time series changes. By evaluating the spatial pressure unevenness and the temporal pressure mutation degree, the potential obstruction risk of the transportation channel is comprehensively reflected.

[0068] A large gradient of pressure change in space usually indicates that there is a risk of obstruction in the transportation channel. Under normal circumstances, the pressure inside the transportation pipeline should be relatively uniform or show a smooth gradual trend. However, when a blockage, deformation or landslide occurs somewhere in the pipeline, an obvious local pressure mutation area will be formed at that location, resulting in a sharp increase in the upstream and downstream pressure difference, thereby forming a high-pressure gradient. If the pressure change measured by the pressure sensor at different locations shows a nonlinear jump, that is, the pressure in a certain area drops or rises suddenly, while other areas remain stable, it means that there is a fluid transport obstruction in the area or a sudden change in the cross-sectional area inside the pipeline. For example, under the influence of a landslide, the pipeline may be partially squeezed and deformed, resulting in a smaller pipe diameter, thereby increasing the fluid resistance in the area, causing the upstream pressure to rise rapidly, while the downstream pressure drops rapidly, forming a high-gradient pressure distribution.

[0069] Under the monitoring window, the specific steps of converting the gradient of pressure change in the screened space into the pressure gradient index by feature extraction technology are as follows:

[0070] In the transmission channel, the pressure values ​​measured by adjacent sensors should show a smooth change in space. However, when the pipeline is blocked, the pressure change in the local area will show a sudden change. Therefore, the pressure transition factor is defined to measure the degree of sudden change of pressure difference in space. The calculation expression of the pressure transition factor is as follows:

[0071]

[0072] , where Λ PG is the pressure transition factor, which represents the degree of pressure difference propagation between adjacent monitoring points. The larger the value, the more drastic the pressure change in the area, and there may be a risk of obstruction of the transmission channel. i Represents the instantaneous pressure value measured by the i-th pressure sensor, and its exponential transformation The detection sensitivity is enhanced by amplifying the influence of the pressure mutation region by power, where γ>1, d i+1，i is the physical distance between the i-th pressure sensor and the i+1-th pressure sensor. This parameter is used to measure the pressure gradient change on the spatial scale. α is the distance correction coefficient, which is used to adjust the influence of the spatial scale on the gradient calculation to adapt to the distribution of different monitoring equipment. β is the pressure difference adjustment parameter, which suppresses excessive response in the case of small pressure fluctuations. It only participates in the calculation significantly when the pressure mutation reaches a certain level, thereby reducing false alarms. δ is the nonlinear gradient enhancement index, which is usually 1<δ<2, and is used to amplify the influence of the pressure sudden change area, so that the contribution of this area is much greater than that of the normal pressure change area. N represents the total number of pressure sensors;

[0073] The purpose of this step is to capture the pressure jump between adjacent sensors. Enhance the impact of pressure mutation and introduce nonlinear gradient enhancement term |P i+1 -P i | δ , ensuring that when the transportation channel is partially blocked, the characteristics of the area are accurately magnified to increase the system's sensitivity to hidden dangers of obstruction.

[0074] In order to further extract the overall obstruction characteristics of the transmission channel, the final pressure gradient index is constructed by combining the gradient characteristics between large-span measuring points. The constructed expression is as follows:

[0075]

[0076] , where P gx is the overall pressure gradient index, which is used to measure the spatial pressure inhomogeneity of the entire pipeline transmission channel and reflect the risk of obstruction. η is the gradient index enhancement factor, η>1, which is used to further amplify the impact of the local mutation area to enhance the sensitivity to the risk of obstruction. -6As a stability adjustment factor, it is used to prevent the denominator from approaching zero, improve the stability of the calculation, and ensure that the calculation results can work reliably in any environment. Γ is used as a high-order gradient correction parameter to introduce the gradient information between large-span measuring points, so that the pressure gradient index not only focuses on the local pressure difference, but also reflects the obstruction characteristics of the transportation channel in a wider range. i+2，i is the physical distance between the i-th pressure sensor and the i+2-th pressure sensor. This parameter is used to characterize the pressure gradient characteristics between measuring points with a larger span. is the spatial span index, usually taken as Used to adjust the scale effect of gradient calculation so that the pressure gradient index can adapt to different pressure sensor distribution conditions;

[0077] The purpose of this step is to construct the final pressure gradient index based on the pressure transition factor calculated in the first step and the gradient information between the larger span measuring points. The numerator of the pressure gradient index strengthens the local pressure mutation, and the denominator introduces a wider range of pressure gradient correction to ensure that when the pipeline is blocked, not only the local sensor data will change, but the pressure pattern of the entire transmission channel can also be identified.

[0078] It can be seen from the pressure gradient index that under the monitoring window, the gradient of pressure change in the screened space is converted into a pressure gradient index through feature extraction technology. The larger the performance value of the pressure gradient index, the higher the risk of obstruction of the transmission channel, and vice versa. The pressure gradient index reveals the unevenness of the pressure distribution inside the pipeline by quantifying the pressure change gradient between different sensors in the monitoring window. When the pressure gradient index increases significantly, it means that there is a drastic pressure jump between adjacent sensors, which is usually caused by pipeline obstruction, deformation or collapse, resulting in a sudden change in local pressure. This abnormal fluctuation indicates that the transmission channel may encounter serious obstruction, affecting the normal delivery of rescue materials, and threatening the survival probability of trapped people. On the contrary, a low pressure gradient index indicates that the pressure distribution in the pipeline is uniform, the change is stable, the transmission channel operates normally, and the risk of obstruction is low.

[0079] A large curvature of the pressure time series curve at the face end usually indicates that the transportation channel may be blocked. Specifically, the curvature of the pressure time series curve reflects the nonlinear characteristics of pressure changes. Under normal transportation conditions, the pressure should be relatively stable and the curve changes relatively slowly. Once the curvature of the curve increases, it means that the pressure value fluctuates greatly in a short period of time, which may be due to blockage, local collapse, pipeline compression deformation, or fluid flow obstruction in the transportation channel. Especially in the case of local negative pressure or abnormal increase in pressure gradient, the curvature of the curve will increase significantly, which means that the pressure state is unstable and material transportation may be hindered.

[0080] In the monitoring window, the specific steps of converting the curvature of the screened gradient and pressure time series curve into the pressure curvature index by feature extraction technology are as follows:

[0081] The pressure gradient change rate tensor is constructed to measure the spatial change trend and mutation degree of the pressure field. The pressure gradient change rate tensor is calculated by the change rate of the pressure gradient, the second-order gradient change, and the third-order gradient change. The constructed expression is as follows:

[0082]

[0083] , where PGVT is the pressure gradient change rate tensor, P represents the instantaneous pressure value, that is, the pressure state at any time within the monitoring window, is the pressure gradient vector, which indicates the rate of change of pressure in the spatial direction. The calculation expression is:

[0084]

[0085] ,in, and They represent the second-order partial derivatives of pressure on the x and y axes, respectively, measuring the local curvature of the pressure field and reflecting the non-uniformity of the local pressure gradient. and Represents the third-order partial derivatives of pressure on the x and y axes, respectively, and measures the degree of pressure mutation, that is, the high-order nonlinear change of the pressure field. ′ It is a nonlinear influencing factor used to adjust the influence weight between the second-order and third-order gradients. It is usually set based on historical monitoring data to ensure that the curvature calculation is suitable for the tunnel transportation environment.

[0086] The purpose of this step is to calculate the pressure gradient rate tensor, and quantify the pressure change trend and mutation in the transmission channel by comprehensively considering the spatial gradient, second-order curvature and third-order rate of change. When the pressure gradient rate tensor is large, it means that the transmission channel may be blocked, the pipeline may be deformed or collapsed, resulting in increased fluid disturbance, which provides a basis for the subsequent calculation of the pressure curvature index.

[0087] After calculating the pressure gradient change rate tensor, the pressure curvature index is further calculated to comprehensively evaluate the potential risk of obstruction of the transmission channel. The pressure curvature index combines the maximum curvature eigenvalue of the pressure time series curve and the cumulative effect of the pressure gradient. The calculation expression is as follows:

[0088]

[0089] , where: P ci is the pressure curvature index, λ max It represents the principal curvature eigenvalue, that is, the maximum curvature of the pressure time series curve in the local area, and is defined as follows: Where H(P) is the Hessian matrix of the pressure time series, and its eigenvalue λ max ,λ min Reflects the main direction change trend of the pressure surface, and selects the maximum eigenvalue λ max represents the direction of the most drastic local pressure change, β′ is the curvature nonlinear amplification factor, which controls the contribution of the principal curvature and is generally set to 1.5-2.0 to enhance the sensitivity to the mutation area, and Ω is the spatial area within the monitoring window, which means that the pressure gradient change in the entire area is integrated to obtain the global obstruction risk assessment;

[0090] The purpose of this step is to quantify the risk of obstruction in the delivery channel by constructing a pressure curvature index. When the pressure curvature index value is high, it means that there is a significant pressure mutation and gradient unevenness inside the pipeline, indicating that the channel may be obstructed or even blocked. By continuously monitoring the pressure curvature index, the abnormal state of the delivery channel can be predicted in advance to ensure that the relief supplies can be delivered stably.

[0091] The eigenvalue λ of the Hessian matrix is ​​solved by the characteristic equation, and the solution expression is:

[0092]

[0093] , solving this second-order characteristic equation can obtain two eigenvalues ​​λ1 and λ2 (where λ max =max(λ1,λ2)), specifically:

[0094]

[0095] , where: max is the larger of the two eigenvalues, λ min is the smaller of the two eigenvalues.

[0096] It can be seen from the pressure curvature index that under the monitoring window, the screened gradient and the curvature of the pressure time series curve are converted into the pressure curvature index through feature extraction technology. The larger the performance value of the pressure curvature index, the higher the risk of obstruction of the transmission channel, and vice versa. This is because the pressure curvature index directly measures the curvature of the pressure time series curve and the drastic change of the pressure gradient. Under the monitoring window, the pressure gradient, local curvature (Hessian matrix eigenvalue) and high-order nonlinear changes (such as third-order derivatives) screened by feature extraction technology jointly determine the value of the pressure curvature index. When the transmission channel is blocked (such as local collapse, pipeline deformation, material accumulation, etc.), the pressure in the pipeline will produce significant spatial and temporal changes, resulting in drastic fluctuations in the second-order and third-order partial derivatives, thereby increasing the pressure curvature index. A higher pressure curvature index means increased pressure inhomogeneity, obstructed flow of fluid or gas, and the formation of abnormal pressure gradients, so it can be used to predict whether there is a serious risk of obstruction in the transmission system. On the contrary, if the pressure curvature index is low, it indicates that the pressure field is relatively stable, the fluid transport is smooth, and there is no obvious sign of obstruction.

[0097] The feature vector is input into a pre-trained machine learning model to predict pipeline obstruction and determine whether there is a risk of pipeline obstruction.

[0098] The characteristic vector pressure gradient index and pressure curvature index are input into a pre-trained machine learning model, and the channel obstruction risk coefficient is generated through the model. The channel obstruction risk coefficient is used to predict pipeline obstruction and determine whether there is a risk of obstruction in the transportation channel.

[0099] A pre-trained machine learning model refers to an algorithm system that fully trains and verifies the model through historical data, experimental data and expert knowledge before the system is officially put into use, so that the model can capture and identify the complex patterns and laws of pressure changes in the tunnel. In this system, engineers first collected a large amount of historical data and experimental data related to tunnel construction, pipeline transportation, and tunnel collapse. These data not only include conventional pressure values, but also cover key parameters generated after various preprocessing and feature extraction, such as pressure gradient index and pressure curvature index. The pressure gradient index and pressure curvature index reflect the rate of change of pressure in the pipeline over time or space and the degree of curvature of the change curve, respectively. They can capture abnormal signals caused by emergencies in subtle pressure fluctuations. Using these data, engineers used supervised learning methods to input a large number of instances with normal and blocked channel states into the machine learning model for training. Through the complex nonlinear relationship contained in the model adaptive learning data, the model gradually mastered how to extract effective information from multidimensional features such as pressure gradient index and pressure curvature index, thereby generating a channel obstruction risk coefficient. The channel obstruction risk coefficient is a numerical indicator that reflects the possibility or risk level of the pipeline transportation channel being blocked under the current state. After training, verification and cross-testing with a large amount of data, the machine learning model has a high accuracy and robustness. In actual applications, it can predict and promptly warn of possible obstructions in the transportation channel based on real-time collected pressure data, thus providing a scientific basis for rescue decisions.

[0100] Pre-trained machine learning models play a vital role in the system because the tunnel construction environment is extremely complex and is affected by many factors, such as surrounding rock pressure, construction vibration, and changes in ambient temperature and humidity, which can cause pressure data to fluctuate to varying degrees. By learning various sudden abnormal situations in historical data, the model can identify those pressure abnormality patterns that are not simple random noise but potential risks. In particular, by inputting the pressure gradient index and pressure curvature index in the feature vector into the model, the model can perform multi-dimensional and in-depth analysis of these features to generate a quantitative channel obstruction risk coefficient. The pre-trained model means that before it is put into use, the model has been fully verified with historical data and parameter adjustment to ensure that it can make judgments quickly and accurately when encountering similar historical situations or newly emerging similar data. Through continuous feedback and online learning, the model can also be further updated and optimized as new data is added, thereby improving its prediction accuracy and adaptability. This early warning mechanism based on machine learning greatly makes up for the shortcomings of traditional static monitoring methods, enabling the entire rescue system to always maintain a high degree of sensitivity and response speed in the face of complex and changeable tunnel environments, thereby ensuring the stable delivery of emergency rescue materials and the safety of trapped people.

[0101] The machine learning model is not specifically limited here, and can realize the pressure gradient index P gx and the pressure curvature index P ci Perform comprehensive analysis to generate the channel obstruction risk coefficient R c In order to realize the technical solution of the present invention, the present invention provides a specific implementation method; the channel obstruction risk coefficient R c The resulting calculation formula is:

[0102]

[0103] , where τ x , τ y They are pressure gradient index P gx and the pressure curvature index P ci The preset proportionality factor, and τ x , τ y All are greater than 0. Preset proportional coefficient (τ x , τ y ) in this formula is used to adjust the pressure gradient index P gx and the pressure curvature index P ci The risk factor of channel obstruction R c In practical applications, the numerical range and changing trend of the pressure gradient index and the pressure curvature index may be different, and simply adding or averaging them may not accurately reflect their contribution to risk assessment. Therefore, τ is introduced. x and τ y As a weight factor, the model calculates the channel obstruction risk coefficient R c When the pressure gradient index P is gx The impact on the risk of channel obstruction is greater, and a larger τ can be assigned x , in order to increase its risk factor R in calculating channel obstruction c ; On the contrary, if the pressure curvature index P ci The contribution is more significant, so increase τ y These preset proportionality factors are usually determined based on historical data analysis, expert experience, or the training results of machine learning models to ensure that the calculated channel obstruction risk factor R c It has more engineering practical significance and prediction accuracy.

[0104] The channel obstruction risk coefficient is used to predict pipeline obstruction and determine whether there is a risk of transmission channel obstruction. The specific steps are as follows:

[0105] If the channel obstruction risk coefficient is greater than the channel obstruction risk coefficient threshold, it is determined that the transport channel has a risk of being obstructed;

[0106] If the channel obstruction risk coefficient is less than or equal to the channel obstruction risk coefficient threshold, it is determined that the transportation channel can achieve efficient emergency rescue material supply.

[0107] When the machine learning model detects that there is a hidden danger of obstruction in the pipeline transportation channel, it automatically switches to the emergency control mode. In the emergency control mode, the small pressure compensation device (such as a micro air pump) installed at the face end is immediately started to automatically release positive pressure air into the pipeline, and dynamically adjust the amount of positive pressure air released according to the pressure change predicted by the model. If the pressure change trend further deteriorates, the output pressure of the air pump is automatically increased. If the pressure returns to normal, the release of positive pressure air is reduced or stopped to maintain the normal flow state inside the pipeline, ensuring that the positive pressure air can timely and continuously compensate for the pressure drop in the pipeline.

[0108] In the emergency control mode, the small pressure compensation device (such as a micro air pump) installed at the face end is immediately started to automatically release positive pressure air into the pipeline, and dynamically adjust the amount of positive pressure air released according to the pressure changes predicted by the model. The specific steps are as follows:

[0109] When a potential obstruction is detected in the pipeline, the initial release demand of the positive pressure air is calculated first, and the pressure compensation device is started. In the emergency control mode, the current pressure compensation intensity factor is calculated based on the difference between the channel obstruction risk coefficient and the channel obstruction risk coefficient threshold. The pressure compensation intensity factor determines the release amplitude of the positive pressure air. The calculation expression is as follows:

[0110]

[0111] , where: C p is the pressure compensation intensity factor, which is used to measure the positive pressure air compensation intensity that needs to be released at present. c is the currently calculated channel obstruction risk coefficient, R th is the channel obstruction risk coefficient threshold, α p is the system adjustment coefficient, which is used to control the reference value of compensation strength, β p is a nonlinear adjustment parameter used to strengthen or weaken the response curve of the compensation strength. p The minimum compensation offset ensures that there is a certain degree of compensation even when the channel obstruction risk factor just exceeds the channel obstruction risk factor threshold;

[0112] This step calculates the positive pressure air release intensity that currently needs to be compensated to ensure that the compensation mechanism is triggered immediately after the model determines that the obstruction risk exceeds the threshold. Nonlinear adjustment parameters are introduced to adapt to different levels of risks and enhance the system's sensitivity to risk changes. The minimum offset ensures that even if there is a slight obstruction, fine-tuning can be done in advance to prevent further deterioration.

[0113] After the positive pressure air release is triggered, the actual released air volume flow is calculated according to the change of real-time pipeline pressure, and the pressure compensation intensity is dynamically adjusted to ensure that the compensation process accurately matches the pipeline obstruction situation. The calculation expression of the actual released air volume flow is:

[0114]

[0115] , where: Q a is the actual air release, indicating the air volume flow rate released by the current air compensation system to the pipeline, V p is the effective volume per unit length of the pipeline, which means the volume of air that can be accommodated per meter of pipeline. δP is the current pipeline pressure change, that is, the difference between the most recent pressure measurement and the pressure measurement at the previous moment. ζ v is a nonlinear weight, controlling the impact of the risk factor of channel obstruction during the adjustment process, Q min The basic air release volume ensures that even when the channel obstruction risk factor just exceeds the channel obstruction risk factor threshold, there is still a minimum air compensation volume;

[0116] The purpose of this step is to ensure that when the pipeline is blocked due to tunnel collapse, the air compensation system can accurately match the pressure changes inside the pipeline and realize dynamic adaptive compensation. The formula calculates the current required air release amount by the relative difference between the channel obstruction risk coefficient and the channel obstruction risk coefficient threshold, and combines the effective volume per unit length of the pipeline to ensure that the calculated air compensation amount is consistent with the actual capacity of the pipeline. In addition, the current pipeline pressure change is used as the pipeline pressure change, which ensures that the system can perceive the pressure fluctuation in real time and dynamically adjust the compensation intensity according to the nonlinear weight, making the system response smoother and more accurate. At the same time, the basic air release amount is introduced to ensure that even if the obstruction risk index just exceeds the threshold, the minimum air compensation can be provided to avoid compensation lag and cause material delivery failure. Overall, this step ensures that the internal pressure of the pipeline can quickly return to the normal range by calculating the optimal air release amount and dynamically adjusting the compensation intensity, thereby maintaining the continuous and efficient delivery of rescue materials and increasing the survival probability of trapped people.

[0117] The present invention adopts high-precision pressure sensors to establish continuous time series data, and converts key pressure change data into feature vectors that can be processed by machine learning models through feature extraction technology, so as to accurately judge the risk of pipeline obstruction. When local negative pressure or transportation blockage trend is detected, the system can automatically switch to emergency control mode, start the small pressure compensation device (such as a micro air pump) at the face end to release positive pressure air, dynamically adjust the compensation force, prevent gas or liquid from flowing back or being completely blocked, and ensure the smooth transportation of rescue materials. Through this intelligent dynamic control scheme, the present invention can effectively improve the survival probability of trapped people in landslide accidents, reduce the serious consequences caused by failure of the transportation system, and significantly improve the safety of tunnel construction and rescue efficiency.

[0118] The present invention provides Figure 2 A tunnel collapse emergency rescue material supply system shown includes a rescue pipeline layout module, a pressure monitoring installation module, a data transmission and monitoring module, a pressure data acquisition and preprocessing module, an obstruction risk prediction module and a dynamic pressure compensation module;

[0119] The rescue pipe laying module is used to lay rescue small-diameter pipes along the tunnel from the tunnel face to the end of the secondary lining before tunnel construction.

[0120] Pressure monitoring installation module: high-precision pressure sensors are installed at the rescue end and the face end respectively to monitor the pressure changes inside the pipeline in real time. The measurement data of all pressure sensors are transmitted back to the monitoring center via wireless data transmission or wired optical fiber for real-time analysis;

[0121] The data transmission and monitoring module collects pressure data at the face end in real time through the installed pressure sensors, establishes continuous time series data, and pre-processes the collected data;

[0122] The pressure data collection and preprocessing module selects key data that can reflect the pressure drop from the preprocessed pressure data, and converts the selected data into feature vectors that can be understood and processed by the machine learning model through feature extraction technology;

[0123] The obstruction risk prediction module inputs the feature vector into a pre-trained machine learning model to predict pipeline obstruction and determine whether there is a risk of obstruction in the transmission channel;

[0124] The dynamic pressure compensation module automatically switches to emergency control mode when the machine learning model detects that the pipeline transportation channel is blocked. In the emergency control mode, the small pressure compensation device installed at the face end starts immediately, automatically releasing positive pressure air into the pipeline, and dynamically adjusts the amount of positive pressure air released according to the pressure changes predicted by the model.

[0125] A tunnel collapse emergency rescue material supply method provided by an embodiment of the present invention is realized by the above-mentioned tunnel collapse emergency rescue material supply system. The specific method and process of a tunnel collapse emergency rescue material supply system are detailed in the embodiment of the above-mentioned tunnel collapse emergency rescue material supply method, which will not be repeated here.

[0126] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0127] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0128] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for supplying emergency rescue materials for tunnel collapse, characterized in that: The following steps are involved: Before tunnel construction, a small-diameter rescue pipe is laid along the tunnel interior from the tunnel face to the end of the secondary lining; High-precision pressure sensors are installed at the rescue end and the face end to monitor the pressure changes inside the pipeline in real time. All the measurement data of the pressure sensors are transmitted back to the monitoring center via wireless data transmission or wired optical fiber for real-time analysis. The pressure sensors installed are used to collect pressure data at the face end in real time, establish continuous time series data, and pre-process the collected data; Filter out key data that can reflect the pressure drop from the preprocessed pressure data, and use feature extraction technology to convert the filtered data into feature vectors that can be understood and processed by the machine learning model; The feature vector is input into a pre-trained machine learning model to predict pipeline obstruction and determine whether there is a risk of pipeline obstruction. When the machine learning model detects that there is a hidden danger of obstruction in the pipeline transportation channel, it automatically switches to emergency control mode. In emergency control mode, the small pressure compensation device installed at the face end is immediately started, automatically releasing positive pressure air into the pipeline, and dynamically adjusting the amount of positive pressure air released according to the pressure changes predicted by the model.

2. A tunnel collapse emergency rescue material supply method according to claim 1, characterized in that: The installation location for the small-diameter rescue pipe is selected on the arch top, arch waist or side wall to ensure that it does not affect normal construction and can be extended synchronously with the tunnel excavation.

3. A tunnel collapse emergency rescue material supply method according to claim 1, characterized in that: The key data that can reflect the pressure drop are screened out from the preprocessed pressure data, wherein the screened key data include the gradient of pressure change in space and the curvature of the pressure time series curve. Under the monitoring window, the screened gradient of pressure change in space and the curvature of the pressure time series curve are converted into a pressure gradient index and a pressure curvature index through feature extraction technology. The pressure gradient index and the pressure curvature index are used as feature vectors that can be understood and processed by the machine learning model to quantify the severity of pressure changes at different positions in the transportation channel and the nonlinearity of the pressure time series changes. By evaluating the spatial pressure unevenness and the temporal pressure mutation degree, the potential obstruction risk of the transportation channel is comprehensively reflected.

4. A tunnel collapse emergency rescue material supply method according to claim 3, characterized in that: The characteristic vector pressure gradient index and pressure curvature index are input into a pre-trained machine learning model, and the channel obstruction risk coefficient is generated through the model. The channel obstruction risk coefficient is used to predict pipeline obstruction and determine whether there is a risk of obstruction in the transportation channel.

5. A tunnel collapse emergency rescue material supply method according to claim 4, characterized in that: The channel obstruction risk coefficient is used to predict pipeline obstruction and determine whether there is a risk of transmission channel obstruction. The specific steps are as follows: If the channel obstruction risk coefficient is greater than the channel obstruction risk coefficient threshold, it is determined that the transport channel has a risk of being obstructed; If the channel obstruction risk coefficient is less than or equal to the channel obstruction risk coefficient threshold, it is determined that the transportation channel can achieve efficient emergency rescue material supply.

6. A tunnel collapse emergency rescue material supply method according to claim 5, characterized in that: In the emergency control mode, the small pressure compensation device installed at the face end is immediately activated to automatically release positive pressure air into the pipeline and dynamically adjust the amount of positive pressure air released according to the pressure changes predicted by the model. The specific steps are as follows: When a potential obstruction is detected in the pipeline, the initial release demand of the positive pressure air is calculated first, and the pressure compensation device is started. In the emergency control mode, the current pressure compensation intensity factor is calculated based on the difference between the channel obstruction risk coefficient and the channel obstruction risk coefficient threshold. The pressure compensation intensity factor determines the release amplitude of the positive pressure air. The calculation expression is as follows: ， Where: C p is the pressure compensation intensity factor, which is used to measure the positive pressure air compensation intensity that needs to be released at present. c is the currently calculated channel obstruction risk coefficient, R th is the channel obstruction risk coefficient threshold, α p is the system adjustment coefficient, which is used to control the reference value of compensation strength, β p is a nonlinear adjustment parameter used to strengthen or weaken the response curve of the compensation strength. p is the minimum compensation offset.

7. A tunnel collapse emergency rescue material supply method according to claim 6, characterized in that: After the positive pressure air release is triggered, the actual released air volume flow is calculated according to the change of real-time pipeline pressure, and the pressure compensation intensity is dynamically adjusted to ensure that the compensation process accurately matches the pipeline obstruction situation. The calculation expression of the actual released air volume flow is: ， Where: Q a is the actual air release, indicating the air volume flow rate released by the current air compensation system to the pipeline, V p is the effective volume per unit length of the pipeline, which means the volume of air that can be accommodated per meter of pipeline. δP is the current pipeline pressure change, that is, the difference between the most recent pressure measurement and the pressure measurement at the previous moment. ζ v is a nonlinear weight, controlling the impact of the risk factor of channel obstruction during the adjustment process, Q min The basic air release volume ensures that there is still a minimum air compensation volume even when the channel obstruction risk factor just exceeds the channel obstruction risk factor threshold.

8. A tunnel collapse emergency rescue material supply method according to claim 3, characterized in that: Under the monitoring window, the specific steps of converting the gradient of pressure change in the screened space into the pressure gradient index by feature extraction technology are as follows: The pressure transition factor is defined to measure the degree of sudden change of pressure difference in space. The calculation expression of the pressure transition factor is as follows: ， Among them, Λ PG is the pressure transition factor, which represents the degree of pressure difference propagation between adjacent monitoring points, P i Represents the instantaneous pressure value measured by the i-th pressure sensor, and its exponential transformation The detection sensitivity is enhanced by amplifying the influence of the pressure mutation region by power, where γ>1, d i+1，i is the physical distance between the i-th pressure sensor and the i+1-th pressure sensor, which is used to measure the pressure gradient change on the spatial scale. α is the distance correction coefficient, which is used to adjust the influence of the spatial scale on the gradient calculation to adapt to the distribution of different monitoring equipment. β is the pressure difference adjustment parameter to reduce false alarms. δ is the nonlinear gradient enhancement index, which takes 1<δ<2, and is used to amplify the influence of the pressure sudden change area. N represents the total number of pressure sensors. In order to further extract the overall obstruction characteristics of the transmission channel, the final pressure gradient index is constructed by combining the gradient characteristics between large-span measuring points. The constructed expression is as follows: ， Among them, P gx is the overall pressure gradient index, which is used to measure the spatial pressure inhomogeneity of the entire pipeline transmission channel and reflect the risk of obstruction. η is the gradient index enhancement factor, η>1, which is used to further amplify the impact of the local mutation area to enhance the sensitivity to the risk of obstruction. -6 As a stability adjustment factor, it is used to prevent the denominator from approaching zero and improve the stability of the calculation. Γ is used as a high-order gradient correction parameter to introduce the gradient information between large-span measuring points, so that the pressure gradient index not only focuses on the local pressure difference, but also reflects the obstruction characteristics of the transportation channel in a wider range. i+2，i is the physical distance between the i-th pressure sensor and the i+2-th pressure sensor, which is used to characterize the pressure gradient characteristics between measurement points with a larger span. is the spatial span index, It is used to adjust the scale effect of gradient calculation so that the pressure gradient index can adapt to different pressure sensor distribution conditions.

9. A tunnel collapse emergency rescue material supply method according to claim 3, characterized in that: Under the monitoring window, the specific steps of converting the curvature of the screened gradient and pressure time series curve into the pressure curvature index by feature extraction technology are as follows: The pressure gradient change rate tensor is constructed to measure the spatial change trend and mutation degree of the pressure field. The pressure gradient change rate tensor is calculated by the change rate of the pressure gradient, the second-order gradient change, and the third-order gradient change. The constructed expression is as follows: ， Among them, PGVT is the pressure gradient change rate tensor, P represents the instantaneous pressure value, that is, the pressure state at any time in the monitoring window, is the pressure gradient vector, which indicates the rate of change of pressure in the spatial direction. The calculation expression is: ， in, and They represent the second-order partial derivatives of pressure on the x and y axes, respectively, measuring the local curvature of the pressure field and reflecting the non-uniformity of the local pressure gradient. and Represents the third-order partial derivatives of pressure on the x and y axes, respectively, and measures the degree of pressure mutation, that is, the high-order nonlinear change of the pressure field. ′ It is a nonlinear influencing factor used to adjust the influence weight between the second-order and third-order gradients to ensure that the curvature calculation is suitable for the tunnel transportation environment; After calculating the pressure gradient change rate tensor, the pressure curvature index is further calculated to comprehensively evaluate the potential risk of obstruction of the transmission channel. The pressure curvature index combines the maximum curvature eigenvalue of the pressure time series curve and the cumulative effect of the pressure gradient. The calculation expression is as follows: ， Where: P ci is the pressure curvature index, λ max It represents the principal curvature eigenvalue, that is, the maximum curvature of the pressure time series curve in the local area, and is defined as follows: Where H(P) is the Hessian matrix of the pressure time series, and its eigenvalue λ max ,λ min Reflects the main direction change trend of the pressure surface, and selects the maximum eigenvalue λ max represents the direction of the most drastic local pressure change, β′ is the curvature nonlinear amplification factor, which controls the contribution of the principal curvature to enhance the sensitivity to the mutation area, and Ω is the spatial area within the monitoring window, which means that the pressure gradient change in the entire area is integrated to obtain the global obstruction risk assessment.

10. A tunnel collapse emergency rescue material supply system, used to implement the tunnel collapse emergency rescue material supply method according to any one of claims 1 to 9, characterized in that: It includes rescue pipeline layout module, pressure monitoring installation module, data transmission and monitoring module, pressure data acquisition and preprocessing module, obstruction risk prediction module and dynamic pressure compensation module; The rescue pipe laying module is used to lay rescue small-diameter pipes along the tunnel from the tunnel face to the end of the secondary lining before tunnel construction. Pressure monitoring installation module, high-precision pressure sensors are installed at the rescue end and the face end respectively, which are used to monitor the pressure changes inside the pipeline in real time. The measurement data of all pressure sensors are transmitted back to the monitoring center via wireless data transmission or wired optical fiber for real-time analysis; The data transmission and monitoring module collects pressure data at the face end in real time through the installed pressure sensors, establishes continuous time series data, and pre-processes the collected data; The pressure data collection and preprocessing module selects key data that can reflect the pressure drop from the preprocessed pressure data, and converts the selected data into feature vectors that can be understood and processed by the machine learning model through feature extraction technology; The obstruction risk prediction module inputs the feature vector into a pre-trained machine learning model to predict pipeline obstruction and determine whether there is a risk of obstruction in the transmission channel; The dynamic pressure compensation module automatically switches to emergency control mode when the machine learning model detects that the pipeline transportation channel is blocked. In the emergency control mode, the small pressure compensation device installed at the face end starts immediately, automatically releasing positive pressure air into the pipeline, and dynamically adjusts the amount of positive pressure air released according to the pressure changes predicted by the model.

Images