Detection device for water pressure outside tunnel lining and drainage capacity evaluation method

Abstract

The invention discloses a device for detecting water pressure outside a tunnel lining and a drainage capacity evaluation method. The method comprises the following steps: constructing a multi-source associated basic data set based on original data collected by a portable device; generating a data stability evaluation result by using time sequence stability evaluation and working condition correlation analysis, and screening a credible test section; a self-adaptive multi-working-condition response model is constructed, and a working-condition external water pressure field is reconstructed and designed through weighted regularization inversion; constructing an external water pressure and drainage capacity coupling model, and performing inversion to generate equivalent drainage capacity parameters by using a global optimization algorithm of proxy model acceleration and spatial smooth constraint; and finally, comprehensively outputting a structure safety evaluation suggestion. The real hydrostatic pressure is obtained through the sealed inverted filter structure of the device, the problems that short-term data interference is large, working condition extrapolation is difficult and inversion is discontinuous are solved in combination with a space coupling inversion algorithm, and high-precision reconstruction of external water pressure and whole-line diagnosis of drainage performance are achieved.

Description

Technical Field

[0001] This invention belongs to the field of water conservancy engineering safety monitoring and structural assessment technology, and particularly relates to a device for detecting external water pressure in tunnel lining and a method for assessing drainage capacity. Background Technology

[0002] External water pressure on the lining is a critical load affecting the structural safety of hydraulic tunnels, and the unobstructed flow of the drainage system directly determines the reduction effect of external water pressure. As the project ages, drainage holes are prone to blockage, leading to abnormally high external water pressure, which can then cause lining cracking or even instability. Therefore, utilizing the brief window of water outage maintenance to accurately obtain the deep pore water pressure of the lining using a portable device, and using this information to inversely deduce the pressure distribution under design conditions and evaluate drainage performance, is of significant research value for guiding project operation and maintenance and reinforcement.

[0003] Currently, the detection of external water pressure in tunnels mainly relies on piezometers pre-embedded during construction or piezometers installed during operation. In terms of detection methods, common approaches include directly inserting simple pipes or using existing drainage holes for open-aperture observation. For data analysis, direct statistical methods are frequently employed, i.e., reading the average pressure over the observation period, or performing simple single-point hydraulic calculations based on Darcy's law. In addition, some studies have attempted to use finite element numerical simulation for back analysis, manually adjusting parameters to approximate the observed values.

[0004] However, existing technologies suffer from technical bottlenecks in both device construction and data analysis methods, leading to insufficient reliability of evaluation results.

[0005] The device suffers from inaccurate pressure readings and inconvenient operation: the existing temporary testing device lacks a reliable deep-hole sealing and back-filtering structure, which makes the test hole prone to leakage (measuring dynamic water pressure instead of static water pressure) or silt blockage; and the dial angle is limited when installed on the side wall, making it difficult to read.

[0006] Short-sequence data has poor anti-interference ability: the data sequence during water outage maintenance is short and easily affected by operation. Existing methods lack effective data cleaning and stability evaluation mechanisms and cannot remove transient noise.

[0007] Lack of operating condition extrapolation and spatial coupling mechanism: The measured data only represent the specific low water level operating conditions during the maintenance period, and the existing algorithm is difficult to scientifically extrapolate them to the design flood level operating conditions; moreover, the traditional single-point inversion ignores the spatial continuity of the drainage system, resulting in drastic changes in the inverted drainage parameters along the line, making it difficult to accurately locate regional siltation. Summary of the Invention

[0008] The purpose of this invention is to provide a device for detecting external water pressure in tunnel linings and a method for assessing drainage capacity, in order to solve one of the aforementioned problems in the prior art.

[0009] Technical solution: A portable detection device for external water pressure in tunnel lining, comprising:

[0010] The pressure tapping pipe assembly defines a flow guiding cavity that extends along the axial direction. The pressure tapping pipe assembly is constructed as a rigid pipe body, which is divided along its length into an insertion section that can extend into the tunnel lining and an exposed section that extends out of the tunnel lining. The insertion section has multiple water inlet holes distributed on its pipe wall that communicate with the flow guiding cavity.

[0011] The filter enclosure assembly is circumferentially wrapped and fastened to the outer peripheral wall of the insertion section, and completely covers the plurality of water inlet holes on the insertion section and the end opening of the insertion section;

[0012] A sealing and filling structure is disposed around the outer periphery of the pressure-sensing pipe assembly and located at the junction of the insertion section and the exposed section; the sealing and filling structure fills the annular gap formed between the outer wall of the pressure-sensing pipe assembly and the inner wall of the installation hole of the tunnel lining, forming a water-stopping plug entity that seals the opening of the installation hole;

[0013] The pressure detection component is airtightly connected to the end of the exposed section away from the insertion section and is in fluid communication with the guide cavity.

[0014] Optionally, the reverse filter coating component includes:

[0015] A flexible, permeable covering layer, which has a multi-layered structure, is stacked and covers the outer peripheral wall of the insertion section;

[0016] A spiral fastener is spirally wound around the outer surface of the flexible permeable coating along the axial direction of the insertion section, applying radial preload inward to fix the flexible permeable coating against the pipe wall of the insertion section and sealing the coating boundary of the flexible permeable coating at both ends of the insertion section.

[0017] Optionally, the pressure tap assembly is constructed as a segmented splicing structure, including:

[0018] A horizontally extending pipe section forms the inserted section and part of the exposed section, with its axis extending in the horizontal direction;

[0019] The vertical guide section, which constitutes the end portion of the exposed section, has its axis extending in the vertical direction;

[0020] A right-angle adapter is rigidly connected between the horizontal extension pipe section and the vertical guide pipe section, defining an orthogonal spatial layout in which the horizontal extension pipe section and the vertical guide pipe section are perpendicular to each other.

[0021] The pressure detection component is installed at the end of the vertical guide section, so that the reading panel of the pressure detection component is kept in the vertical plane.

[0022] Optionally, the insertion section is constructed as a rigid PVC perforated pipe, and the insertion section has a predetermined axial length, which is configured such that the depth to which the insertion section extends into the tunnel lining is not less than 80% of the thickness of the tunnel lining;

[0023] The plurality of water inlets are distributed at intervals along the circumferential and axial directions of the insertion section, and the water inlets penetrate the wall of the rigid PVC perforated pipe to connect the flow guide cavity with the reverse filter coating assembly.

[0024] Optionally, the pressure detection component includes:

[0025] A flow control valve having an inlet end and an outlet end, wherein the inlet end is coaxially connected to the end of the vertical guide tube section via a threaded sealing interface;

[0026] A digital pressure sensor is connected to the outlet end of the flow control valve via a threaded sealing interface;

[0027] The flow control valve is constructed as a ball valve and is used to switch the fluid path between the flow guide cavity and the digital pressure sensor between the on and off states.

[0028] Optionally, the sealing and filling structure is constructed as a solid waterstop that is cured in the field;

[0029] The water-stop plug body is formed by filling the annular gap with a fluid sealing medium and then curing it. The sealing medium is selected from quick-setting cement mortar or expanding rubber material. The water-stop plug body has a predetermined sealing length along the axial direction of the mounting hole, and the sealing length is less than the length of the exposed section.

[0030] Optionally, the end of the insertion section away from the exposed section is defined as an axially open end port;

[0031] The flexible permeable coating layer is constructed as a closed bag-like structure or a folded coating structure. The flexible permeable coating layer not only covers the water inlet hole on the pipe wall circumferentially, but also extends and covers the end pipe opening, thereby forming an axial water inlet filtration interface at the end pipe opening.

[0032] Optionally, the right-angle adapter is constructed as an independent 90-degree elbow fitting;

[0033] The horizontal extension pipe section, the 90-degree elbow fitting, and the vertical guide pipe section are respectively connected by threaded screw fitting or socket bonding fitting to form a rigid communication assembly; the outer diameter of the horizontal extension pipe section is smaller than the outer diameter of the switch valve connected to the exposed section, so as to allow the insertion section to be pushed into the depth of the mounting hole without obstruction.

[0034] According to another aspect of this application, a method for condition inversion and equivalent drainage capacity assessment based on pressure test data is also provided, comprising:

[0035] Based on the collected raw data from the pressure test, combined with external operating condition data and the structure of the drainage system, a basic dataset containing a multi-source data association index is constructed.

[0036] Based on the aforementioned basic dataset, the time series stability of the pressure test time series data is evaluated, data stability evaluation results are generated, and reliable test segment datasets are selected.

[0037] Based on the aforementioned reliable measurement segment dataset, a multi-condition response model for external water pressure is constructed, and the external water pressure field data under the design conditions is reconstructed through inversion calculation.

[0038] Based on the external water pressure field data of the design conditions, a coupled model of external water pressure and drainage capacity is constructed in conjunction with the structure of the drainage system, and equivalent drainage capacity parameters are generated by iterative solution.

[0039] Optionally, the step of constructing a multi-condition response model of external water pressure and reconstructing the external water pressure field data under design conditions includes:

[0040] Extract time series samples of each measurement point from the trusted measurement segment dataset, and assign numerical weights to the time series samples of each measurement point according to the stability level and confidence level recorded in the data stability evaluation results, and construct weighted multi-condition training sample data.

[0041] Analyze the correlation between external water pressure and external operating condition variables in the weighted multi-condition training sample data and the sample coverage, adaptively adapt the optimal response model type, and generate response model structure selection data; and

[0042] Based on the response model structure selection data, a weighted regularized objective function is constructed, and the weighted multi-condition training sample data is fitted and solved to identify the external water pressure inversion parameter data that characterizes the relationship between external water pressure and water level and rainfall function. The external water pressure response value under the design condition is then calculated using the external water pressure inversion parameter data.

[0043] Optionally, the steps of adaptively adapting to the optimal response model type and constructing a weighted regularized objective function include:

[0044] The number of reliable samples and the reservoir water level fluctuation in the weighted multi-condition training sample data are evaluated. When the number of reliable samples is lower than a preset threshold or the reservoir water level fluctuation is narrow, a linear model is selected as the response model type; otherwise, a piecewise linear model or a saturated model is selected.

[0045] An external water pressure response function is constructed, incorporating reservoir water level response coefficient, rainfall response coefficient, rainfall lag time, and background head offset. An optimization equation is then established to minimize the weighted sum of squared residuals.

[0046] A regularization term for constraining the fluctuation amplitude of the response coefficient is introduced into the optimization equation. The unknown parameters in the external water pressure response function are solved using the weighted multi-condition training sample data to obtain the initial value data of the external water pressure inversion parameters.

[0047] Optionally, the method further includes a step of robustly iteratively correcting the initial values ​​of the external water pressure inversion parameters:

[0048] The prediction residual of each sample in the weighted multi-condition training sample data is calculated using the initial value data of the external water pressure inversion parameters, and abnormal samples are identified based on the statistical distribution of the residuals.

[0049] The numerical weights of the abnormal samples are dynamically reduced to generate weighted multi-condition training sample correction data.

[0050] Based on the weighted multi-condition training sample correction data, the optimization equation is reconstructed and solved, and the parameter values ​​in the external water pressure response function are updated until the parameter change amplitude meets the convergence condition, generating the final parameter data of external water pressure inversion.

[0051] Optionally, the step of constructing a coupled model of external water pressure and drainage capacity and generating equivalent drainage capacity parameters includes:

[0052] Extract the external water pressure values ​​of each segment from the external water pressure field data of the design working condition, and match them spatially with the corresponding drainage structure parameters in the drainage system structure to construct a pressure-drainage structure coupled dataset.

[0053] Establish an external water pressure-drainage capacity coupled calculation model to describe the hydraulic dependence between external water pressure response, drainage structure parameters, and equivalent drainage capacity;

[0054] Based on the pressure-drainage structure coupling dataset, with the goal of making the output of the external water pressure-drainage capacity coupling calculation model approximate the external water pressure value under the design conditions, the equivalent drainage capacity of each segment is identified by inversion, and equivalent drainage capacity parameters are generated.

[0055] Optionally, the step of inverting and identifying the equivalent drainage capacity of each segment includes:

[0056] The external water pressure-drainage capacity coupled calculation model is discretized or numerically fitted in a preset parameter space to construct external water pressure-drainage capacity proxy model data that can quickly map drainage parameters to external water pressure response.

[0057] The weight index of each segment is extracted from the data stability evaluation results. Combined with the target value of external water pressure under design conditions in the pressure-drainage structure coupling dataset, a local inversion objective function data containing weighted fitting error term and prior parameter constraint term is constructed.

[0058] The weighted fitting error term quantifies the difference between the theoretical pressure value calculated based on the external water pressure-drainage capacity proxy model data and the target external water pressure value under design conditions.

[0059] Optionally, the step of inverting and identifying the equivalent drainage capacity of each segment further includes performing a global optimization process based on spatial smoothness constraints:

[0060] Based on the local inversion objective function data of each segment, a spatial smoothing penalty term is introduced to constrain the difference in equivalent drainage capacity between adjacent segments, and a global inversion objective function covering the target tunnel range is constructed.

[0061] The global inversion objective function is solved using an iterative optimization algorithm. Under the premise of satisfying the spatial continuity constraint, the estimated value of the equivalent drainage capacity of each segment is updated synchronously to generate the equivalent drainage capacity inversion result data.

[0062] Residual analysis and reliability grading are performed on the equivalent drainage capacity inversion results data to generate the equivalent drainage capacity parameters.

[0063] Optionally, the step of evaluating the time-series stability of the pressure test data and generating data stability evaluation results includes:

[0064] Extract the time series of the pressure test readings from the basic dataset, calculate the coefficient of variation and the adjacent time step difference value that characterize the relative fluctuation level, and generate the pressure test time series statistical feature data.

[0065] Correlation analysis is performed on the pressure test readings and the upstream reservoir water level and rainfall sequences in the basic dataset. Multidimensional correlation coefficients are calculated to generate pressure test time series correlation feature data that quantitatively characterizes the sensitivity of external water pressure to external operating conditions.

[0066] Based on the preset stability judgment rules, the time series statistical feature data of the pressure test and the time series correlation feature data of the pressure test are combined to divide the time series data of each test point into different confidence levels and stability categories, and generate data stability evaluation results, which are used to map sample weights in the subsequent inversion model construction.

[0067] Beneficial effects: This invention obtains the true hydrostatic pressure through the sealed reverse filter structure of the device, and combined with the spatial coupling inversion algorithm, it solves the problems of large short-term data interference, difficulty in extrapolating operating conditions and discontinuous inversion, and realizes high-precision reconstruction of external water pressure and full-line diagnosis of drainage performance. Attached Figure Description

[0068] Figure 1 This is a schematic diagram of the structure of this application.

[0069] Figure 2 This is a physical image of the object used in this application.

[0070] Figure 3 This is a flowchart of this application. Detailed Implementation

[0071] Example 1: This example provides a portable detection device for external water pressure in tunnel lining. The device is configured to be installed in a drainage hole or a manually drilled mounting hole in the tunnel lining 7, and is particularly suitable for in-situ detection of external water pressure behind the lining during tunnel water outage maintenance.

[0072] like Figure 1 As shown, the device mainly comprises the following macroscopic structures: pressure tapping pipe assemblies 3 and 6, reverse filter coating assembly 5, sealing filling structure 4, and pressure detection assemblies 1 and 2. Specifically, the structure includes:

[0073] The pressure-sensing tube assembly, which forms the main framework of the device, defines an axially continuous flow-guiding cavity for draining deep water. Physically, this assembly is constructed using a rigid tube, such as a PVC (polyvinyl chloride) water pipe, to ensure sufficient axial stiffness during insertion. Spatially, the pressure-sensing tube assembly is divided into two parts along its length:

[0074] Insertion section 6, also known as the perforated pipe section: This section is constructed as a pipe that can extend deep into the tunnel lining 7. To achieve water collection, the pipe wall of the insertion section has multiple inlet holes (i.e.,...) that connect to the inner cavity of the guide tube. Figure 1 (As shown by the small circles on the inner wall of the central pipe). These inlet holes form the channels through which water flows into the inner cavity of the guide tube.

[0075] Exposed section 3: This section extends beyond the tunnel lining 7 (i.e. Figure 1 (The cave space shown). In this embodiment, the exposed section and the inserted section are integrally connected, transmitting the collected water pressure to the cave space.

[0076] Filter coating components, such as Figure 1 As shown, to prevent silt particles behind the lining from entering the pipe and causing blockage, the reverse filter covering assembly (e.g., Figure 1 The reverse filtration device 5).

[0077] The filter cover assembly 5 is made of filter cloth with good water permeability and strength. The filter cloth is circumferentially wrapped and fastened to the outer surface of the insertion section 6, and its coverage area must completely cover all water inlet holes on the insertion section and the end opening of the insertion section.

[0078] This fully enclosed structure forms a physical filtration barrier around the insertion section 6, ensuring that only water flow can pass through the filter cloth into the guide cavity, while impurities are blocked outside the tube, thus guaranteeing the long-term stability and accuracy of the pressure test.

[0079] Sealed filling structure, corresponding Figure 1 The sealing material 4 is disposed around the outer periphery of the pressure tube assembly, and specifically located in the junction area between the insertion section 6 and the exposed section 3 (i.e., near the inner wall surface of the lining 7).

[0080] The sealing and filling structure 4 fills the annular gap formed between the outer wall of the pressure pipe assembly (specifically the root of the pressure water pipe 3) and the inner wall of the installation hole of the tunnel lining 7.

[0081] The structure is formed by filling and compacting a sealing material (such as rubber, cement mortar, or a special sealing component).

[0082] The sealing filling structure 4 forms a solid water-stop plug at the orifice, blocking the channel for water to leak from the installation hole into the cavity. The design intent is to transform the originally open drainage hole into a closed pressure chamber, so that the pressure gauge connected later can measure the hydrostatic pressure behind the lining, rather than the dynamic water pressure caused by leakage.

[0083] 4. A pressure detection component, wherein the pressure detection component is airtightly connected to the end of the exposed section (pressure water pipe 3) away from the insertion section.

[0084] The component includes a switching valve 2 and a pressure gauge 1 connected in series. The switching valve 2 is installed at the end of the pressure-inlet water pipe 3, and the pressure gauge 1 (preferably a smart digital pressure gauge) is installed downstream of the switching valve 2.

[0085] When the switch valve 2 is opened, the inner cavity of the guide is in fluid communication with the pressure gauge 1, and the deep water pressure acts directly on the sensing element of the pressure gauge 1, thereby displaying the specific water pressure value. The presence of the switch valve 2 also facilitates the replacement of the pressure gauge or the venting of the system without depressurization.

[0086] In summary, this embodiment, through the coordinated operation of the aforementioned components and utilizing existing or newly drilled holes, constructs a complete external water pressure detection system that draws water from deep within the water body, protects it via reverse filtration, maintains pressure through orifice sealing, and ultimately receives readings from terminal instruments.

[0087] Example 2: Based on Example 1, this example further describes in detail the specific structure of the insertion section in the pressure tapping pipe assembly and the assembly details of the reverse filter covering assembly, aiming to enable the device to penetrate deep into the tunnel lining and effectively collect pore water pressure, while preventing pipeline blockage.

[0088] like Figure 1 As shown, the insertion section (i.e., the perforated pipe section 6) is constructed as a rigid PVC perforated pipe, using polyvinyl chloride material with a certain bending strength to ensure that the pipe body will not undergo severe deflection when advancing into the deep hole. The insertion section 6 has a predetermined axial length, which is configured such that the insertion section can extend into the tunnel lining 7 to a depth of not less than 80% of the lining thickness. This depth parameter is designed to avoid stress release zones or unsaturated zones affected by air in the shallow lining area, thereby ensuring that the collected water pressure can accurately reflect the pore water pressure conditions in the deep lining.

[0089] Multiple water inlet holes are distributed circumferentially and axially along the pipe wall of the insertion section 6, forming the primary channel for water flow into the guide cavity. To prevent sediment particles carried by the water flow from clogging these inlet holes or subsequent instrument pipelines, the reverse filter covering assembly 5 adopts a multi-layered permeable geotextile (reverse filter cloth) layered covering structure. The reverse filter covering assembly 5 not only circumferentially wraps the outer peripheral wall of the insertion section 6, but also extends and folds to cover the axial end port of the insertion section 6 away from the exposed section. This fully enclosed covering structure forms a complete filtration interface around and at the end of the insertion section 6, eliminating the bottleneck of sediment entry.

[0090] To ensure that the filter coating assembly 5 does not slip or shift during the process of pushing the pressure tube into the narrow and rough borehole, this embodiment employs a helical fastener (not shown in the figure, but...). Figure 2 In the actual product, the filter cloth is secured by a spiral fastener (which is visible as a wound iron wire). The spiral fastener is spirally wound around the outer surface of the filter covering assembly 5 along the axial direction of the insertion section 6, applying a radial preload inward to tightly abut and fix the filter cloth to the PVC pipe wall. This spiral winding combined with the full-section fixing assembly method effectively resists the frictional resistance of the hole wall during installation, ensuring the integrity of the filter layer after the device is in place.

[0091] Example 3: This example describes the spatial form and connection relationship of the exposed part of the pressure tapping pipe assembly, aiming to solve the problems of reading convenience and installation adaptability when observing water pressure outside the tunnel sidewall.

[0092] The exposed section of the pressure-feeding pipe assembly (pressure-feeding water pipe 3) is not a single straight pipe, but rather constructed as a segmented, L-shaped structure (right-angle type). Specifically, this structure includes a horizontally extending pipe section extending out of the orifice in the horizontal direction, a vertically extending guide pipe section, and a right-angle adapter (such as a 90-degree PVC elbow) rigidly connecting the two. The horizontally extending pipe section is coaxially connected to the insertion section 6, responsible for leading the water flow horizontally out of the lining surface; the right-angle adapter changes the flow direction by 90 degrees, allowing the subsequent vertical guide pipe section to extend vertically upward (or downward).

[0093] This orthogonal spatial layout offers significant ergonomic advantages: it allows the reading panel of pressure gauge 1, installed at the end of the vertical guide pipe section, to remain naturally in the vertical plane. Compared to installing pressure gauges directly at the horizontal pipe end, which results in the dial facing upwards or downwards, the structure of this embodiment allows workers to read the values ​​directly at eye level when inspecting along the tunnel sidewall, greatly improving observation efficiency and reading accuracy.

[0094] In terms of assembly and connection, the aforementioned horizontal extension pipe section, right-angle adapter, and vertical guide pipe section all employ standard threaded connections or adhesive bonding, forming a rigid and sealed connecting assembly. Furthermore, considering the operability of the installation process, the outer diameter of the horizontal extension pipe section (and the entire insertion portion) is designed to be smaller than the outer diameter of the subsequently connected on / off valve 2. This dimensional difference ensures that the upstream piping structure can pass unobstructed through the sealing area or protective cover of the orifice before the valves and instruments are installed. After the position is adjusted and sealed, the larger valve and instrument assemblies can then be externally connected.

[0095] Example 4 describes the specific selection and connection of the pressure detection component and the physical form of the sealing and filling structure to ensure the airtightness and data accuracy of the detection system.

[0096] The pressure detection assembly, located at the end of the entire device, includes a flow control valve (i.e., the on / off valve 2 shown in the figure) and a digital pressure sensor (i.e., the pressure gauge 1 shown in the figure). The inlet of the flow control valve is coaxially connected to the end of the vertical guide tube section of the pressure-conducting water pipe 3 via a threaded interface with sealing PTFE tape. Structurally, the flow control valve is constructed as a manual ball valve with a rotatable handle for switching the fluid path between the guide chamber and the outside environment between a fully open and fully closed state. The digital pressure sensor is further connected to the outlet of the flow control valve via a threaded sealing interface. The use of an intelligent digital pressure gauge not only provides high-precision readings but also facilitates direct data reading in dimly lit tunnel environments.

[0097] Regarding the sealing and filling structure (sealing material 4), it is the core entity that enables the device to transform from drainage to pressure measurement. In actual assembly, this structure is constructed as a field-cured water-stop plug. Specifically, after the pressure-inducing pipe assembly is inserted into place, the construction personnel fill the annular gap between the pressure-inducing water pipe 3 and the inner wall of the installation hole of the tunnel lining 7 with a fluid or plastic sealing medium. Depending on the site conditions, the sealing medium is selected from quick-setting cement mortar or rubber materials with water-swelling properties. After filling, the sealing medium cures or expands, tightly bonding with the pipe wall and the hole wall to form a sealed plug section with a certain length along the axial direction of the installation hole. The presence of this water-stop plug can withstand external water pressure that may be as high as 76.5 kPa or higher behind the lining, ensuring that the water pressure in the perforated pipe section 6 can be transmitted to the pressure gauge 1 without loss, avoiding low readings due to leakage at the orifice.

[0098] Example 5: This example describes the application environment and collaborative working method of the above-mentioned device in a real project (such as the Shenzhen Northern Water Source Project), aiming to demonstrate how the device can achieve high-precision external water pressure detection in conjunction with specific boundary condition control.

[0099] When using this device for testing, specific test boundary conditions need to be constructed to eliminate the interference of adjacent drainage channels on the seepage field. The specific procedure is as follows: after selecting the test hole for which this device is installed, other drainage holes or water seepage holes within a 5m range upstream and downstream of the test hole (a total range of 10m) are temporarily sealed. This regional sealing operation, combined with the device's own sealing and filling structure 4, can effectively restore the hydrostatic pressure field around the test point, thereby ensuring that the test results accurately reflect the external water pressure level behind the lining.

[0100] This device is particularly suitable for tunnel maintenance during periods of water outage. After installation on the lining sidewall, pressure gauge 1 is positioned at a certain vertical height (e.g., 68 cm) from the floor. During testing, continuous monitoring is performed using this device, for example, for at least three consecutive days, with pressure values ​​recorded daily at fixed times, along with the prevailing weather conditions (e.g., rainfall) and the operating water level of a nearby reservoir. For example, in a specific test, the external water pressure at a certain station (e.g., E0+645) measured by this device remained stable between 76.4 kPa and 76.5 kPa over three days. This stable test data verifies the anti-clogging effectiveness of the filter enclosure component 5 and the reliability of the sealing filling structure 4 within the device, indicating that the device can meet the accuracy requirements for safety testing of hydraulic tunnels.

[0101] Example 6 describes a method for inverting external water pressure conditions and assessing drainage capacity in tunnel linings using a portable device. This example describes, from a macroscopic perspective, how to utilize measured data from short-term, unsteady-state water outage maintenance periods, through a multi-level inversion algorithm, to ultimately obtain long-term design parameters and drainage system status assessment conclusions that guide engineering safety. This example comprehensively covers all technical aspects of the process, from data acquisition, preprocessing, pressure field reconstruction, drainage parameter inversion to comprehensive safety assessment.

[0102] Step 101: Based on the collected raw pressure test data, combined with external operating condition data and drainage system structure, construct a basic dataset containing a multi-source data association index. In other words, use a portable pressure testing device to collect raw pressure test data during tunnel water outage maintenance, and combine it with basic information data of measuring points, external operating condition data, and drainage system structure to construct a basic dataset containing a multi-source information index association.

[0103] Specifically, this step involves the acquisition, cleaning, and fusion of multi-source heterogeneous data. First, using a portable pressure testing device, high-frequency sampling is performed at pre-set measuring points during tunnel maintenance shutdowns to acquire raw pressure test data. This raw data is a sequence of pressure readings arranged chronologically, typically including observation timestamps, pressure gauge readings, and equipment status indicators. Simultaneously, external operating condition data is also collected, primarily including upstream reservoir water level time series and rainfall time series near the measuring area, which directly influence the groundwater seepage field. These data usually originate from reservoir scheduling systems and meteorological monitoring stations. Simultaneously, drainage system structure data needs to be collected, describing the engineering structural characteristics near the measuring points, such as the diameter, depth, and spacing of drainage holes, as well as the material and layout of drainage pipes. Furthermore, basic measuring point information data needs to be integrated, including spatial attributes such as the measuring point's mileage, elevation, and cross-sectional type. The process of constructing the basic dataset essentially involves using the measuring point number and timestamp as dual primary keys to uniformly encode and align the data scattered across different systems. For example, the sampling frequency of the pressure test data and the recording frequency of the water level data are unified to the same time step through interpolation methods, ensuring that at any time t, the system can index the corresponding pressure value P(t), water level H(t), and rainfall R(t). This step provides a standardized data foundation for subsequent complex time series analysis and physical inversion.

[0104] In some alternative implementations, to improve data quality, preliminary outlier removal can be performed on the raw stress test data before constructing the base dataset. For example, the Raida criterion (3σ criterion) can be used to identify and remove instantaneous spikes caused by equipment failure or human intervention. For external operating condition data, if data is missing, linear interpolation or a completion algorithm based on historical trends can be used to fill in the missing data.

[0105] Step 102: Based on the basic dataset, perform a time series stability evaluation on the time series data of the pressure test, generate data stability evaluation results, and select reliable test segment datasets.

[0106] Specifically, since the water outage maintenance period is usually short and subject to significant interference from tunnel operations, directly using all the original data for inversion may lead to distorted results. Therefore, this step introduces a data quality evaluation mechanism. Temporal stability evaluation refers to quantifying the fluctuation characteristics and physical correlation of the pressure test data using statistical methods. For example, the coefficient of variation (standard deviation divided by mean) of a measuring point over a period of time is calculated. If the coefficient of variation is too large, it indicates that the data at that measuring point is extremely unstable and may be affected by dynamic water pressure or leakage. Simultaneously, the correlation coefficient between pressure readings and reservoir water levels is calculated. If the two show a significant positive correlation, it indicates that the measuring point can sensitively respond to changes in external water pressure, and the physical meaning of the data is clear. The generated data stability evaluation results include quality labels (such as high reliability, medium reliability, and low reliability) for each measuring point and each time period, along with corresponding quantitative weighting indicators. Selecting reliable measurement segment datasets is based on the above evaluation results. Data segments that meet the preset quality threshold are retained, while data that is severely disturbed or exhibits abnormal physical patterns are removed, thereby ensuring that the input to the subsequent inversion model has a high signal-to-noise ratio.

[0107] In some alternative implementations, stability assessment can also incorporate rainfall lag effects, i.e., the time delay of rainfall on groundwater recharge is considered when calculating correlations. Data deemed to have low stability are not necessarily discarded directly, but can be assigned very low weights in subsequent inversion steps to minimize their impact on model parameters, thereby making full use of the limited observation sample.

[0108] Step 103: Based on the reliable measurement segment dataset, construct a multi-condition response model for external water pressure, and reconstruct the external water pressure field data under the design conditions through inversion calculation.

[0109] Specifically, this step extracts the inherent patterns of external water pressure changes from short-term measured data and extrapolates them to the design conditions. Constructing a multi-condition response model for external water pressure essentially involves establishing a mathematical function describing how external water pressure changes with environmental variables such as reservoir water level and rainfall. For example, a linear or non-linear relationship can be assumed between external water pressure and reservoir water level, and the parameters of this function can be fitted or inverted using sample points in a reliable measurement segment dataset. Once the model parameters are determined, extreme water levels under the design conditions (such as the check flood level) can be substituted into the model to calculate the predicted external water pressure value at that measurement point under the design conditions. Reconstructing the external water pressure field data under the design conditions involves combining the predicted values ​​of each discrete measurement point with their spatial location information and using interpolation or fitting methods to extend them into a continuous pressure distribution curve covering the entire tunnel route. This process solves the problem that measured data only represents the current operating conditions and cannot be directly used for structural safety verification, achieving a leap from measured values ​​to design values.

[0110] In some alternative implementations, the response model can be a multiple regression model that includes a rainfall lag term, or a nonlinear model based on machine learning algorithms such as support vector machines and neural networks. When reconstructing the pressure field, Kriging interpolation or spline function interpolation can be used, with data stability evaluation results used as interpolation weights to ensure that high-confidence measuring points have a greater impact on the surrounding area.

[0111] Step 104: Based on the external water pressure field data of the design conditions, and in conjunction with the drainage system structure, construct a coupled model of external water pressure and drainage capacity, and generate equivalent drainage capacity parameters through iterative solution.

[0112] Specifically, this step aims to quantitatively evaluate the actual performance of the drainage system. The coupled external water pressure and drainage capacity model is a physical model based on the principles of seepage mechanics. It describes the quantitative relationship between the external water pressure field and drainage capacity (usually expressed as the equivalent permeability coefficient or drainage efficiency factor) under given geological conditions and drainage structure parameters (such as orifice diameter and spacing). Since the external water pressure field data under design conditions has been obtained in step 103, and the drainage structure parameters are also known, the only unknown in the model at this point is the equivalent drainage capacity parameter, which represents the overall performance of the drainage system. By constructing an optimization objective function, the theoretical external water pressure calculated by the model is made as close as possible to the design external water pressure reconstructed in step 103, thereby deriving the equivalent drainage capacity parameter for each segment. This parameter reflects the drainage efficiency of the drainage system under the current actual operating conditions. If this value is significantly lower than the design value, it suggests that the drainage orifices may be clogged or malfunctioning.

[0113] In some alternative implementations, to improve inversion efficiency, a response surface model or surrogate model of external water pressure-drainage capacity can be pre-constructed to replace complex finite element seepage calculations. Iterative solution algorithms can employ the Levenberg-Marquardt algorithm (LM algorithm) or particle swarm optimization (PSO) to quickly find the global optimum.

[0114] Furthermore, as a complete extension of this embodiment, based on the aforementioned equivalent drainage capacity parameters and external water pressure field data under design conditions, structural safety assessment and reinforcement recommendation data can also be generated by combining the ultimate structural bearing capacity of the tunnel lining. For example, the external water pressure reduction factor under actual working conditions can be calculated and compared with the design assumption value; according to the degree of pressure exceeding the limit and the degree of drainage capacity attenuation, the tunnel can be divided into different levels of risk areas (such as safe areas, monitoring areas, and treatment areas), and specific dredging, drainage hole addition, or structural reinforcement measures can be proposed for high-risk areas. This step transforms the algorithm output into directly usable engineering decision-making information.

[0115] In other words, according to one aspect of this application, by combining the external water pressure field data of the design working condition, the equivalent drainage capacity parameters, and the data stability evaluation results, a multi-dimensional assessment of the safety status of the tunnel lining structure is conducted to generate structural safety assessment and reinforcement recommendation data.

[0116] Example 7 describes how to generate data stability evaluation results through quantitative calculation of statistical features and correlation features, and transform them into numerical weights in the subsequent inversion model, thereby achieving optimal use of good data and underutilization of inferior data at the algorithm level.

[0117] Step 201: Extract the time series of the pressure test readings from the basic dataset, calculate the coefficient of variation and adjacent time step difference values ​​that characterize the relative fluctuation level, and generate pressure test time series statistical feature data.

[0118] Specifically, for any measurement point i, first extract its pressure test reading sequence P within the observation window T from the basic dataset. _i (t). To quantify the volatility of this sequence, its time average P needs to be calculated. _mean_i and standard deviation σ _P_i Based on this, the coefficient of variation (CV) is calculated. _i It is defined as the ratio of the standard deviation to the mean, i.e., CV. _i =σ _P_i / P _mean_i The coefficient of variation can eliminate the influence of the absolute value of pressure and objectively reflect the relative dispersion of the data. Simultaneously, it calculates the difference between adjacent time steps Δ. _P_i (t)=|P _i (t)-P _i (t-1)|, and calculate its maximum and mean values. The generated timing statistical characteristic data of the pressure test includes the above indicators. In a physical sense, if the coefficient of variation CV _i If the value exceeds the preset threshold (e.g., 0.1), or if there are frequent abnormal jumps between adjacent differences, it usually means that the measuring point has been interfered with by non-seepage factors (such as construction disturbance or poor equipment contact), and the physical representativeness of the data is poor.

[0119] In some alternative implementations, the autocorrelation coefficient of the series can be calculated or a stationarity test (such as the ADF test) can be performed to further evaluate the time-series characteristics of the data from a statistical perspective. For data containing a clear trend, detrending can be performed before calculating volatility indicators.

[0120] Step 202: Perform correlation analysis on the pressure test readings and the upstream reservoir water level and rainfall sequence in the basic dataset, calculate the multidimensional correlation coefficient, and generate pressure test time series correlation feature data that quantitatively characterizes the sensitivity of external water pressure to external operating conditions.

[0121] Specifically, this step aims to verify the physical plausibility of the data. Theoretically, the external water pressure on the lining should increase with the rise of the upstream reservoir water level H(t) and with the increase of rainfall R(t) (possibly with a lag). Therefore, the pressure test reading sequence P is calculated separately. _i The Pearson correlation coefficient ρ between H(t) and the water level sequence H(t) _PH And the Pearson correlation coefficient ρ with the rainfall sequence R(t) _PR For rainfall correlation, multiple lag times τ (e.g., 0 to 72 hours) can be tried, and the lag time with the highest correlation coefficient can be selected as the optimal lag time τ. _best The generated timing correlation feature data of the pressure test recorded ρ _PH ρ _PR and τ _best If ρ _PH A value close to 1 indicates good connectivity between the measuring point and the reservoir water level, suggesting highly reliable data; if ρ _PH If the value is close to 0 or even negative, or if it has no correlation with rainfall, it suggests that the measuring point may be in a hydraulically isolated zone or that the data is abnormal.

[0122] In some alternative implementations, cross-correlation functions can be used to analyze the dynamic correlation characteristics between time series. For regions with significant nonlinear responses, Spearman's rank correlation coefficient can be used instead of Pearson's correlation coefficient.

[0123] Step 203: Based on the preset stability judgment rules, the time series statistical feature data of the pressure test and the time series correlation feature data of the pressure test are combined to divide the time series data of each test point into different confidence levels and stability categories, and generate data stability evaluation results, which are used to map sample weights in the subsequent inversion model construction.

[0124] Specifically, this step integrates the aforementioned statistical and correlation indicators into the final evaluation conclusion. The preset stability determination rule can adopt a scoring system or a decision tree format. For example, the rule could be set as follows: if CV... _i If the value is less than 0.05 and ρ_PH > 0.8, it is considered to have high stability and is assigned a weight w. _i =1.0; if 0.05 <= CV _i If the value is less than 0.15 and ρ_PH > 0.5, it is considered to have moderate stability and is assigned a weight w. _i =0.6; otherwise, it is judged as low stability and assigned a weight w. _i=0.1. The generated data stability evaluation results bind and store these labels and weight values ​​with the measurement point numbers. This weight mapping mechanism is crucial, as it ensures that in the subsequent multi-condition inversion step 103, the model will prioritize fitting high-quality data with small fluctuations and strong regularity, while automatically suppressing the interference of low-quality data on the inversion parameters, thereby significantly improving the robustness of the inversion results.

[0125] In some alternative implementations, the weight w _i It can also be designed as a continuous function, such as w _i =exp(-k*CV _i )*ρ _PH , where k is an adjustment coefficient, thus achieving a more refined weight allocation. Furthermore, measurement points marked as having extremely low stability can be directly removed from the reliable measurement segment dataset and excluded from subsequent calculations.

[0126] Example 8 describes how to adaptively select the model structure based on data characteristics and construct a weighted regularized objective function to solve for the external water pressure response parameters at a single point.

[0127] Step 301: Extract time-series samples of each measurement point from the trusted measurement segment dataset, and assign numerical weights to the time-series samples of each measurement point according to the stability level and trust level recorded in the data stability evaluation results, and construct weighted multi-condition training sample data.

[0128] Specifically, for measurement point i, the time series sample set {(t,P} is read from the reliable measurement segment dataset. _i (t),H(t),R(t))}. Using the weight information generated in Example 2, a weight w is assigned to each sample point t. _i (t). If the overall stability of the measuring point is high, the basic weight of all samples at that measuring point will be high; if the data fluctuates abnormally during certain specific periods (such as during heavy rain), the weight of the samples during that period will be reduced accordingly. The constructed weighted multi-condition training sample data not only includes physical variables, but also carries confidence information reflecting data quality, thus preparing the data for subsequent weighted regression analysis.

[0129] Step 302: Analyze the correlation between external water pressure and external operating condition variables and the sample coverage in the weighted multi-condition training sample data, adaptively adapt to the optimal response model type, and generate response model structure selection data.

[0130] Specifically, to avoid overfitting or underfitting, this step introduces an adaptive model selection mechanism. First, the number of reliable samples, N_sample, and the variation range of the reservoir water level H(t), Δ_H, are counted. If N_sample is less than a preset threshold (e.g., 10 data points) or Δ_H is less than a certain value (e.g., 0.5 meters), it indicates limited data information. In this case, a simple linear model is forcibly selected as the response model type to ensure the stability of the inversion. Conversely, if the sample size is sufficient and the water level variation is large, and the data exhibits obvious nonlinear characteristics (judged by scatter plots or curvature), a piecewise linear model or a saturated model (e.g., exponential or logarithmic function) is selected as the response model type. The generated response model structure selection data records the model ID and the corresponding number of parameters for each measurement point.

[0131] In some alternative implementations, model selection can also be performed automatically based on the Akaike Information Criterion (AIC) or the Bayesian Information Criterion (BIC), that is, by testing multiple models and selecting the model structure with the best score.

[0132] Step 303: Based on the response model structure, select data to construct a weighted regularized objective function, fit and solve the weighted multi-condition training sample data, identify the external water pressure inversion parameter data that characterizes the relationship between external water pressure and water level and rainfall function, and use the external water pressure inversion parameter data to calculate the external water pressure response value under the design conditions.

[0133] Specifically, taking the most commonly used linear model as an example, the external water pressure response function is constructed as follows: Among them, h ext(t) For external water pressure head, a _i Let c be the water level response coefficient. i Let τ be the rainfall response coefficient. _i b is the rainfall lag time. _i Let J be the background head offset. Construct a weighted regularization objective function J. _i Its form is: In the formula, the first term is the weighted sum of squared residuals, ensuring the model's fitting accuracy to high-weight samples; the second term is the regularization term (L2 norm), where λ is the regularization coefficient used to constrain parameter a. _i and c _i There will be no physically unreasonablely large values. J is minimized using the least squares method or gradient descent method. _i Solve for the optimal parameter set (a) i ,c i ,b i ,τ i This refers to the external water pressure inversion parameter data. Finally, the water level H under the design conditions is... design and typical rainfall Rdesign Substituting into the above function, the predicted external water pressure h_ext under the design conditions is calculated. design .

[0134] In some alternative implementations, for piecewise linear models, breakpoint locations can be introduced as additional optimization parameters. For nonlinear models, solving the objective function may require nonlinear optimization algorithms. Numerical example: Suppose that the inversion at a certain measurement point yields a... _i =0.6, b _i =10m, if the design water level H design =100m, ignoring the impact of rainfall, then the external water pressure head h_ext under the design conditions... design =0.6*100+10=70m, which translates to a pressure of approximately 0.7MPa.

[0135] Example 9 describes how, after initially obtaining the external water pressure inversion parameters, abnormal samples are identified through residual diagnosis, and the influence of random disturbances on the inversion results is eliminated by using a dynamic weighted iteration mechanism, thereby obtaining highly robust final parameters.

[0136] Step 401: Calculate the prediction residual of each sample in the weighted multi-condition training sample data using the initial value data of the external water pressure inversion parameters, and identify abnormal samples based on the statistical distribution of the residuals.

[0137] Specifically, in Example 3, the initial values ​​of the external water pressure inversion parameters (e.g., the initial water level response coefficient a) are obtained. _init Initial rainfall response coefficient c _init After that, substitute it into the response function to calculate the predicted external water pressure head h for each time sample t. _pred (t). Then, the predicted value and the measured value h_ are calculated. meas The prediction residual r(t) between (t) and (t) is equal to |h _meas (t)-h _pred (t)|. To identify outliers, this embodiment employs the box plot rule or standard deviation multiple method from statistics. For example, the standard deviation σ_r of all residuals r(t) is calculated. If the residual r(t) of a sample is greater than 3*σ_r, or exceeds the 95th quantile of the residual distribution, then the sample is determined to be an outlier. These outliers may originate from transient water hammer effects or sensor drift that were not identified in the time series stability evaluation (Example 2). If left untreated, these outliers will significantly skew the regression parameters.

[0138] Step 402: Dynamically reduce the numerical weights of the abnormal samples to generate weighted multi-condition training sample correction data.

[0139] Specifically, once an outlier sample is identified, its weight needs to be penalized and reduced. The weight adjustment strategy can employ either a hard thresholding method or a soft thresholding method. In the hard thresholding method, the weight of the outlier sample is directly reset to 0 or a very small value (such as 0.01), meaning that the sample is almost ignored in subsequent calculations. In the soft thresholding method, the Cauchy weight function or the Huber weight function can be used for continuous adjustment. For example, the updated weight w_new(t) = w_old(t) / (1 + (r(t) / k)) 2 ), where k is the tuning constant. In this way, weighted multi-condition training sample correction data is generated, in which samples that deviate too much from the model prediction are given a very low confidence level, while the weights of samples that conform to physical laws are retained.

[0140] Step 403: Based on the weighted multi-condition training sample correction data, reconstruct and solve the optimization equation, update the parameter values ​​in the external water pressure response function until the parameter change amplitude meets the convergence condition, and generate the final parameter data of external water pressure inversion.

[0141] Specifically, the weighted regularization objective function J in Example 3 is reconstructed using the modified weights w_new(t). _i The optimal parameters are then solved again. This process constitutes an iterative closed loop of diagnosis-weighting-solution. In each iteration, as the parameters are updated, the residual distribution changes, triggering a readjustment of the weights. The iterative process continues until the magnitude of the parameter change between two adjacent iterations (e.g., |a|) is reached. _iter_k -a _iter_k-1 The convergence value is less than the preset convergence threshold (e.g., 1e-4), or the maximum number of iterations is reached. For example, in a numerical experiment, the initial inversion water level response coefficient 'a' was 0.85. However, after identifying abnormal data during a period of heavy rainfall and adjusting the weights, 'a' was corrected to 0.72 in the second iteration, and stabilized at 0.71 in the third iteration. This final convergence value of 0.71 is the final parameter data of the external water pressure inversion, which more accurately reflects the seepage response characteristics of the tunnel lining under normal working conditions than the initial value.

[0142] Example 10: How to use the external water pressure inversion results to calculate the equivalent drainage capacity of a tunnel drainage system. This example introduces a surrogate model technique, which solves the problem of excessively long calculation time in traditional finite element inversion, and constructs a local inversion objective function containing prior constraints.

[0143] Step 501: Extract the external water pressure values ​​of each segment from the external water pressure field data of the design working condition, and perform spatial location matching with the corresponding drainage structure parameters in the drainage system structure to construct a pressure-drainage structure coupled dataset.

[0144] Specifically, for each segment or measuring point i of the tunnel, the target pressure value P at that location is first extracted from the external water pressure field data of the design conditions generated in Example 1. target_i Simultaneously, the geometric parameters at this location are extracted from the drainage system structure, specifically including the drain hole diameter d_hole, drain hole spacing s_spacing, drain hole depth L_depth, and drain pipe diameter d. _pipe Wait. The target pressure value P... target_i With a set of structural parameters D at this location _i ={d _hole ,s _spacing ,L _depth The parameters , ..., are bound together to form a pressure-drainage structure coupled dataset. This dataset clarifies the result (observed pressure) and part of the cause (geometric structure) of the inversion problem, and the remaining unknown cause is the equivalent drainage capacity parameter k to be solved. eq .

[0145] Step 502: Discretize or numerically fit the external water pressure-drainage capacity coupled calculation model in a preset parameter space to construct external water pressure-drainage capacity proxy model data that can quickly map drainage parameters to external water pressure response.

[0146] Specifically, the physical relationship between external water pressure and drainage capacity is P=F(D,k eq H) is typically obtained by solving the seepage differential equation governed by Darcy's Law, which is computationally expensive. To meet the need for rapid on-site assessment in engineering projects, this embodiment innovatively constructs a surrogate model. First, within a reasonable parameter range (e.g., k... eq Using values ​​ranging from 1e-7 m / s to 1e-3 m / s and intervals s ranging from 1 m to 4 m, orthogonal experiments or Latin hypercube sampling are designed to generate a set of discrete parameter combinations. These combinations are then batch-calculated using high-precision finite element software (such as Seep / W or COMSOL) to obtain the corresponding theoretical external water pressure values. Subsequently, algorithms such as polynomial response surface methodology, radial basis function (RBF) neural networks, or Gaussian process regression (Kriging) are used to fit the input parameters {D, k...} eq The mapping relationship between H and the output pressure P is used to generate a surrogate model. This surrogate model is essentially a mathematical formula or black-box function P that can be calculated in milliseconds. proxy =F_approx(D,k eq The calculation error (H) is within the allowable range of engineering (e.g., less than 1%).

[0147] In some alternative implementations, the surrogate model can be constructed in segments. For example, three independent sub-surrogate models can be trained for three different structural forms: no drainage holes, single row of drainage holes, and quincunx arrangement of drainage holes. These sub-surrogate models are then automatically invoked based on the actual structural type of the measurement point during use.

[0148] Step 503: Extract the weight index of each segment from the data stability evaluation results, and combine it with the target value of external water pressure under design conditions in the pressure-drainage structure coupling dataset to construct a local inversion objective function data containing weighted fitting error terms and prior parameter constraint terms.

[0149] Specifically, for each segment i, construct a system with equivalent drainage capacity k. eq_i J is the local objective function of the independent variable. loc_i (k eq_i The function consists of two parts: the first part is the fitting error term, i.e., w. _i *(P proxy (D _i ,k eq_i ,H_des)-P target_i ) 2 Among them, P proxy The theoretical pressure, P, is calculated using a proxy model. target_i It is the reconfiguration pressure under design conditions, w _i It is derived from the weights of the data stability evaluation results. If the data stability of a certain measurement point is high, then w _i A larger value forces the model to fit that point precisely; conversely, a smaller value forces the model to fit that point precisely. _i The second part is the prior constraint term (regularization term), i.e., λ_k*(k eq_i -k _prior_i ) 2 Where k _prior_i This is a reference value for the permeability coefficient based on design data or geological surveys, where λ_k is the constraint coefficient. This term serves to prevent the inverse calculation of k when observational data is insufficient. eq_i Deviates from common sense in physics (e.g., negative values ​​or abnormally large values ​​appear).

[0150] Example 11: To address the issue that single-point inversion may lead to abrupt changes and discontinuities in the results of adjacent segments, a spatial smoothing constraint mechanism is introduced, and a global optimization algorithm is used to achieve joint inversion of the entire drainage capacity.

[0151] Step 601: Based on the local inversion objective function data of each segment, a spatial smoothing penalty term is introduced to constrain the difference in equivalent drainage capacity between adjacent segments, and a global inversion objective function covering the target tunnel range is constructed.

[0152] Specifically, geological conditions and the aging of drainage facilities are generally spatially continuous and do not undergo drastic changes over short distances. Based on this physical assumption, this embodiment applies all segmented local objective functions J... loc_i Accumulate, and add an additional spatial smoothing penalty term J. _smooth Global inversion objective function J global It can be represented as: J global =Sum(J loc_i )+μ*Sum((k eq_i -k eq_j ) 2 The second term is the summation of all adjacent segment pairs (i,j), and (k... eq_i -k eq_j ) 2 Differences in drainage capacity parameters between adjacent segments are penalized, with μ serving as a smoothing coefficient. When μ is large, the inversion results tend to be spatially flat, suppressing parameter oscillations caused by single-point measurement errors; when μ is small, the model allows for larger abrupt changes in local parameters to capture local geological defects.

[0153] In some alternative implementations, the spatial smoothing penalty term can be in the form of a second-order difference (i.e., the Laplace operator) to constrain the curvature of the parameter spatial distribution, making the inverted permeability coefficient curve smoother and more natural.

[0154] Step 602: Solve the global inversion objective function using an iterative optimization algorithm, and synchronously update the estimated equivalent drainage capacity of each segment under the premise of satisfying the spatial continuity constraint, thereby generating the equivalent drainage capacity inversion result data.

[0155] Specifically, since the global objective function is coupled with the parameters of all segments, it cannot be solved point-by-point as in local inversion; a joint solution is necessary. This embodiment employs a large-scale nonlinear optimization algorithm, such as the conjugate gradient method, the quasi-Newton method (BFGS), or the trust region algorithm. During the iteration process, the algorithm simultaneously adjusts k for all segments. eq The goal is to find an optimal solution vector that minimizes the pressure fitting error at each point while maintaining overall spatial continuity. Thanks to the surrogate model introduced in Example 5, even with joint inversion involving hundreds of parameters, the calculation can be completed within minutes. The final equivalent drainage capacity inversion result is a parameter sequence distributed along the tunnel mileage, clearly demonstrating the distribution of drainage performance along the entire line.

[0156] Example 12 describes how to use the pressure field data and drainage parameter data generated in the aforementioned examples to output specific structural safety assessment conclusions and reinforcement recommendations.

[0157] Step 701: Calculate the external water pressure reduction factor based on the equivalent drainage capacity parameters, and conduct a multi-dimensional safety assessment in conjunction with the external water pressure field data under the design conditions.

[0158] Specifically, for each segment, the equivalent drainage capacity k obtained through inversion is used. eq Calculate the external water pressure reduction factor η under actual working conditions. real This coefficient reflects the reduction effect of the drainage system on the water pressure in the surrounding rock. Let η real The design reduction factor η specified in the design documents design Compare them. If η real Significantly less than η design This indicates that the actual effectiveness of the drainage system is better than the design expectations; conversely, if η real Greater than η design This indicates insufficient drainage capacity. Simultaneously, the external water pressure P under the reconstructed design conditions will be... target With respect to the design bearing capacity P of the lining structure limit Compare and calculate the safety margin factor S=P limit / P target .

[0159] Step 702: Based on the external water pressure field data of the design working condition, the equivalent drainage capacity parameters, and the data stability evaluation results, a multi-dimensional assessment of the safety status of the tunnel lining structure is conducted to generate structural safety assessment and reinforcement recommendation data.

[0160] Specifically, this step constructs a comprehensive diagnostic logic matrix. First, based on the safety margin factor S, the structural stress state is divided into three levels: safe, warning, and dangerous. Second, based on the equivalent drainage capacity k... eq The degree of attenuation categorizes the drainage system status into three levels: unobstructed, slightly clogged, and severely clogged. Finally, based on the data stability evaluation results (reliability), the system outputs final reinforcement recommendations. For example, for areas with a high degree of risk, severe clog, and high reliability, the system automatically generates recommendations for immediate high-pressure cleaning of drainage holes and aeration of drainage holes; for areas with a high degree of alert, unobstructed, and low reliability, the system recommends increasing monitoring frequency and verifying sensor status. The final structural safety assessment and reinforcement recommendation data can be presented in the form of charts, reports, or Geographic Information System (GIS) layers, directly assisting engineering management departments in developing maintenance plans.

[0161] In summary, this invention, through a series of closely integrated data processing and inversion algorithms, successfully achieves a leap from short-term measurement of a single point to long-term safety assessment of a line, providing scientific and quantitative decision support for the operation and maintenance management of hydraulic tunnels.

[0162] The embodiment employs a portable device comprising a perforated tube filter casing assembly and an orifice sealing and filling structure. The filter assembly effectively prevents siltation in deep holes, while the orifice sealing structure converts the open drainage hole into a closed pressure measuring chamber, ensuring that the collected data represents the true pore hydrostatic pressure; simultaneously, the L-shaped pressure guiding structure solves the problem of sidewall readings.

[0163] In summary, the embodiments construct a time-series stability evaluation model, quantify the physical reliability of the data using the coefficient of variation and multidimensional correlation analysis, and map it to numerical weights in the inversion model. This mechanism can automatically identify and suppress transient noise and non-seepage interference, ensuring that subsequent physical inversion is based on high-quality data.

[0164] Secondly, an adaptive model structure selection and robust iterative correction algorithm were adopted. By automatically matching linear or nonlinear models according to the sample size and operating condition variation, and introducing regularization constraints, a robust response relationship between external water pressure, water level, and rainfall was successfully established, realizing the scientific extrapolation from short-term measured operating conditions to long-term design operating conditions.

[0165] Finally, a surrogate model technique was introduced to replace time-consuming finite element calculations, significantly improving inversion efficiency. By constructing a global optimization objective function that includes a spatial smoothing penalty term, isolated measurement point inversions were transformed into joint inversions along the entire tunnel. This method utilizes spatial continuity constraints to correct single-point errors, obtaining equivalent drainage capacity parameters that smoothly vary along the tunnel mileage. This enables precise location of regional drainage system blockages or failures, providing a quantitative basis for engineering reinforcement decisions.

Claims

1. A device for detecting external water pressure in tunnel lining, characterized in that, include: The pressure tapping pipe assembly defines a flow guiding cavity that extends along the axial direction. The pressure tapping pipe assembly is constructed as a rigid pipe body, which is divided along its length into an insertion section that can extend into the tunnel lining and an exposed section that extends out of the tunnel lining. The insertion section has multiple water inlet holes distributed on its pipe wall that communicate with the flow guiding cavity. The filter enclosure assembly is circumferentially wrapped and fastened to the outer peripheral wall of the insertion section, and completely covers the plurality of water inlet holes on the insertion section and the end opening of the insertion section; A sealing and filling structure is disposed around the outer periphery of the pressure-sensing pipe assembly and located at the junction of the insertion section and the exposed section; the sealing and filling structure fills the annular gap formed between the outer wall of the pressure-sensing pipe assembly and the inner wall of the installation hole of the tunnel lining, forming a water-stopping plug entity that seals the opening of the installation hole; The pressure detection component is airtightly connected to the end of the exposed section away from the insertion section and is in fluid communication with the guide cavity.

2. The apparatus according to claim 1, characterized in that, The filter coating component includes: A flexible, permeable covering layer, which has a multi-layered structure, is stacked and covers the outer peripheral wall of the insertion section; A spiral fastener is spirally wound around the outer surface of the flexible permeable coating along the axial direction of the insertion section, applying radial preload inward to fix the flexible permeable coating against the pipe wall of the insertion section and sealing the coating boundary of the flexible permeable coating at both ends of the insertion section.

3. The apparatus according to claim 1, characterized in that, The pressure-feeding tube assembly is constructed as a segmented splicing structure, including: A horizontally extending pipe section forms the inserted section and part of the exposed section, with its axis extending in the horizontal direction; The vertical guide section, which constitutes the end portion of the exposed section, has its axis extending in the vertical direction; A right-angle adapter is rigidly connected between the horizontal extension pipe section and the vertical guide pipe section, defining an orthogonal spatial layout in which the horizontal extension pipe section and the vertical guide pipe section are perpendicular to each other. The pressure detection component is installed at the end of the vertical guide section, so that the reading panel of the pressure detection component is kept in the vertical plane.

4. The apparatus according to claim 1, characterized in that, The insertion section is constructed as a rigid PVC perforated pipe, and the insertion section has a predetermined axial length, which is configured such that the depth to which the insertion section extends into the tunnel lining is not less than 80% of the thickness of the tunnel lining; The plurality of water inlets are distributed at intervals along the circumferential and axial directions of the insertion section, and the water inlets penetrate the wall of the rigid PVC perforated pipe to connect the flow guide cavity with the reverse filter coating assembly.

5. The apparatus according to claim 3, characterized in that, The pressure detection component includes: A flow control valve has an inlet end and an outlet end, the inlet end being coaxially connected to the end of the vertical guide tube section via a threaded sealing interface; and a digital pressure sensor connected to the outlet end of the flow control valve via a threaded sealing interface; wherein the flow control valve is constructed as a ball valve structure for switching the fluid path between the guide cavity and the digital pressure sensor between an on state and a off state.

6. The apparatus according to claim 1, characterized in that, The sealing and filling structure is constructed as a solid water-stop plug that is cured on-site; The water-stop plug is formed by filling the annular gap with a fluid sealing medium and then curing it. The sealing medium is selected from quick-setting cement mortar or expanding rubber material. The water-stop plug body has a predetermined sealing length along the axial direction of the mounting hole, and the sealing length is less than the length of the exposed section.

7. A method for condition inversion and equivalent drainage capacity assessment based on pressure test data, characterized in that, The methods include: Based on the collected raw data from the pressure test, combined with external operating condition data and the structure of the drainage system, a basic dataset containing a multi-source data association index is constructed. Based on the basic dataset, the time series stability of the pressure test time series data is evaluated, and the data stability evaluation results containing stability level and reliability level are generated, and the reliable test segment dataset is selected. Based on a reliable measurement segment dataset, a multi-condition response model for external water pressure is constructed, and the external water pressure field data under the design conditions is reconstructed through inversion calculation. Based on the external water pressure field data under design conditions, a coupled model of external water pressure and drainage capacity is constructed in conjunction with the structure of the drainage system. The equivalent drainage capacity parameters are generated through iterative solution.

8. The method according to claim 7, characterized in that, The steps for constructing a multi-condition response model of external water pressure and reconstructing the external water pressure field data under design conditions include: Extract time series samples of each measurement point from the reliable measurement segment dataset, and assign numerical weights to the time series samples of each measurement point according to the stability level and reliability level to construct weighted multi-condition training sample data. The correlation between external water pressure and external operating condition variables and the sample coverage in the weighted multi-condition training sample data are analyzed. The optimal response model type is adaptively adapted, and response model structure selection data is generated. Based on the response model structure selection data, a weighted regularized objective function is constructed. The weighted multi-condition training sample data is fitted and solved to identify the external water pressure inversion parameter data that characterizes the relationship between external water pressure and water level and rainfall function. The external water pressure response value under the design condition is then calculated using the external water pressure inversion parameter data.

9. The method according to claim 7, characterized in that, The steps for constructing a coupled model of external water pressure and drainage capacity and generating equivalent drainage capacity parameters include: Extract the external water pressure values ​​of each segment from the external water pressure field data under design conditions, and match them spatially with the corresponding drainage structure parameters in the drainage system structure to construct a pressure-drainage structure coupled dataset. Establish an external water pressure-drainage capacity coupled calculation model to describe the hydraulic dependence between external water pressure response, drainage structure parameters, and equivalent drainage capacity; Based on the pressure-drainage structure coupling dataset, with the goal of making the output of the external water pressure-drainage capacity coupling calculation model approximate the external water pressure value under the design conditions, the equivalent drainage capacity of each segment is identified by inversion, and equivalent drainage capacity parameters are generated.

10. The method according to claim 7, characterized in that, The steps for evaluating the time-series stability of the time-series data from the stress test and generating the data stability evaluation results include: Extract the time series of pressure test readings from the basic dataset, calculate the coefficient of variation and adjacent time step difference values ​​that characterize the relative fluctuation level, and generate pressure test time series statistical feature data. Correlation analysis was performed on the pressure test readings and the upstream reservoir water level and rainfall sequences in the basic dataset. Multidimensional correlation coefficients were calculated to generate time-series correlation feature data of pressure test that quantitatively characterizes the sensitivity of external water pressure to external operating conditions. Based on the preset stability judgment rules, the time series data of each test point are divided into different confidence levels and stability categories by combining the statistical feature data of the pressure test time series and the correlation feature data of the pressure test time series, and the data stability evaluation results are generated.

Images