Shield tunnel sensor health monitoring continuation method

Abstract

The invention relates to the technical field of tunnel engineering, and particularly discloses a shield tunnel sensor health monitoring continuation method, which comprises the steps of collecting and analyzing original data of an old sensor, obtaining theoretical data through structural software simulation calculation, comparing the theoretical data with actual data, and determining final accurate data before the old sensor fails. The method comprises the following steps: establishing and processing an original data set to obtain credible data; using a fitting function to calculate a current most possible predicted value of an old sensor; selecting a new sensor according to the predicted value and a new sensor installation environment; installing the new sensor; and the monitoring continuity is ensured. The method improves the accuracy of the monitoring data of the shield tunnel sensor, optimizes the replacement and correction process of the new and old sensors, is widely suitable for various sensors of the shield tunnel, and guarantees the safety of tunnel engineering.

Description

Technical Field

[0001] This invention relates to the field of tunnel engineering technology, and in particular to a method for continuous health monitoring of sensors in shield tunnels. Background Technology

[0002] Existing technologies for health monitoring of shield tunnel structures primarily employ sensors such as earth pressure gauges, permeability gauges, steel reinforcement stress gauges, and concrete strain gauges. These sensors are pre-embedded or installed in different parts of the tunnel structure to monitor its stress state, deformation, and other relevant parameters. The monitoring data is used to assess the mechanical properties and damage level of the tunnel structure, ensuring traffic safety. The installation locations of tunnel structure health monitoring sensors are mainly divided into two types based on the monitoring location: pre-embedded within the structure during pouring and installed on the surface after concrete pouring. For monitoring equipment installed on the concrete surface, repair or in-situ replacement only requires simple data migration and correction. However, for monitoring equipment pre-embedded within the structure during pouring, direct repair is extremely difficult if it malfunctions, is damaged, or operates abnormally. Furthermore, the sensors have a limited lifespan, and ensuring the continuity and accuracy of monitoring data during sensor replacement is a significant challenge. Summary of the Invention

[0003] The purpose of this invention is to provide a method for continuous health monitoring of shield tunnel sensors, so as to solve the problems of discontinuous monitoring data, low accuracy of monitoring data, and complex sensor replacement process in the prior art.

[0004] To address the aforementioned technical problems, this invention provides a method for continuous health monitoring of sensors in shield tunnels, comprising the following steps:

[0005] S1. Collect and analyze the raw data of the old sensor that needs to be replaced;

[0006] S2. Based on the original data, theoretical data of the old sensor is obtained through simulation calculation using structural software. This data is then compared and analyzed with the actual data of the old sensor in the original data to determine the last accurate data before the old sensor failed. The original data, theoretical data of the old sensor, and last accurate data are used to establish the original dataset. The original dataset is then processed to obtain a reliable original dataset.

[0007] S3. Based on the reliable original dataset, use the fitting function to estimate the most likely current prediction value of the old sensor. ;

[0008] S4. Select a new sensor based on the most likely predicted value and the installation environment of the new sensor;

[0009] S5. Install the new sensor and perform initial data acquisition and correction on the new sensor data;

[0010] S6. Convert the corrected new sensor data into actual monitoring values, and then use this to perform health monitoring and tracking on the new sensor.

[0011] Preferably, in step S2, the raw data includes structural drawings, original monitoring requirements, technical solutions, calculation sheets, and the type, range, accuracy, calibration table, and installation records of the old sensors.

[0012] Preferably, in step S2, the theoretical data of the old sensor obtained by structural software simulation calculation includes establishing a geometric model of the tunnel structure, specifying material properties for each structural unit, defining the boundary conditions of the tunnel structure, performing numerical simulation calculations using finite element calculation software and finite difference calculation software, analyzing the theoretical data of the old sensor at the current stage, and determining the data status based on the actual data of the old sensor.

[0013] Preferably, in step S2, the processing of the original dataset includes denoising, filtering, correction, visualization, stationarity testing, linear pattern recognition, and identification of disordered data.

[0014] Preferably, in step S3, a multinomial regression model is used for fitting and prediction:

[0015] ;

[0016] in, This is the prediction result of a multinomial regression model. It's the frequency of the old sensor. These are the parameters of the multinomial regression model;

[0017] The residuals of the multinomial regression model are used to make prediction corrections using a neural network model, resulting in the predicted correction values. ;

[0018] The current data prediction formula for old sensors is:

[0019] ;

[0020] The original dataset is used to train a model through a model fusion strategy to obtain a trained meta-learner. The most likely prediction is obtained by inputting the data into the trained meta-learner. .

[0021] Preferably, in step S4, the range of the new sensor should be no less than 2-5 times the most likely predicted value.

[0022] Preferably, in step S5, a floodproof box is set around the sensor. The floodproof box body is made of welded steel plates. The rear end of the box body is provided with an outward flange. The outward flange is provided with multiple bolt holes and multiple ribs welded on. The front end of the floodproof box body is provided with a channel steel groove and a floodproof box door. Valves are provided at both the upper and lower ends of the floodproof box body, and the valves are connected to steel pipes.

[0023] Preferably, in step S5, factory data is collected before the new sensor is installed. After the concrete structure at the location where the new sensor will be installed reaches the design strength and its strength stabilizes and increases slowly, initial data from the new sensor will be collected. ;

[0024] The initial data is corrected to obtain the actual data from the new sensor. :

[0025] ;

[0026] In step S6, the corrected new sensor data is used. Converted to actual monitoring values :

[0027] .

[0028] Preferably, the types of old and new sensors are earth pressure gauges, seepage pressure gauges, concrete strain gauges, or steel stress gauges.

[0029] Preferably, in step S5, the new sensor type is an earth pressure gauge. A truncated cone-shaped slot larger than the earth pressure gauge is excavated at the determined tunnel location. The larger end of the slot is placed in the surrounding rock, and the smaller end is placed in the inner arc surface of the tunnel segment. The inner arc surface is larger than the earth pressure gauge. The slot depth penetrates the original grouting layer of the tunnel and can expose the surrounding rock and penetrate into the surrounding rock by 20mm. The earth pressure gauge is placed in the slot with the sensing diaphragm facing the deformation direction of the tunnel soil pressure.

[0030] The new sensor type is a seepage pressure timer, and a 20mm gap should be left between it and the surrounding rock of the tunnel during installation.

[0031] The beneficial effects of this invention are as follows:

[0032] This invention significantly improves the accuracy of monitoring data by precisely identifying the type, range, accuracy, and installation location of old sensors. Using linear and regression analysis, it effectively predicts data before the old sensor malfunctions and further enhances the reliability of monitoring results by predicting current data through a fitting function. When replacing sensors, new sensors are selected based on design conditions and theoretical formulas, and correction formulas are determined by comparing old and new data, ensuring data continuity and accuracy and optimizing the sensor replacement and data correction process. This invention is widely applicable to various sensors pre-embedded in shield tunnel segments, such as earth pressure gauges, permeability gauges, rebar stress gauges, and concrete strain gauges. It can flexibly cope with complex construction environments and changing geological conditions, not only improving the accuracy and reliability of the monitoring system but also playing a crucial role in ensuring traffic safety. It is applicable to multiple fields such as railway traffic management, subway and highway traffic management, providing strong technical support and guarantees for tunnel monitoring and traffic safety. Attached Figure Description

[0033] Figure 1 This is a flowchart illustrating the scenario where the sensor needs to be replaced in an embodiment of the present invention;

[0034] Figure 2 This is a side view of the floodproof tank according to an embodiment of the present invention;

[0035] Figure 3 This is a front view structural diagram of the flood-proof box according to an embodiment of the present invention.

[0036] In the diagram, 101 is the tunnel segment; 102 is the grouting layer; 103 is the surrounding rock; 104 is the operating space; 201 is the flood control tank body; 202 is the outward flange; 203 is the rib plate; 204 is the bolt hole; 205 is the groove; 206 is the tank door; 207 is the rib; 208 is the valve; 209 is the steel pipe; 210 is the rubber strip; and 211 is the expansion bolt. Detailed Implementation

[0037] The embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and should not be construed as limiting the scope of the invention.

[0038] Sensor problems can be mainly classified into the following four types: (1) The sensor malfunctions and cannot collect data; (2) The sensor collects unstable and irregular data, which deviates significantly from the actual data; (3) The sensor range cannot meet the monitoring requirements and the actual data exceeds the sensor range; (4) The sensor collects regularly changing data, but there is a regular deviation from the actual data.

[0039] For case (4), recalibrating or refitting the data will basically meet the subsequent needs; for cases (1), (2), and (3), where the sensor cannot collect data, only has irregular data, or has exceeded the sensor's range, a new sensor needs to be installed to approximately replace the old sensor. The types of old and new sensors include earth pressure gauges, seepage pressure gauges, concrete strain gauges, and steel stress gauges, such as... Figure 1 As shown, this embodiment provides a method for continuous health monitoring of sensors in a shield tunnel, including the following steps:

[0040] S1. Collect and analyze the original data of the old sensors that need to be replaced, including structural drawings, original monitoring requirements, technical solutions, calculation sheets, as well as the type, range, accuracy, calibration table and installation records of the old sensors; check the original calculation sheets to understand how the old sensor data is used for structural safety assessment, and theoretically analyze the stress state of the tunnel, the surrounding rock, and the changes in stress boundary conditions.

[0041] S2. Based on the original data, theoretical data of the old sensor is obtained through structural simulation calculation using structural software. The theoretical data of the old sensor includes establishing a geometric model of the tunnel structure, specifying material properties for each structural unit, such as elastic modulus and Poisson's ratio, defining the boundary conditions of the tunnel structure, including fixed ends and stress points, etc. Numerical simulation calculations are performed using finite element calculation software and finite difference calculation software to analyze the theoretical stress situation and theoretical calculation values ​​at the current location of the old sensor. Based on the actual data of the old sensor, the data status is determined, including data that cannot be collected, data with irregular deviation, data that exceeds the range, and data with regular deviation. The theoretical data of the old sensor is compared and analyzed with the actual data of the old sensor in the original data to determine the last accurate data before the old sensor failed.

[0042] The raw data, theoretical data from old sensors, and the final accurate data are used to create the original dataset. This dataset is then processed, including:

[0043] Denoising is performed using wavelet transform, assuming the original signal is... The denoised signal is The formula for wavelet transform is:

[0044] ;

[0045] in, These are wavelet transform coefficients. It is a wavelet function. It is a scale parameter. It is a translation function. It is the sampling period.

[0046] Using a Butterworth filter for filtering, its transfer function is:

[0047] ;

[0048] in, It is the complex frequency in the Laplace transform. It is the cutoff frequency. It is the order of the filter.

[0049] Correction is achieved by comparing with reference data, assuming... The original data, The baseline data is the corrected data. for:

[0050] ;

[0051] Perform data visualization operations, namely, draw time series plots. By drawing time series plots, we can initially observe whether the data shows a clear trend, observe the trend of the data over time, and the point where the data changes from order to disorder.

[0052] To determine the data characteristics, a stationarity test is performed. Monitoring data for the structural health of shield tunnels should exhibit stable changes under relatively stable external environmental conditions. Time series data can be stationary or non-stationary but exhibit a consistent trend, meeting the prerequisites for the ADF test.

[0053] Use the ADF test to determine whether the data is stationary:

[0054] ;

[0055] in, It is a first-order difference. At a certain point in time Data values, It is a time trend item. It is the lag order. It is a constant term. It is a trend item. It is the autoregressive coefficient. It is the coefficient of the lagged term. It is an error term;

[0056] The lag order in the model needs to be determined. Information criteria are typically used to automatically select the lag order. ,

[0057] Run the ADF test and calculate the ADF test statistic. :

[0058] ;

[0059] in, For autoregressive coefficients The estimated value, yes Standard error;

[0060] The calculated ADF statistic is compared with the corresponding critical value table. If the ADF statistic is less than the critical value at a certain significance level, the null hypothesis is rejected, and the time series is considered to be stationary.

[0061] Linear patterns can be identified and linear models can be constructed. For data with possible first, second, and third linear patterns, different linear models can be constructed to fit the data.

[0062] Linear model:

[0063] ;

[0064] Quadratic linear model:

[0065] ;

[0066] Cubic linear model:

[0067] ;

[0068] in, These are model parameters. It is the dependent variable, stress value. The independent variable is time. It refers to the number of data points;

[0069] Outlier detection is performed using statistical methods. Fractional method for identifying outliers:

[0070] ;

[0071] in, It is the value of the data point. It is the mean of the dataset. It is the standard deviation of the dataset;

[0072] Calculate the mean of the data :

[0073] ;

[0074] in, It is the total number of data points. It is the value of each data point;

[0075] Calculate the standard deviation of the data :

[0076] ;

[0077] when If the value is >, then the data point is considered an outlier, thus determining which data points might be outliers.

[0078] Based on the identification of messy and disordered data, data filtering is performed to obtain a reliable original dataset.

[0079] S3. Based on the reliable original dataset, use the fitting function to estimate the most likely current prediction value of the old sensor. Use a multinomial regression model for fitting and prediction:

[0080] ;

[0081] in, This is the prediction result of a multinomial regression model. It's the frequency of the old sensor. These are the parameters of the multinomial regression model;

[0082] A neural network model is used to predict and correct the residuals of a multinomial regression model. The neural network combines gradient descent and Gauss-Newton methods, and its weight adjustment formula is as follows:

[0083] ;

[0084] in, It is a Jacobian matrix. It is the identity matrix. It's about adjusting parameters. This is the expected output. This is the actual output, yielding the predicted correction value. ;

[0085] The current data prediction formula for old sensors is:

[0086] ;

[0087] The model is trained using the original dataset through a model fusion strategy, which includes blending training and blending testing. The blending training process involves splitting the data into a training set Mtrain, a validation set Mv, and a test set Mtest. Individual learners are trained on the training set and validated on the validation set. The validation results of all individual learners are horizontally concatenated to form a new feature matrix, which is then fed into a meta-learner for training. The blending testing process involves using the trained individual learners and the meta-learner to predict on the test set Mtest. The prediction results of the individual learners are used as new features input into the meta-learner to obtain the final prediction result. Assuming there are N individual learners with prediction results y1, y2, ..., yN, the input features of the meta-learner are X = [y1, y2, ..., yN]. The prediction results of the meta-learner... for:

[0088] ;

[0089] in, It is the prediction function of the meta-learner;

[0090] Obtain the trained meta-learner, and The most likely prediction is obtained by inputting the data into the trained meta-learner. .

[0091] S4. Based on the most probable predicted value and the installation environment of the new sensor, select the new sensor and determine its technology type, range, accuracy, and other performance parameters. The range of the new sensor should not be less than the most probable predicted value. 2-5 times.

[0092] Vibrating wire sensors, due to their high accuracy and stability, are suitable for applications requiring long-term, continuous monitoring. Grating sensors, on the other hand, are suitable for harsh environments or applications requiring large-scale measurements due to their strong anti-interference capabilities and wide measurement range.

[0093] When selecting a technology type, the following factors should also be considered: environmental adaptability, whether the sensor can adapt to environmental conditions such as humidity, temperature, and chemical corrosion inside the tunnel; durability, the durability and long-term stability of the sensor to ensure normal operation throughout the entire life cycle of the shield tunnel; maintenance cost, the cost differences in maintenance and calibration for different types of sensors; and compatibility, whether the new sensor is compatible with existing monitoring systems and data processing software.

[0094] The selection of the measurement range directly affects whether the sensor can accurately capture changes in actual physical quantities. For the structural health monitoring of shield tunnels, the selection of the measurement range needs to consider the following factors:

[0095] (1) Historical data: Analyze historical monitoring data to determine the maximum and minimum values ​​of the physical quantities monitored by the sensor, so as to select a range that can cover these ranges.

[0096] (2) Predict the data and use the fitting function to predict the future data change trend to ensure that the range can cover the predicted maximum value.

[0097] (3) Safety margin: When selecting the range, a certain safety margin should be reserved to deal with possible extreme situations or emergencies.

[0098] (4) Cost-effectiveness: Under the premise of meeting the monitoring needs, select the range with the highest cost-effectiveness.

[0099] Accuracy is a key indicator for evaluating sensor performance, directly impacting the accuracy of monitoring data. When selecting an accuracy level, the following factors need to be considered:

[0100] (1) Monitoring requirements: Determine the required accuracy level based on the specific requirements of shield tunnel structure health monitoring.

[0101] (2) Data comparison: By comparing the data with the data of the sensor before the failure, the required accuracy of the new sensor is determined to ensure the continuity and consistency of the data.

[0102] (3) Error analysis: Analyze the difference between the sensor output signal and the true value, and select the accuracy that can control the error within an acceptable range.

[0103] (4) Industry standards: Refer to the industry standards and recommended values ​​for similar monitoring items, and select the accuracy that meets the industry standards.

[0104] Based on the most likely data and environment after sensor failure, select the appropriate sensor technology type, such as vibrating wire or grating type.

[0105] The measurement range refers to the range of physical quantities that a sensor can measure. For different application scenarios, it is crucial to select the appropriate measurement range. The appropriate measurement range should be selected based on the most likely data and trends after the sensor is damaged.

[0106] The range of the new sensor should be no less than 2-5 times the most likely predicted data at the current stage, in order to ensure that the data can be displayed correctly as the long-term monitoring data value increases.

[0107] Accuracy is an important parameter of sensor performance. It determines the difference between the sensor output signal and the true value. Based on the most likely data after sensor failure and the past fluctuation of the data, an appropriate accuracy is selected.

[0108] Before actual installation, prototype testing of the selected sensors is conducted to verify their performance and adaptability.

[0109] S5. The new sensor type is an earth pressure timer. At the determined tunnel location, excavate a frustum-shaped slot larger than the earth pressure gauge. The larger diameter end of the slot is placed in the surrounding rock, and the smaller diameter end is placed on the inner arc surface of the tunnel segment. The inner arc surface is larger than the earth pressure gauge. The slot depth penetrates the original grouting layer of the tunnel, extending 20mm into the surrounding rock. Place the earth pressure gauge inside the slot, with its sensing diaphragm facing the direction of deformation of the tunnel soil pressure. For seepage pressure timers, a 20mm gap should be maintained between the new sensor and the surrounding rock during installation.

[0110] A flood-proof box is constructed to prevent water leakage during the removal of the tunnel lining concrete and grouting layer. If leakage occurs, the box door is sealed, and grouting is performed to plug the leak. The size and shape of the flood-proof box are matched to the new sensor. The box is installed around the sensor to protect it from environmental damage. Figure 2 , 3 As shown, the flood control tank body 201 is welded from steel plates. The rear end of the tank body has an outward-flared edge 202 with multiple bolt holes 204 and multiple welded ribs 203. The front end of the flood control tank body 201 has a channel steel groove 205 and a flood control tank door 206. Valves 208 are provided at both the upper and lower ends of the flood control tank body 201, and the valves 208 are connected to steel pipes 209. The operating space 104 encompasses the pipe segments 101, the grouting layer 102, and the surrounding rock 103. In this embodiment, the specific structure of the flood control tank is as follows:

[0111] The flood control tank body 201 is reinforced and welded with 5mm steel plate, with a side length of 500mm and a height of 100-150mm. A 100mm outward-facing flange is welded to one end near the pipe segment. Several sets of ribs 203 are welded to the outward-facing flange 202. Multiple bolt holes 204 are provided around the outward-facing flange. Channel steel grooves 205 are welded to both sides of the other end. The flood control tank door 206 is installed and fixed using 5mm steel plate, reinforced with welded ribs 207 on the outside. 50mm holes are opened and welded to the upper and lower end faces of the tank body. The diameter valve 208 is installed with a 100mm long steel pipe 209, which has an external thread at one end connected to the valve 208 and an external thread at the other end. The steel pipe 209 is used to connect to the drainage or grouting pipeline. When the groove leaks, the valve 208 on the lower end is used for drainage and grouting, and the valve 208 on the upper end is used for pressure relief. The flood prevention box is fixed to the inner arc surface of the concrete pipe segment 101 by expansion bolts 211. The steel plate and the inner arc surface of the pipe segment 101 are directly pressed together by water-swellable rubber strips 210.

[0112] Install new sensors:

[0113] The new sensor to be installed is an earth pressure gauge. Based on design requirements and the monitoring purpose of the earth pressure gauge, its installation location was determined. During the later installation of the earth pressure gauge, the original force balance was disrupted, and the newly installed earth pressure sensor could not achieve the same force state as the original sensor. Therefore, it was necessary to use the predictive data from the old sensor for correction. At the determined location, a truncated cone-shaped slot, slightly larger than the earth pressure gauge, was excavated. The larger circular surface was located in the surrounding rock, providing sufficient space for drilling and other tools. The smaller circular surface was located on the inner arc surface of the tunnel segment, slightly larger than the earth pressure gauge. The slot depth penetrated the original grouting layer, exposing the surrounding rock and extending 20mm into it. The slot size was ensured to be large enough to accommodate the earth pressure gauge, with adequate space for installation and fixation. The truncated cone shape of the slot facilitated... For the soil pressure gauge and micro-expansion concrete to bear the load in the later stage, the soil surface at the bottom of the trench needs to be flat, uniform and dense. Place the soil pressure gauge in the trench and ensure that its orientation is correct, with the sensing membrane facing the direction of soil pressure deformation. Use expansion bolts to temporarily fix the soil pressure gauge to the surrounding rock. Use tools to compact the soil around the soil pressure gauge to ensure that the soil pressure gauge is in full contact with the soil surface and bears the load. In order to ensure that the pressure surface of the soil pressure gauge is in full contact with the surrounding rock, grouting pipes are simultaneously buried around the side of the soil pressure gauge. Fill and compact the micro-expansion concrete with the same strength grade and approximate mix ratio. After the micro-expansion concrete reaches the standard strength, use grouting pipes to inject cement grout around the side of the soil pressure gauge to fill the gap between the soil pressure gauge and the surrounding rock, so that the soil pressure gauge is in close contact with the surrounding rock and the concrete.

[0114] The newly installed sensor is a seepage pressure gauge, which measures the water pressure of groundwater outside the pipe segment. After installation, once the groundwater pressure returns to its normal stress state, the seepage pressure can be directly measured. Therefore, unlike the data correction method of the soil pressure sensor, the seepage pressure gauge does not need to be in close contact with the surrounding rock and should have a 20mm gap between it and the surrounding rock to ensure effective transmission of groundwater seepage pressure. It should be filled and compacted with micro-expansion concrete of the same strength grade and approximate mix ratio.

[0115] Earth pressure gauges and permeability gauges are installed on the outer side of the outer arc surface of the tunnel segment to withstand the action of the surrounding rock and groundwater. The location and number of measurement points are determined according to the actual engineering needs and structural characteristics, and should be as similar as possible to the state represented by the sensor whose function needs to be replaced. The corresponding tools and materials are prepared. Since the holes for earth pressure gauges and permeability gauges are relatively deep, earth pressure gauges and permeability gauges should not be installed in the same hole position as concrete strain gauges and steel stress gauges. If secondary grouting is considered to make the earth pressure gauges tightly attached, then permeability gauges should not be installed in the same hole position as earth pressure gauges.

[0116] The new sensors to be installed are concrete strain gauges and steel stress gauges. The installation steps are the same as those for the soil pressure gauges and seepage pressure gauges mentioned above. If the installation location is shallow and there will be no water leakage, the floodproof box does not need to be installed. Drill holes at the determined suitable location, install the new sensors according to the monitoring direction of the old sensors, and fill them with micro-expansion concrete of the same strength grade and approximate mix ratio.

[0117] Extend the observation cable and loosely lay it out to ensure that the cable will not be damaged during subsequent construction. During excavation and backfilling, special care should be taken to protect the earth pressure gauge to avoid damage caused by improper construction. Ensure that the observation cable is not damaged during laying and is not subjected to excessive stretching or compression when backfilling the soil. During the backfilling process, the earth pressure gauge readings should be monitored with a reading device throughout the process. If any abnormality is found, remedial measures should be taken immediately.

[0118] After installing the new sensor, initial data acquisition and correction are performed, collecting factory data prior to the sensor installation. Connect the sensor to the data acquisition unit. Once the new concrete structure reaches the design strength and stabilizes, and the strength increase becomes slow, collect the initial data from the new sensor. ;

[0119] The initial data is corrected to obtain the actual data from the new sensor. :

[0120] ;

[0121] The correction of osmotic pressure gauges differs from that of earth pressure gauges, concrete strain gauges, and steel stress gauges because newly installed osmotic pressure gauges directly receive the effects of water pressure and osmotic pressure in the surrounding rock, while earth pressure gauges, concrete strain gauges, and steel stress gauges are results of secondary stress. The calculation formulas for different types of new sensors are as follows:

[0122] Actual osmotic pressure value of the osmotic pressure gauge :

[0123] ;

[0124] in, These are the initial data read after the new osmotic pressure gauge was installed. This is the factory data before the new osmotic pressure gauge was installed;

[0125] Actual earth pressure value of earth pressure gauge :

[0126] ;

[0127] in, These are the initial data read after the new earth pressure gauge was installed. This is the factory data before the new earth pressure gauge was installed. This is the most likely predictive data for the new earth pressure gauge at this stage, namely the data mentioned above. ;

[0128] Actual concrete stress value of concrete strain gauge :

[0129] ;

[0130] in, These are the initial data read after the new concrete strain gauges were installed. This is the factory data before the installation of the new concrete strain gauge. This is the most likely predictive data for the new concrete strain gauge at this stage, namely the data mentioned above. ;

[0131] Actual steel stress value of steel bar stress gauge :

[0132] ;

[0133] in, These are the initial data read after the new rebar stress gauge was installed. This is the factory data before the installation of the new rebar stress gauge. This is the most likely predictive data for the new steel bar stress gauge at this stage, namely the data mentioned above. ;

[0134] S6. Correct the actual data of the new sensor. Converted to actual monitoring values :

[0135] ;

[0136] Subsequent monitoring values ​​will be based on actual values. Health monitoring and tracking of the new sensors.

[0137] Based on the above-mentioned method for continuous health monitoring of shield tunnel sensors, the following calculation example is provided:

[0138] Suppose we have the following data from a vibrating wire strain sensor:

[0139]

[0140] S1. Data Preprocessing:

[0141] (1) Denoising: Wavelet transform is used for denoising. Assuming the Daubechies wavelet is selected, the scale is 2, and the translation is 1, the denoised signal is assumed to be as follows:

[0142]

[0143] (2) Filtering: A Butterworth filter with a cutoff frequency of 10Hz and an order of 2 is used. The filtered signal is assumed to be as follows:

[0144]

[0145] (3) Correction, assuming the baseline data is:

[0146]

[0147] The corrected data is as follows:

[0148]

[0149] S2, Data Validation and Outlier Detection

[0150] Calculate the mean and standard deviation, calculate the mean and standard deviation of the corrected readings, calculate the Z-score for each corrected reading, identify outliers, and consider the data point as an outlier when |Z| > 3. In this example, the absolute values ​​of all Z-scores are less than 3, so there are no outliers. Data screening: Based on the outlier detection results, all data points are reliable, so all data points are retained.

[0151] S3. Establish a multinomial regression model

[0152] Train a third-order polynomial regression model using the corrected data:

[0153] ;

[0154] Model parameters were obtained by fitting using the least squares method. .

[0155] S4. Establish a neural network model

[0156] Design an LMP network comprising an input layer (sensor frequency), two hidden layers (10 neurons each), and an output layer (predicted concrete strain). Use mean squared error as the loss function and train the network using backpropagation. The trained network is capable of predicting new sensor data based on new frequency data.

[0157] S5. Use model fusion strategy

[0158] Suppose we have a new frequency data point: .

[0159] (1) Polynomial regression prediction:

[0160] ;

[0161] (2) Neural network prediction:

[0162] Assume the concrete strain predicted by the neural network is ;

[0163] (3) Model fusion:

[0164] Assuming a simple average value is used as the fusion strategy:

[0165] ;

[0166] (4) Final prediction data of concrete strain gauge for:

[0167] ;

[0168] Assuming the new concrete strain gauge has a factory-set strain value of 5 με, and the actual concrete strain value is subsequently measured, based on the aforementioned actual concrete strain value... This yields the converted values ​​of the new sensor data, which are then used for subsequent monitoring and tracking.

[0169] The embodiments of the present invention are given for illustrative and descriptive purposes only, and are not intended to be exhaustive or to limit the invention to the forms disclosed. Many modifications and variations will be apparent to those skilled in the art. The embodiments were chosen and described in order to better illustrate the principles and practical application of the invention, and to enable those skilled in the art to understand the invention and to design various embodiments with various modifications suitable for a particular purpose.

Claims

1. A method for continuous health monitoring of sensors in a shield tunnel, characterized in that, Includes the following steps: S1. Collect and analyze the raw data of the old sensor that needs to be replaced; S2. Based on the original data, theoretical data of the old sensor is obtained through simulation calculation using structural software. This data is then compared and analyzed with the actual data of the old sensor in the original data to determine the last accurate data before the old sensor failed. The original data, theoretical data of the old sensor, and last accurate data are used to establish the original dataset. The original dataset is then processed to obtain a reliable original dataset. S3. Based on the reliable original dataset, use the fitting function to estimate the most likely current prediction value of the old sensor. ; S4. Select a new sensor based on the most likely predicted value and the installation environment of the new sensor; S5. Install the new sensor and perform initial data acquisition and correction on the new sensor data; S6. Convert the corrected new sensor data into actual monitoring values, and then use this to perform health monitoring and tracking on the new sensor.

2. The method for continuous health monitoring of shield tunnel sensors according to claim 1, characterized in that, In step S1, the original data includes structural drawings, original monitoring requirements, technical solutions, calculation sheets, as well as the type, range, accuracy, calibration table, and installation records of the old sensors.

3. The method for continuous health monitoring of shield tunnel sensors according to claim 1, characterized in that, In step S2, the theoretical data of the old sensor is obtained by simulation calculation using structural software. This includes establishing a geometric model of the tunnel structure, specifying material properties for each structural unit, defining the boundary conditions of the tunnel structure, performing numerical simulation calculations using finite element calculation software and finite difference calculation software, analyzing the theoretical data of the old sensor at this stage, and determining the data status based on the actual data of the old sensor.

4. The method for continuous health monitoring of shield tunnel sensors according to claim 1, characterized in that, In step S2, the processing of the original dataset includes denoising, filtering, correction, visualization, stationarity testing, linear pattern recognition, and identification of disordered data.

5. The method for continuous health monitoring of shield tunnel sensors according to claim 1, characterized in that, In step S3, a multinomial regression model is used for fitting and prediction: ； in, This is the prediction result of a multinomial regression model. It's the frequency of the old sensor. These are the parameters of the multinomial regression model; The residuals of the multinomial regression model are used to make prediction corrections using a neural network model, resulting in the predicted correction values. ; The current data prediction formula for old sensors is: ； The original dataset is used to train a model through a model fusion strategy to obtain a trained meta-learner. The most likely prediction is obtained by inputting the data into the trained meta-learner. .

6. The method for continuous health monitoring of shield tunnel sensors according to claim 1, characterized in that, In step S4, the range of the new sensor should be no less than 2-5 times the most likely predicted value.

7. The method for continuous health monitoring of shield tunnel sensors according to claim 1, characterized in that, In step S5, a floodproof box is set around the sensor. The floodproof box body (201) is made of welded steel plates. The rear end of the floodproof box body (201) is provided with an outward flange (202). The outward flange (202) is provided with multiple bolt holes (204) and multiple ribs (203) are welded on it. The front end of the floodproof box body (201) is provided with a channel steel groove (205) and a floodproof box door (206). Valves (208) are provided at both the upper and lower ends of the floodproof box body (201). The valves (208) are connected to steel pipes (209).

8. The method for continuous health monitoring of shield tunnel sensors according to claim 1, characterized in that, In step S5, factory data before the new sensor is installed is collected. After the concrete structure at the location where the new sensor will be installed reaches the design strength and its strength stabilizes and increases slowly, initial data from the new sensor will be collected. ; The initial data is corrected to obtain the actual data from the new sensor. : ; In step S6, the corrected new sensor data is used. Converted to actual monitoring values : 。 9. The method for continuous health monitoring of shield tunnel sensors according to claim 1, characterized in that, The old and new sensors are earth pressure gauges, seepage pressure gauges, concrete strain gauges, or steel stress gauges.

10. The method for continuous health monitoring of shield tunnel sensors according to claim 9, characterized in that, In step S5, the new sensor type is an earth pressure gauge. A frustum-shaped slot larger than the earth pressure gauge is excavated at the determined tunnel location. The larger diameter end of the slot is placed in the surrounding rock, and the smaller diameter end is placed in the inner arc surface of the tunnel segment. The inner arc surface is larger than the earth pressure gauge. The slot depth penetrates the original grouting layer of the tunnel and can expose the surrounding rock and penetrate into the surrounding rock by 20mm. The earth pressure gauge is placed in the slot with the sensing diaphragm facing the deformation direction of the tunnel soil pressure. The new sensor type is a seepage pressure timer, and a 20mm gap should be left between it and the surrounding rock of the tunnel during installation.

Images