Real-time detection system for shield tail sealing grease cavity

Abstract

The invention discloses a shield tail sealing grease cavity real-time detection system, and belongs to the technical field of shield construction monitoring, the shield tail sealing grease cavity real-time detection system comprises a multi-physics field monitoring module and an intelligent early warning analysis module, the multi-physics field monitoring module comprises a pressure difference monitoring unit and a moisture monitoring unit, the intelligent early warning analysis module comprises an edge data processing unit and a risk assessment unit; a pressure difference monitoring unit collects grease pressure data in a grease cavity and external environment water pressure data, a pressure difference dynamic model is established, a pressure difference space-time distribution map is generated, and a high-risk leakage area is marked through anomaly recognition and pressure calculation; the moisture monitoring unit combines the constructed moisture diffusion model and the three-dimensional coordinate system to generate a moisture invasion path map, and a potential leakage area is marked; and the edge data processing unit extracts dynamic characteristics, the risk assessment unit outputs a leakage risk level through Bayesian network fusion analysis, and triggers secondary early warning when triple abnormality is detected, so that real-time accurate monitoring and early warning of grease cavity leakage are realized.

Description

Technical Field

[0001] This invention belongs to the field of shield tunneling construction monitoring technology, specifically a real-time detection system for the sealing grease cavity at the tail of the shield. Background Technology

[0002] During tunnel boring machine (TBM) construction, the sealing performance of the tail seal grease chamber is crucial, directly impacting the safe and efficient operation of the TBM. Current technologies present several unresolved issues. For example, single-point pressure detection cannot accurately pinpoint leaks. When a problem occurs in the tail seal, it's difficult to quickly identify the specific leak point, delaying repairs. Similarly, moisture detection faces location difficulties, hindering timely detection and handling of seal failures caused by moisture intrusion. Furthermore, multi-parameter independent early warning modes are prone to false alarms. Since each parameter is judged individually, various factors interfere with each other in the complex TBM construction environment, making it easy to generate erroneous alarm signals based solely on a single parameter's threshold. This reduces the trust of construction personnel in early warning information, affecting timely handling of actual faults. In addition, the transmission of massive amounts of monitoring data to the cloud for processing suffers from high latency. TBM construction is a dynamic and rapid process with extremely high real-time requirements. The latency of cloud processing cannot meet the needs of real-time monitoring of the tail seal status, making it difficult to provide accurate information for construction decisions and potentially leading to serious engineering accidents. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention proposes a real-time detection system for the grease cavity of a shield tail seal, comprising a multi-physics field monitoring module and an intelligent early warning analysis module. The multi-physics field monitoring module includes a pressure difference monitoring unit and a moisture monitoring unit, while the intelligent early warning analysis module includes an edge data processing unit and a risk assessment unit. The pressure difference monitoring unit collects grease pressure data within the grease cavity and external environmental water pressure data, establishes a dynamic pressure difference model, and generates a spatiotemporal distribution map of the pressure difference. After anomaly identification and pressure calculation, high-risk leakage areas are marked. The moisture monitoring unit combines the constructed moisture diffusion model and a three-dimensional coordinate system to generate a moisture intrusion path map and marks potential leakage areas. The edge data processing unit extracts dynamic features, and the risk assessment unit outputs the leakage risk level through Bayesian network fusion analysis. When three anomalies are detected, a secondary early warning is triggered, achieving real-time and accurate monitoring and early warning of grease cavity leakage.

[0004] To achieve the above objectives, the present invention provides the following technical solution: A real-time detection system for the sealing grease cavity of a shield tail includes: a multi-physics field monitoring module and an intelligent early warning analysis module; the multi-physics field monitoring module includes a pressure difference monitoring unit and a moisture monitoring unit; the intelligent early warning analysis module includes an edge data processing unit and a risk assessment unit. The pressure difference monitoring unit collects oil pressure data in the oil cavity and water pressure data in the external environment, establishes a dynamic model of pressure difference and generates a spatiotemporal distribution map of pressure difference. Then, it identifies pressure imbalance points based on the circumferential center of the oil cavity, calculates the pressure difference between the pressure imbalance point and the circumferentially symmetrical position point, generates an abnormal distribution map of pressure difference, and locates abnormal deformations based on fluid mechanics principles. Based on a preset threshold, it marks high-risk leakage areas and outputs high-risk leakage area data and corresponding pressure monitoring signals. The moisture monitoring unit collects moisture data at the bottom of the oil chamber, constructs a moisture diffusion model using a support vector machine algorithm, and outputs a three-dimensional moisture distribution map. Then, a three-dimensional coordinate system is established with the center of the bottom of the oil chamber as the origin, the diffusion rate gradient of each monitoring point is calculated, and a moisture intrusion path map is generated. Based on the moisture intrusion path map, potential leakage areas are marked according to a preset moisture rise rate threshold, and potential leakage area data and corresponding moisture monitoring signals are output. Based on data from high-risk leakage areas, pressure monitoring signals, potential leakage areas, and moisture monitoring signals, the edge data processing unit uses the isolated forest algorithm to extract dynamic features; these dynamic features include the rate of change of pressure difference and the rate of moisture rise. The risk assessment unit outputs the leakage risk level based on the extracted dynamic features and model parameters trained by combining historical fault data through Bayesian network fusion analysis. When triple evidence of a sudden drop in grease viscosity, a sudden increase in moisture, and an excessive pressure difference is detected during the fusion analysis, a level-two warning is triggered.

[0005] Specifically, the pressure difference monitoring unit includes two types of sensors; The first type of sensor is a pressure sensor array that is uniformly arranged in at least two rings on the inner wall of the grease chamber. The ring closest to the tail brush is the first pressure array, which is used to collect grease pressure data in the grease chamber. The other ring is the second pressure array, which is used to assist in verifying the uniformity of pressure distribution. The second type of sensor is an external water pressure sensor group embedded on the outside of the shield tail shell, corresponding to the grease chamber, used to collect external environmental water pressure data; Both the first type of sensor and the second type of sensor are connected to a pressure data acquisition device; the pressure data acquisition device is used to convert the pressure signals collected by the sensors into digital pressure signals and to synchronize the uploading of intracavity pressure and external water pressure data.

[0006] Specifically, the process of constructing a dynamic model of the pressure difference and generating a spatiotemporal distribution map of the pressure difference is as follows: The pressure sensor array of the pressure differential monitoring unit and the external water pressure sensor group synchronously collect data to generate grease pressure data in the grease chamber and external environmental water pressure data, respectively, and transmit them to the pressure data acquisition instrument. The pressure data acquisition instrument timestamps the oil pressure data in the oil chamber and the external environmental water pressure data, and then converts them into digital pressure signals before transmitting them to the pressure difference analysis processor. The pressure difference analysis processor receives the synchronized digital pressure signal through the data interface, analyzes the measured grease pressure value of each pressure sensor node in the cavity and the corresponding external measured water pressure value, and calculates the measured pressure difference at each spatial location; the measured pressure difference is the difference between the external measured water pressure value and the measured grease pressure value. Based on the fluid dynamics pressure transmission theory, the measured pressure difference data of time series is used as input parameters. The measured pressure difference data of multiple consecutive time moments are associated with the corresponding spatial location information. The time series analysis algorithm is used to fit the dynamic change trend of the measured pressure difference, establish a dynamic pressure difference model, and output the theoretical pressure difference value of each point in the grease cavity at different time moments. Meanwhile, the measured pressure difference of each node is compared one by one with the theoretical pressure difference value of the corresponding position under the same real-time pressure data calculated by the pressure difference dynamic model to obtain the deviation between the measured and theoretical pressure difference of each node. Using the spatial coordinates of all sensor nodes and the corresponding measured and theoretical pressure difference deviation data as input, a three-dimensional interpolation algorithm is used for data fusion and spatial estimation to deduce the pressure difference deviation at all locations on the inner wall of the grease cavity. Combined with the changing trend in the time dimension, the spatial distribution of the pressure difference deviation is combined with the dynamic changes in time to transform it into a pressure difference spatiotemporal distribution map. The pressure difference spatiotemporal distribution map reflects the abnormal distribution and evolution trend of pressure difference in the entire grease cavity at different times.

[0007] Specifically, the process of identifying pressure imbalance points based on the circumferential center of the grease cavity, calculating the pressure difference between the pressure imbalance point and a circumferentially symmetrical point, and generating a pressure difference anomalous distribution map includes: The pressure difference analysis processor uses the circumferential center of the grease chamber as a reference to perform a 360-degree circumferential scan of the pressure difference spatiotemporal distribution map, and obtains the position information of all sensor nodes in the pressure difference spatiotemporal distribution map and the deviation between the measured and theoretical pressure difference. If the absolute error between the measured and theoretical pressure difference of any sensor node is greater than the first preset deviation threshold, then the node is marked as a primary pressure imbalance point. For the marked primary pressure imbalance point, the pressure difference analysis processor determines its symmetrical position point about the circumferential center of the grease cavity, obtains the measured pressure difference at the symmetrical position point, and then calculates the absolute error between the measured pressure difference between the primary pressure imbalance point and the symmetrical position point to obtain the measured symmetrical pressure difference value. If the measured symmetrical pressure difference is greater than the second preset symmetrical deviation threshold, then the primary pressure imbalance point is confirmed as the final pressure imbalance point. The pressure difference analysis processor collects all final pressure imbalance points and their corresponding measured symmetrical pressure difference values, calculates the pressure difference anomalous coefficient for each final pressure imbalance point, and generates the pressure difference anomalous distribution map using the contour line method.

[0008] Specifically, the method of locating pressure-weak areas using fluid dynamics principles and marking high-risk leakage areas based on preset thresholds includes: The pressure differential analysis processor, based on fluid mechanics principles and combined with the generated pressure differential distribution map, analyzes the correlation between the pressure differential coefficient and fluid permeation risk, and locates the pressure-weak areas in the grease cavity; the pressure-weak areas are defined as continuous regions with a pressure differential coefficient greater than 1.2, that is, regions where the external water pressure is significantly higher than the grease pressure inside the cavity; The identified pressure-weak areas are analyzed, and areas with pressure difference coefficient values ​​greater than a preset pressure difference threshold in the pressure difference distribution map, or areas with measured symmetrical pressure difference values ​​greater than a preset pressure difference threshold, are marked as high-risk leakage areas. High-risk leakage area data, including location, area, maximum pressure difference coefficient value, and maximum measured symmetrical pressure difference value, is output. Simultaneously, the real-time digital pressure signal corresponding to the triggering of the high-risk leakage area marking is extracted and output as a pressure monitoring signal. The preset pressure difference threshold is determined based on the anti-permeability grade of the shield tail grease.

[0009] Specifically, the moisture monitoring unit includes multiple high-frequency capacitive moisture sensors deployed in the sedimentation zone at the bottom of the oil chamber; the high-frequency capacitive moisture sensors are connected to a data acquisition unit, which is used to receive and upload the digital signals of moisture content collected by the sensors.

[0010] Specifically, the moisture monitoring unit collects moisture data from the bottom of the oil chamber, constructs a moisture diffusion model using a support vector machine algorithm, and outputs a three-dimensional moisture distribution map, including: The sensors in the moisture monitoring unit collect moisture content data from multiple monitoring points at the bottom of the oil chamber, convert it into a digital signal of moisture content, and then transmit it to the data acquisition unit. The data acquisition unit transmits the digital signal of moisture content to the moisture analysis processor; The moisture analysis processor uses the received digital signal of moisture content as the training sample, takes the coordinates of each monitoring point in the three-dimensional coordinate system established with the center of the bottom of the oil cavity as the origin as the input feature, takes the moisture content value of the corresponding monitoring point as the output label, and uses the support vector machine regression algorithm for training and learning to construct a moisture diffusion model that predicts the moisture content of any point in the oil cavity. The moisture analysis processor is based on a moisture diffusion model. It divides the oil cavity space into a three-dimensional mesh and predicts the moisture content at each point within the mesh using the moisture diffusion model, thereby generating the three-dimensional moisture distribution map.

[0011] Specifically, the step of establishing a three-dimensional coordinate system with the center of the bottom of the oil cavity as the origin, calculating the diffusion rate gradient at each monitoring point, and generating a water intrusion path map includes: The moisture analysis processor invokes the established three-dimensional coordinate system with the center of the bottom of the oil chamber as the origin; Based on the three-dimensional distribution map of moisture from multiple consecutive time series, the moisture content data of each monitoring point at different time points are extracted, and the rate of change of moisture content at each monitoring point per unit time, i.e., the diffusion rate, is calculated. By combining the spatial coordinates of each monitoring point and the corresponding diffusion rate, the diffusion rate gradient of each monitoring point on its spatial coordinates is calculated to obtain the gradient vector. Based on the gradient vectors of all monitoring points, a streamline tracing algorithm is used to simulate the direction and path of water movement, generating the water intrusion path map; the water intrusion path map indicates the trend and channels of water spreading from the bottom of the cavity upwards and around the cavity.

[0012] Specifically, the step of marking potential leakage areas based on a preset water rise rate threshold according to a water intrusion path map includes: The moisture analysis processor analyzes the generated moisture intrusion path map and identifies one or more core intrusion paths with the largest diffusion rate gradient modulus in the map. Along the identified core intrusion path, extract the diffusion rate data of each segment on the core intrusion path; The diffusion rate of each segment on the core intrusion path is compared with a preset moisture rise rate threshold. If the diffusion rate of any segment of the core intrusion path is greater than the preset moisture rise rate threshold, the area corresponding to the core intrusion path is marked as the potential leakage area. The moisture analysis processor outputs potential leakage area data, including the location, range, and maximum diffusion rate of the potential leakage area. It also extracts the real-time moisture content signal corresponding to the triggering of the potential leakage area marker and uses it as a moisture monitoring signal.

[0013] Specifically, the risk assessment unit, based on model parameters trained using extracted dynamic features and historical fault data, outputs a leakage risk level after Bayesian network fusion analysis, including: The risk assessment unit acquires the real-time dynamic features extracted by the edge data processing unit, as well as the parameters of the Bayesian network model trained based on historical fault data; the Bayesian network model parameters include the conditional probability table of the Bayesian network. The risk assessment unit inputs real-time dynamic features as evidence into the corresponding input nodes of the Bayesian network model, and initiates the probabilistic reasoning process in conjunction with the built-in conditional probability table. The Bayesian network performs probabilistic inference based on the real-time dynamic features of the input and the conditional probability table, and updates the posterior probability of all nodes in the Bayesian network. The probability of each risk level corresponding to the output node of the Bayesian network is calculated based on the inference results, and the level with the highest probability is taken as the final output result; the level includes low risk, medium risk and high risk.

[0014] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention proposes a real-time detection system for the grease chamber of the shield tail seal, and optimizes and improves its architecture, operation steps and processes. The system has the advantages of simple process, low investment and operating costs and low production costs.

[0015] 2. This invention proposes a real-time detection system for the grease cavity of a shield tail seal. This system achieves all-round real-time perception of the grease cavity through a multi-physics field monitoring module. The pressure difference monitoring unit accurately locates high-risk leakage areas based on a pressure difference dynamic model and high-risk area analysis. The moisture monitoring unit efficiently identifies potential leakage areas by using a support vector machine model and diffusion path tracking. The combination of the two realizes multi-dimensional fault early warning from structural deformation to moisture intrusion, improving the comprehensiveness and accuracy of grease cavity anomaly detection.

[0016] 3. This invention proposes a real-time detection system for the grease chamber of a shield tail seal. The intelligent early warning analysis module extracts key dynamic features through the edge data processing unit and combines them with the Bayesian network fusion analysis technology of the risk assessment unit. It can not only output a quantitative leakage risk level, but also trigger a secondary early warning when it detects three abnormalities: a sudden drop in grease viscosity, a sudden increase in moisture, and an excessive pressure difference. This realizes intelligent processing of the entire process from data acquisition and feature extraction to risk assessment, reducing the operational risks caused by leakage failures. Attached Figure Description

[0017] Figure 1 This is a diagram illustrating the architecture of a real-time detection system for the sealing grease chamber at the tail of a shield, according to the present invention. Figure 2 This is a flowchart illustrating the principle of a real-time detection system for a shield tail sealing grease cavity according to the present invention. Figure 3 This is a flowchart illustrating the process of generating a spatiotemporal distribution map of pressure difference in a real-time detection system for a shield tail sealing grease cavity according to the present invention. Detailed Implementation

[0018] Example 1 Please see Figures 1-2One embodiment of the present invention provides a real-time detection system for the grease chamber of a shield tail seal, comprising: The system includes a multiphysics monitoring module and an intelligent early warning analysis module; the multiphysics monitoring module includes a pressure difference monitoring unit and a moisture monitoring unit; the intelligent early warning analysis module includes an edge data processing unit and a risk assessment unit. The pressure difference monitoring unit collects grease pressure data in the grease chamber and water pressure data in the external environment, establishes a dynamic model of pressure difference and generates a spatiotemporal distribution map of pressure difference. Then, it identifies pressure imbalance points based on the circumferential center of the grease chamber, calculates the pressure difference between the pressure imbalance point and the circumferentially symmetrical position point, generates an abnormal distribution map of pressure difference, and locates pressure weak areas based on fluid mechanics principles. Based on a preset threshold, it marks high-risk leakage areas and outputs high-risk leakage area data and corresponding pressure monitoring signals. The moisture monitoring unit collects moisture data at the bottom of the oil chamber, constructs a moisture diffusion model using a support vector machine algorithm, and outputs a three-dimensional moisture distribution map. Then, a three-dimensional coordinate system is established with the center of the bottom of the oil chamber as the origin, the diffusion rate gradient of each monitoring point is calculated, and a moisture intrusion path map is generated. Based on the moisture intrusion path map, potential leakage areas are marked according to a preset moisture rise rate threshold, and potential leakage area data and corresponding moisture monitoring signals are output. Based on data from high-risk leakage areas, pressure monitoring signals, potential leakage areas, and moisture monitoring signals, the edge data processing unit uses the isolated forest algorithm to extract dynamic features; these dynamic features include the rate of change of pressure difference and the rate of moisture rise. The isolated forest algorithm is existing technology in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0019] It should be noted that the edge data processing unit is deployed at the edge of data acquisition, enabling real-time data processing and analysis, reducing latency and bandwidth consumption during data transmission to the cloud. The isolated forest algorithm has efficient anomaly detection and feature extraction capabilities, accurately extracting dynamic features reflecting changes in the operating status of the grease chamber from a large amount of monitoring data. The rate of change of pressure difference reflects the imbalance trend between external water pressure and internal grease pressure, and is the core driving indicator of water intrusion. The rate of water rise reflects the speed of water intrusion. These two dynamic features are key indicators for assessing the risk of grease chamber leakage. By extracting these dynamic features, concise and crucial input parameters are provided for subsequent risk assessment, improving the efficiency and accuracy of risk assessment.

[0020] The risk assessment unit outputs the leakage risk level based on the extracted dynamic features and model parameters trained by combining historical fault data through Bayesian network fusion analysis. When triple evidence of a sudden drop in grease viscosity, a sudden increase in moisture, and an excessive pressure difference is detected during the fusion analysis, a level-two warning is triggered.

[0021] It should be noted that the architecture of this system is designed to closely match the core failure mechanism of the shield tail seal: the multi-physics monitoring module captures the root cause of failure—the imbalance between external water pressure and internal grease pressure—through the pressure difference monitoring unit, and tracks the deterioration process of grease performance after moisture intrusion through the moisture monitoring unit. The two work together to achieve full-chain monitoring from the source of risk to the result of failure. The intelligent early warning analysis module, through the fusion of edge data processing and Bayesian networks, transforms multi-dimensional parameters into quantitative risk levels, avoiding the limitations of traditional single-parameter monitoring. This ensures early identification and accurate early warning of seal failure risks, providing a scientific basis for timely on-site measures such as replenishing grease and adjusting pressure.

[0022] Example 2 The pressure difference monitoring unit described in this embodiment includes two types of sensors; The first type of sensor is a pressure sensor array that is uniformly arranged in at least two rings on the inner wall of the grease chamber. The ring closest to the tail brush is the first pressure array, which is used to collect grease pressure data in the grease chamber. The other ring is the second pressure array, which is used to assist in verifying the uniformity of pressure distribution. The second type of sensor is an external water pressure sensor group embedded on the outside of the shield tail shell, corresponding to the grease chamber, used to collect external environmental water pressure data; Both the first type of sensor and the second type of sensor are connected to a pressure data acquisition device; the pressure data acquisition device is used to convert the pressure signals collected by the sensors into digital pressure signals, and to complete the time synchronization upload of the internal pressure and external water pressure data to ensure the consistency of the data in the time dimension.

[0023] It should be noted that in this invention, the dual-ring array layout enables focused monitoring of high-risk areas near the tail brush: the first pressure array can capture local pressure anomalies caused by tail brush wear and uneven grease distribution at close range, while the second pressure array provides a reference for the overall pressure distribution. The combination of the two forms a pressure distribution verification mechanism. The corresponding arrangement of the external water pressure sensor group and the internal pressure sensor can directly calculate the pressure difference at the same spatial location, avoiding pressure difference calculation errors caused by positional deviations and significantly improving the accuracy of pressure difference monitoring.

[0024] Please see Figure 3 The process of constructing a dynamic model of the pressure difference and generating a spatiotemporal distribution map of the pressure difference is as follows: A1: The pressure sensor array of the pressure differential monitoring unit and the external water pressure sensor group synchronously collect data to generate grease pressure data in the grease chamber and external environmental water pressure data, respectively, and transmit them to the pressure data acquisition instrument. Furthermore, the process by which the pressure sensor array and the external water pressure sensor group convert the grease pressure data inside the grease chamber and the external environmental water pressure data into changes in the wavelength of an optical signal includes: (1) The sensitive element of the pressure sensor array, such as the elastic diaphragm, is in direct contact with the inner wall of the grease chamber, while the sensitive element of the external water pressure sensor group is in contact with the water body outside the shield tail. When the grease injection pressure in the grease chamber or the external environmental water pressure changes, the pressure is transmitted to the surface of the sensitive element through the medium, such as grease or water, causing the elastic diaphragm to produce a mechanical deformation proportional to the pressure change: when the pressure increases, the diaphragm bulges outward, and when the pressure decreases, the diaphragm is concave inward. The degree of deformation increases linearly with the increase of the pressure change. (2) The elastic diaphragm and the fiber grating are rigidly connected. The mechanical deformation of the diaphragm is transmitted to the fiber grating body through the encapsulation structure. When the diaphragm bulges, the encapsulation structure generates an axial tensile force on the fiber grating, which increases the period of the grating. When the diaphragm is concave, the encapsulation structure generates an axial compressive force on the fiber grating, which decreases the period of the grating. During this process, the change in the period of the grating is strictly linearly related to the deformation of the elastic diaphragm, and then linearly related to the change in pressure. The encapsulation structure of the fiber grating refers to the metal sleeve or epoxy resin adhesive layer. (3) The core reflection characteristics of the fiber grating are determined by the Bragg equation, and the center wavelength of the reflected light varies with the grating period and the effective refractive index of the fiber; wherein, the Bragg equation is the prior art in this field and is not an inventive solution of this application, and will not be described in detail here; (4) The built-in broadband light source of the sensor emits a continuous spectrum of light signal, which is injected into the fiber optic grating through the fiber coupler. The fiber optic grating only reflects a specific wavelength light signal that matches the center wavelength of its current reflected light. Other wavelength light signals are transmitted or absorbed. Then, the reflected light signal containing wavelength offset information is transmitted to the light signal output end through the same coupler. At this time, the wavelength change of the output light signal has formed a one-to-one correspondence with the original pressure change, completing the conversion of pressure data into the wavelength change of the light signal.

[0025] A2: The pressure data acquisition instrument timestamps the oil pressure data in the oil chamber and the external environmental water pressure data, and then converts them into digital pressure signals before transmitting them to the pressure difference analysis processor; In this invention, the pressure data acquisition unit of the pressure differential monitoring unit is connected to the pressure sensor array and the external water pressure sensor group. It is specifically responsible for acquiring the wavelength change of the optical signal output by the sensor array. The pressure data acquisition unit can perform high-precision demodulation and conversion of the optical signal, accurately convert the wavelength change of the optical signal into a digital pressure signal, and stably transmit the digital pressure signal to the pressure differential analysis processor. During the demodulation process, the pressure data acquisition unit filters out noise interference through signal processing algorithms to ensure the quality of the digital pressure signal.

[0026] Furthermore, the specific steps of A2 include: (1) The pressure data acquisition instrument receives the analog pressure signals output by the pressure sensor array and the external water pressure sensor group through the industrial Ethernet interface. First, the weak voltage signal output by the sensor is amplified to the standard range through the signal conditioning circuit to reduce the error of subsequent analog-to-digital conversion. In this invention, the weak voltage signal refers to 0-5V, and the standard range refers to 0-10V. (2) Start the built-in synchronous clock module to assign an independent timestamp to each sensor signal. The timestamp accuracy reaches 1ms. By comparing the timestamps of each sensor signal, data with a time deviation of more than 5ms due to transmission delay are eliminated to ensure that the intracavity pressure and external water pressure data correspond one-to-one at the same time. (3) A 16-bit analog-to-digital converter is used to convert the conditioned analog pressure signal into a digital pressure signal. The conversion rate is 1kHz. During the conversion process, multiple samples are taken to obtain the average value, which further reduces random noise interference. The analog-to-digital converter is a prior art in this field and is not an inventive solution of this application. It will not be described in detail here. (4) Verify the validity of digital pressure signals: If any signal exceeds the normal measurement range for 3 consecutive times, or if the fluctuation of two adjacent data exceeds 0.1 MPa, it is marked as abnormal data and the backup sensor data is automatically activated. If there is no backup sensor, the sliding average of the first 5 data is used to replace it to ensure data continuity. The normal measurement range refers to the internal pressure of 0-2 MPa and the external water pressure of 0-3 MPa. (5) Encode the digital pressure signal according to the Modbus-TCP communication protocol and add data check bits to avoid data tampering or loss during transmission. The Modbus-TCP communication protocol is existing technology in this field and is not an inventive solution of this application. It will not be described in detail here. (6) The encoded digital pressure signal is transmitted to the pressure difference analysis processor via optical fiber, along with timestamp information to ensure time synchronization during subsequent pressure difference calculation.

[0027] A3: The pressure difference analysis processor receives the synchronized digital pressure signal through the data interface, analyzes the measured grease pressure value of each pressure sensor node in the cavity and the corresponding external measured water pressure value, and calculates the measured pressure difference at each spatial location; the measured pressure difference is the difference between the external measured water pressure value and the measured grease pressure value. Furthermore, the pressure difference analysis processor is linked with the tunnel boring machine's PLC system through a data interface, enabling it to acquire real-time operating parameters such as the tunnel boring machine's advance speed and cutterhead rotation speed. When operating parameters change abruptly, such as the advance speed suddenly increasing from 30 mm / min to 50 mm / min, the processor automatically increases the pressure difference calculation frequency to ensure that pressure fluctuations caused by changes in operating conditions are captured. At the same time, during the analysis process, sensor nodes are numbered and bound to spatial coordinates to ensure that each pressure difference data corresponds one-to-one with the actual location, providing a positional reference for the subsequent generation of spatial distribution maps.

[0028] A4: Based on the fluid dynamics pressure transmission theory, the measured pressure difference data of time series is used as input parameters. The measured pressure difference data of multiple consecutive time moments are associated with the corresponding spatial location information. The time series analysis algorithm is used to fit the dynamic change trend of the measured pressure difference, establish a dynamic pressure difference model, and output the theoretical pressure difference value of each point in the grease cavity at different time moments. Furthermore, the specific steps for A4 include: (1) Clarify the hydrodynamic characteristics of the grease chamber: The inside of the grease chamber is high-viscosity shield tail grease, and the outside is a mixture of groundwater and mud. The pressure transmission follows Darcy's law. Based on this, determine the correlation between pressure difference and fluid seepage risk. Darcy's law is the prior art in this field and is not an inventive solution of this application. It will not be elaborated here. (2) The pressure difference analysis processor receives two key data: one is the real-time pressure difference data after synchronization. Secondly, the real-time operating parameters transmitted by the tunnel boring machine PLC system: First, the two key data are preprocessed, including: time synchronization calibration to ensure that the timestamps of the operating parameters and the pressure difference data are completely consistent, and abnormal data is removed. (3) Based on Darcy's law and actual working condition parameters, an initial pressure difference dynamic model is constructed, and the initial pressure difference dynamic model is initially calibrated by historical normal working condition data to determine the coefficient range; the expression of the initial pressure difference dynamic model is a weighted sum of real-time working condition parameters and time; the historical normal working condition data is obtained through construction periods with no leakage records. (4) The autoregressive differential moving average model is used to dynamically fit the real-time pressure difference data for 5 consecutive minutes, optimize the initial pressure difference dynamic model parameters, and obtain the final pressure difference dynamic model. The autoregressive differential moving average model is the prior art in this field and is not an inventive solution of this application. It will not be described in detail here. (5) Substitute the real-time operating parameters into the final pressure difference dynamic model to calculate the theoretical pressure difference at different times and different spatial locations.

[0029] It should be explained that the pressure difference dynamic model not only considers the pressure change trend over time, but also incorporates the tunnel boring machine's operating parameters. This can effectively distinguish between normal pressure fluctuations caused by changes in operating conditions and abnormal pressure differences caused by seal failure, avoiding misjudgments caused by adjustments in propulsion speed or changes in cutterhead load. The theoretical pressure difference value output by the model provides a dynamic benchmark for judging whether the measured pressure difference is abnormal, which is more in line with the actual construction scenario than a fixed threshold.

[0030] A5: At the same time, the measured pressure difference of each node is compared one by one with the theoretical pressure difference value of the corresponding position under the same real-time pressure data calculated by the pressure difference dynamic model to obtain the deviation between the measured and theoretical pressure difference of each node. Furthermore, this precise point-to-point comparison allows for accurate localization of areas with abnormal pressure. This deviation calculation method effectively amplifies subtle pressure anomalies. For example, if the theoretical pressure difference at any node consistently exceeds 0.05 MPa, it indicates that the pressure difference change at that location exceeds the operating condition's influence range, potentially indicating a risk of grease leakage or a localized increase in external water pressure. The deviation calculation results can also be used for sensor fault diagnosis. If the theoretical pressure difference at any node suddenly increases to above 0.2 MPa, and there is no significant change in adjacent nodes, it is preliminarily identified as a sensor fault, triggering the sensor calibration process.

[0031] A6: Using the spatial coordinate information of all sensor nodes and the corresponding measured and theoretical pressure difference deviation data as input, a three-dimensional interpolation algorithm is used for data fusion and spatial estimation to deduce the pressure difference deviation at all locations on the inner wall of the grease cavity. Combined with the change trend in the time dimension, the spatial distribution of the pressure difference deviation is combined with the dynamic change in time to transform it into a pressure difference spatiotemporal distribution map. The pressure difference spatiotemporal distribution map reflects the abnormal distribution and evolution trend of pressure difference in the entire grease cavity at different times.

[0032] Furthermore, the specific steps in A6 include: (1) Organize the basic data of all sensor nodes; the basic data includes two parts: one is the three-dimensional spatial coordinates of each node on the inner wall of the grease chamber, a cylindrical coordinate system is established with the circumferential center of the grease chamber as the origin, and then the cylindrical coordinates are converted into rectangular coordinates; the other is the deviation between the measured and theoretical pressure difference corresponding to each node. (2) The inner wall surface of the grease cavity is discretized into a regular three-dimensional grid as a spatial carrier for estimating the difference between measured and theoretical pressure. The specific division method is as follows: grid lines are divided at 1° intervals along the circumference and at 5 mm intervals along the axial direction. The two intersect to form a grid cell. The vertex of each grid cell is the unknown point for estimating the difference between measured and theoretical pressure. The grid density is determined according to the number of sensor nodes and the size of the grease cavity. The coordinates of each grid vertex are determined by linear interpolation. Then, the coordinates of all grid vertices form an ordered set of spatial points as the output target point of the three-dimensional interpolation algorithm. The linear interpolation is the prior art in this field and is not an inventive solution of this application. It will not be elaborated here. (3) Data fusion and spatial estimation are performed using the Kriging interpolation algorithm, including: By analyzing the spatial correlation between the measured and theoretical pressure difference deviations at known nodes, a spherical variogram is used to fit the spatial variation structure. The spherical variogram is a prior art in this field and is not an inventive solution of this application, so it will not be described in detail here. For each grid vertex, based on the measured and theoretical pressure difference deviation data of its surrounding known nodes and the variogram model, the weighting coefficient is calculated, and the estimated value of the measured and theoretical pressure difference deviation of the vertex is obtained by weighted summation. (4) Convert the estimated pressure deviation values ​​of the three-dimensional mesh vertices into a spatiotemporal distribution map of pressure difference. The specific process is as follows: a. The pressure deviation is represented by a color gradient method. For example, blue represents a negative deviation and red represents a positive deviation. The color depth increases with the absolute value of the deviation. The intermediate value is represented by green. A mapping table between the deviation value and the RGB color value is established. Positive deviation means that the measured pressure difference is greater than the theoretical pressure difference, and negative deviation means that the measured pressure difference is less than the theoretical pressure difference. The specific calibration process of the color gradient method is the prior art in this field and is not an inventive solution of this application. It will not be described in detail here. b. For each grid cell, the color of any point inside the cell is calculated using bilinear interpolation based on the pressure deviation value and corresponding color of its four vertices. Then, using surface rendering technology in computer graphics, all grid cells are stitched together to form a complete surface graphic of the inner wall of the grease cavity. The bilinear interpolation and surface rendering technologies are existing technologies in this field and are not the inventive solutions of this application, so they will not be described in detail here. c. Adding coordinate scales, color rulers, and sensor node location markers to the graph allows it to not only visually display the spatial distribution of pressure deviations but also trace the location of the original data.

[0033] Furthermore, the three-dimensional interpolation algorithm can make full use of limited sensor node data and extend it to all positions on the inner wall of the grease cavity through spatial interpolation, realizing the transformation from discrete point data to continuous surface data. The generated pressure difference spatiotemporal distribution map intuitively reflects the abnormal pressure distribution of the entire inner wall of the grease cavity, enabling staff to grasp the overall pressure status of the grease cavity at a glance, and providing a visual tool for quickly identifying abnormal areas.

[0034] The process of identifying pressure imbalance points using the circumferential center of the grease cavity as a reference, calculating the pressure difference between the pressure imbalance point and a circumferentially symmetrical point, and generating a pressure difference anomalous distribution map includes: B1: The pressure difference analysis processor uses the circumferential center of the grease chamber as a reference to perform a 360-degree circumferential scan of the pressure difference spatiotemporal distribution map, and obtains the position information of all sensor nodes in the pressure difference spatiotemporal distribution map and the deviation between the measured and theoretical pressure difference. B2: If the absolute error of the deviation between the measured and theoretical pressure difference of any sensor node is greater than the first preset deviation threshold, then the node is marked as the primary pressure imbalance point. B3: For the marked primary pressure imbalance point, the pressure difference analysis processor determines its symmetrical position point about the circumferential center of the grease cavity, obtains the measured pressure difference at the symmetrical position point, and then calculates the absolute error between the measured pressure difference between the primary pressure imbalance point and the symmetrical position point to obtain the measured symmetrical pressure difference value. B4: If the measured symmetrical pressure difference is greater than the second preset symmetrical deviation threshold, then the primary pressure imbalance point is confirmed as the final pressure imbalance point. It should be noted that, under normal circumstances, the pressure values ​​at symmetrical positions on the inner wall of the grease chamber should be approximately equal. By comparing symmetrical positions, misjudgments caused by sensor errors or local interference can be further eliminated, improving the accuracy of pressure imbalance point identification. The setting of the second preset symmetry deviation threshold provides a strict standard for the final confirmation of pressure imbalance points, ensuring that only pressure points with genuine structural abnormalities are confirmed.

[0035] B5: The pressure difference analysis processor collects all the final pressure imbalance points and their corresponding measured symmetrical pressure difference values, calculates the pressure difference anomalous coefficient of each final pressure imbalance point, and generates the pressure difference anomalous distribution map by using the contour line method.

[0036] Among them, the calculation of the pressure difference anomaly coefficient can quantify the degree of high-risk area at the pressure imbalance point; the application of the contour line method presents the situation of high-risk area in an intuitive map form; the pressure difference anomaly distribution map clearly shows the location, range and degree of high-risk area on the inner wall of the grease cavity.

[0037] Furthermore, the specific steps of B5 include: (1) Obtain all final pressure imbalance points and their corresponding measured symmetrical pressure difference values; (2) Select sensor nodes in the inner wall of the grease chamber without abnormal areas, i.e. nodes that are not marked as abnormal points, extract the measured pressure difference of nodes in the non-abnormal areas under the current real-time pressure, and calculate the average pressure difference by the arithmetic mean method, which is used as the pressure difference benchmark under the normal stress state of the grease chamber under the current working condition. If the number of nodes in the non-abnormal areas selected is less than the preset threshold, such as less than 30% of the total number of nodes, call the established pressure difference dynamic model, input the current real-time pressure data, calculate the theoretical normal pressure difference of each position in the grease chamber under the current working condition, and take the average value of all theoretical normal pressure differences as the pressure difference benchmark. (3) For each final pressure imbalance point, based on the determined pressure difference benchmark, the local pressure difference value is calculated by combining the measured symmetrical pressure difference value corresponding to the point: Since the measured symmetrical pressure difference value reflects the degree of pressure difference deviation of the final pressure imbalance point relative to the symmetrical position, and the deviation originates from the pressure anomaly of the point itself, the pressure difference benchmark is added to the measured symmetrical pressure difference value to obtain the local pressure difference value of each final pressure imbalance point. This value directly reflects the actual pressure difference level at the anomaly point. (4) Based on the obtained pressure difference benchmark and local pressure difference value, calculate the pressure difference anomalous coefficient for each final pressure imbalance point: the pressure difference anomalous coefficient is the ratio of the local pressure difference value to the pressure difference benchmark. The larger the ratio, the more serious the deviation of the pressure difference at the final pressure imbalance point from the normal state. The pressure anomalous coefficient can be used to quantify the degree of pressure abnormality. (5) Using the spatial coordinates of the final pressure imbalance point and the corresponding pressure difference coefficient as known sample points, the pressure difference coefficients at all locations on the inner wall of the grease cavity are estimated by the inverse distance weighted interpolation method: the weight is determined according to the spatial distance between each unknown location and the known sample point, where the closer the distance, the greater the weight. Then, the pressure difference coefficient at the unknown location is obtained by weighted calculation, and the discrete sample point data is expanded into continuous pressure difference coefficient distribution data covering the entire inner wall of the grease cavity, so as to realize the pressure anomaly distribution characterization from point to surface. The inverse distance weighted interpolation method is the prior art in this field and is not an inventive solution of this application, so it will not be described in detail here. (6) The risk levels are divided according to the numerical range of the pressure difference coefficient, including three levels: low risk, medium risk and high risk. Each level corresponds to a different color, such as blue, yellow and red, which is easy to distinguish visually. (7) Overlay contour lines of different levels onto the three-dimensional model of the inner wall of the oil cavity, mark the pressure difference coefficient value corresponding to each contour line, and add color scale and coordinate scale to the edge of the map. At the same time, mark the actual position of the final pressure imbalance point in the map with triangles so that the original data and the interpolation results correspond to each other.

[0038] The method of locating abnormal deformations using fluid dynamics principles and marking high-risk leakage areas based on preset thresholds includes: C1: The pressure difference analysis processor, based on fluid mechanics principles and combined with the generated pressure difference constant distribution map, analyzes the correlation between the pressure difference constant coefficient and the fluid permeation risk to locate the pressure weak area in the grease cavity; the pressure weak area is defined as a continuous area with a pressure difference constant coefficient greater than 1.2, that is, an area where the external water pressure is significantly higher than the grease pressure inside the cavity. The fluid mechanics principle is prior art in this field and is not an inventive solution of this application, so it will not be elaborated here. Furthermore, the specific steps of C1 include: (1) The pressure difference analysis processor extracts features from the generated pressure difference distribution map. The extracted features include: the range of pressure difference coefficient values ​​in each region, the range of spatial coordinates in high-risk areas, the gradient rate of change of pressure difference coefficients, and the coordinates of the maximum pressure difference coefficient. (2) Preprocessing the extracted feature data: Remove isolated high-value points in the map caused by interpolation errors. These points are usually characterized by a single grid node with a pressure difference abnormality coefficient that is significantly higher than all surrounding nodes. For example, the coefficients of surrounding nodes are all less than or equal to 1.0, but the coefficient of this point is greater than or equal to 1.5, and there are no continuous high-coefficient nodes associated with it. They do not reflect the real pressure abnormality state, but are only interpolation calculation deviations. Then, the abrupt change region of the pressure difference abnormality coefficient gradient change is corrected by the smoothing algorithm so that the gradient change conforms to the continuous law of pressure transmission in fluid mechanics. The continuous law of pressure transmission is that the spatial change of the pressure difference abnormality coefficient should be smooth and gradual, without any unreasonable sudden rise or fall. The smoothing algorithm is the existing technology in this field and is not an inventive solution of this application. It will not be elaborated here. (3) Based on the principles of fluid mechanics, the correlation between the pressure differential coefficient and the risk of fluid seepage is established. The core principles include: a Linear relationship between pressure difference and permeation flow rate: According to Darcy's law, when the characteristics of the medium, such as oil or water, and the permeation area are fixed, the permeation flow rate is linearly positively correlated with the pressure difference. The pressure difference anomalous coefficient is equal to the ratio of the local pressure difference to the normal pressure difference baseline. Therefore, the larger the pressure difference anomalous coefficient, the larger the local pressure difference, the larger the corresponding fluid permeation flow rate, and the higher the permeation risk. bPressure transmission characteristics of the grease chamber: The grease chamber is filled with high-viscosity shield tail grease. Under normal operating conditions, the pressure inside the chamber is in balance with the external water pressure. The high viscosity of the grease can effectively prevent the penetration of external water. When the pressure difference coefficient is greater than 1.2, the external water pressure is significantly higher than the pressure of the grease inside the chamber. The anti-permeability of the grease cannot offset the effect of the water pressure, and the risk of fluid penetration increases sharply. The larger the coefficient, the more obvious the increase in the risk of penetration. This characteristic provides a core basis for subsequent location of pressure-weak areas. (4) Based on the above principles and combined with the preset definition of pressure weak areas, the location basis of pressure weak areas is determined as follows: In the pressure difference distribution map, the area with a pressure difference coefficient greater than 1.2 is the area with increased fluid permeation risk, which can be preliminarily identified as a pressure weak area; the coordinate point where the maximum pressure difference coefficient is located corresponds to the location with the most serious permeation risk, which should be the focus of attention. The preset definition of pressure weak area refers to a continuous area with a pressure difference coefficient greater than 1.2, that is, an area where the external water pressure is higher than the pressure of the grease inside the cavity. (5) Identify the spatial boundaries of high-risk areas in the pressure difference distribution map, including: a Set the high pressure difference constant threshold to 1.2, traverse all grid nodes in the pressure difference constant distribution map, and mark the nodes with pressure difference constant coefficient greater than or equal to 1.2 as suspected weak nodes. The area formed by all suspected weak nodes is the suspected pressure weak zone. b Starting from the extracted core high-value point, a region growing algorithm is used to gradually merge suspected weak nodes that are adjacent to the core high-value point and have a pressure difference anomalous coefficient greater than or equal to 1.2, forming multiple independent connected regions. Each connected region is an independent suspected pressure weak area, ensuring that no isolated suspected weak nodes are missed. The core high-value point refers to the coordinate point where the maximum pressure difference anomalous coefficient is located. The region growing algorithm is existing technology in this field and is not an inventive solution of this application, so it will not be described in detail here. c For each suspected pressure weak area, its spatial boundary is extracted using an edge detection algorithm: in the circumferential dimension, the minimum and maximum angles of the area coverage are determined, such as from 30 degrees to 60 degrees; in the axial dimension, the minimum and maximum heights of the area coverage are determined, such as from 50 mm to 150 mm from the bottom of the grease cavity, and the range of each suspected pressure weak area is defined in the form of a rectangular bounding box. The edge detection algorithm is prior art in this field and is not an inventive solution of this application, so it will not be described in detail here. (6) Verify the rationality of the suspected pressure weak zone by combining fluid mechanics principles, and eliminate false areas caused by non-real permeation risk factors.

[0039] C2: Analyze the identified pressure-weak areas, and mark areas in the pressure difference distribution map where the pressure difference coefficient value is greater than the preset pressure difference threshold, or areas where the measured symmetrical pressure difference value is greater than the preset pressure difference threshold, as high-risk leakage areas. Output high-risk leakage area data including the location, area, maximum pressure difference coefficient value, and maximum measured symmetrical pressure difference value of the high-risk leakage areas. At the same time, extract the real-time digital pressure signal corresponding to the high-risk leakage area marking, and output it as a pressure monitoring signal. The preset pressure difference threshold is determined according to the anti-permeability grade of the shield tail grease, and is usually set to 0.2 MPa.

[0040] The moisture monitoring unit includes multiple high-frequency capacitive moisture sensors deployed in the sedimentation zone at the bottom of the oil chamber; the high-frequency capacitive moisture sensors are connected to a data acquisition unit, which receives and uploads the digital signals of moisture content collected by the sensors.

[0041] The moisture monitoring unit collects moisture data from the bottom of the oil chamber, constructs a moisture diffusion model using a support vector machine algorithm, and outputs a three-dimensional moisture distribution map, including: D1: The sensors of the moisture monitoring unit collect moisture content data from multiple monitoring points at the bottom of the oil chamber, convert it into a digital signal of moisture content, and then transmit it to the data acquisition unit; The moisture monitoring unit includes multiple high-frequency capacitive moisture sensors deployed in the sedimentation area at the bottom of the oil chamber. The sedimentation area at the bottom of the oil chamber is a region where moisture easily accumulates. Deploying the sensors here can capture the signal of moisture intrusion earliest and most accurately. The deployment of multiple sensors forms a multi-point monitoring network, which can comprehensively reflect the moisture distribution at the bottom of the oil chamber, avoid the missed detection problem that may occur in single-point monitoring, and improve the reliability of moisture monitoring.

[0042] D2: The data acquisition unit transmits the digital signal of moisture content to the moisture analysis processor; D3: The moisture analysis processor uses the received digital signal of moisture content as the training sample, takes the coordinates of each monitoring point in the three-dimensional coordinate system established with the center of the bottom of the oil cavity as the origin as the input feature, and takes the moisture content value of the corresponding monitoring point as the output label. It uses the support vector machine regression algorithm for training and learning to construct a moisture diffusion model that predicts the moisture content of any point in the oil cavity. The support vector machine regression algorithm is the prior art in this field and is not an inventive solution of this application, so it will not be described in detail here. D4: The moisture analysis processor is based on a moisture diffusion model. It divides the oil cavity space into three-dimensional grids and predicts the moisture content of each point in the grid using the moisture diffusion model to generate the three-dimensional moisture distribution map.

[0043] Among them, the three-dimensional mesh splitting divides the oil cavity space into small mesh units, ensuring the spatial accuracy of the prediction; the moisture diffusion model provides a reliable basis for the prediction of moisture content in each mesh unit; the generated three-dimensional moisture distribution map intuitively shows the spatial distribution of moisture in the oil cavity, enabling staff to clearly understand the distribution status and accumulation area of ​​moisture in the oil cavity, and providing a visual reference for judging the degree and scope of moisture intrusion.

[0044] Furthermore, the specific steps of D4 include: (1) Call the constructed moisture diffusion model and clarify the core parameters and applicable scope of the model; the core principle of the moisture diffusion model is to fit the nonlinear law of spatial moisture distribution by the moisture data of known monitoring points; (2) Establish a three-dimensional coordinate system with the center of the bottom of the oil cavity as the origin, and delineate the grid range according to the actual size of the oil cavity to ensure that the entire space of the oil cavity is covered, including the bottom deposition area, the inner wall area and the top space; (3) Determine the grid size according to the required accuracy of moisture monitoring, and form a cubic grid unit. The vertex of each grid unit is the spatial point where the moisture content needs to be predicted. (4) Calculate the three-dimensional coordinates of each grid vertex according to the grid division rules, and sort all the vertex coordinates according to their spatial positions to form an ordered set of three-dimensional coordinate points, which will be used as the input for moisture content prediction. (5) Predict the moisture content of each grid vertex using the moisture diffusion model, including: converting the three-dimensional coordinates of the grid vertex into the feature grid used during model training to obtain the standardized grid vertex coordinates; inputting the standardized grid vertex coordinates into the moisture diffusion model in batches, and the moisture diffusion model calculates the predicted moisture content of each vertex through the built-in support vector machine regression function; performing local verification on the prediction results, focusing on checking the predicted values ​​of grid vertices near known monitoring points. (6) Convert the predicted moisture content of the grid vertices into an intuitive three-dimensional moisture distribution map, including: dividing the levels according to the numerical range of moisture content and establishing the mapping relationship between moisture content and RGB color values; for each cubic grid cell, according to the predicted moisture content of its 8 vertices and the corresponding colors, using trilinear interpolation to calculate the color value of any point inside the cell, and then using volume rendering technology in computer graphics to combine all grid cells into a complete three-dimensional spatial model of the oil cavity; for the boundary areas such as the inner wall and bottom of the oil cavity, highlighting the spatial distribution of high moisture zones by adjusting the transparency; adding three-dimensional coordinate system scales, moisture content color scales and key position marks to the map; eliminating the jagged visual error of the grid edges through a smoothing algorithm, making the moisture distribution in the map transition naturally, while preserving the boundary clarity of the high moisture zone; (7) Compare the newly collected moisture monitoring data with the predicted values ​​at the corresponding locations in the three-dimensional moisture distribution map, calculate the overall average deviation, and if the deviation exceeds the preset threshold, trigger the model to retrain and optimize the moisture diffusion model parameters with the latest data.

[0045] The process of establishing a three-dimensional coordinate system with the center of the bottom of the oil cavity as the origin, calculating the diffusion rate gradient at each monitoring point, and generating a water intrusion path map includes: E1: The moisture analysis processor calls the established three-dimensional coordinate system with the center of the bottom of the oil chamber as the origin; E2: Based on the three-dimensional distribution map of moisture from multiple consecutive time series, extract the moisture content data of each monitoring point at different time points, and calculate the rate of change of moisture content at each monitoring point per unit time, i.e., the diffusion rate. Furthermore, the moisture content at each monitoring point per unit time is calculated by calculating the difference in moisture content change at each monitoring point per unit time, while the rate of change of moisture content at each monitoring point per unit time is calculated by calculating the ratio of the moisture content at each monitoring point per unit time to the unit time, where the time interval between two adjacent moments is used. As a unit of time for calculation; E3: Combining the spatial coordinates of each monitoring point with the corresponding diffusion rate, calculate the diffusion rate gradient of each monitoring point on its spatial coordinates to obtain the gradient vector; The diffusion rate gradient reflects the rate of change of water diffusion speed in space, while the gradient vector indicates the direction of the fastest change in water diffusion speed. By calculating the diffusion rate gradient, we can understand the diffusion law and dynamic mechanism of water in the oil cavity.

[0046] E4: Based on the gradient vectors of all monitoring points, the streamline tracing algorithm is used to simulate the direction and path of water movement, and generate the water intrusion path map; the water intrusion path map indicates the trend and channel of water spreading from the bottom of the cavity upwards and around. The streamline tracing algorithm is existing technology in this field and is not an inventive solution of this application, so it will not be described in detail here.

[0047] Furthermore, the streamline tracing algorithm can simulate the flow path of water in the oil cavity based on the direction of the gradient vector, intuitively showing the trend and channels of water spreading from the bottom of the cavity upwards and around; the generation of the water intrusion path map enables staff to clearly grasp the specific path of water intrusion.

[0048] The step of marking potential leakage areas based on a water intrusion path map and a preset water rise rate threshold includes: F1: The moisture analysis processor analyzes the generated moisture intrusion path map and identifies one or more core intrusion paths with the largest diffusion rate gradient modulus in the map. Furthermore, the diffusion rate gradient modulus reflects the magnitude of the diffusion rate gradient. The larger the modulus, the more drastic the change in water diffusion rate along the core intrusion path, indicating that it is the main channel for water intrusion. Identifying the main diffusion path can focus on those water intrusion channels that pose the greatest threat to the safety of the grease cavity.

[0049] Furthermore, the specific steps in F1 include: (1) Read the generated water intrusion path map; the water intrusion path map contains diffusion rate gradient vector data of each point in the three-dimensional space of the oil cavity, and the gradient vector of each point consists of three components. Composition, respectively representing along x , y , z The rate of change of diffusion rate in the axial direction; (2) Calculate the diffusion rate gradient magnitude; the diffusion rate gradient magnitude is obtained based on the components of the gradient vector and using the Euclidean distance calculation formula; (3) Statistically analyze the numerical range of the gradient modulus of the entire domain and divide it into low modulus region, medium modulus region and high modulus region from low to high. Each region corresponds to a different path significance level. Among them, the high modulus region usually corresponds to the active path of water invasion and is the key area for path identification. (4) A density clustering algorithm is used to cluster the grid nodes in the high modulus region. Spatially continuous nodes with modulus lengths all above the threshold are aggregated into multiple independent high modulus regions. Each clustered region represents the core zone of a potential intrusion path. The cluster density and average modulus length reflect the activity level of the path. The density clustering algorithm is prior art in this field and is not an inventive solution of this application, so it will not be described in detail here. (5) In each high modulus clustering region, the node with the largest gradient modulus is selected as the seed point of the path. This seed point is the location where the diffusion rate changes most drastically on the path, usually corresponding to the vicinity of the source of water intrusion or the key channel. (6) Starting from the seed point, the path is traced in the opposite direction of the gradient vector. Specifically, the node with the smallest angle between the gradient vector direction and the current node direction and the second largest modulus length is searched in the neighborhood of the current node, and it is taken as the next node of the path. This process is repeated until the boundary of the grease cavity is traced or the modulus length drops below the threshold of the low modulus length region, forming a complete potential path. (7) For all extracted potential paths, analyze their spatial intersection relationship. If two paths intersect at any node and the subsequent tracking direction is consistent, they are merged into one path. If a path branches out, that is, multiple directions are extended from a node, the branch with the largest modulus is taken as the main path and the rest are taken as secondary branches. Finally, a path topology network composed of the main path and secondary branches is formed, where the main path is a potential intrusion path. (8) For each potential path, calculate the average gradient magnitude of all its nodes. The higher the average value, the higher the overall activity level of the path. (9) Sort the paths from highest to lowest according to the average gradient magnitude, and select the top N paths as the core intrusion paths; (10) In the three-dimensional model of the inner wall of the grease cavity, the core invasion path is drawn with thick colored lines. The core invasion path is drawn in bright red, and the secondary path is drawn in orange. The line width is proportional to the peak value of the path modulus, so as to intuitively distinguish the importance of the path.

[0050] F2: Extract the diffusion rate data of each segment along the identified core intrusion path; F3: Compare the diffusion rate of each segment on the core intrusion path with a preset moisture rise rate threshold. If the diffusion rate of any segment of the core intrusion path is greater than the preset moisture rise rate threshold, then mark the area corresponding to the core intrusion path as the potential leakage area. F4: The moisture analysis processor outputs potential leakage area data including the location, range, and maximum diffusion rate value of the potential leakage area, and extracts the real-time moisture content signal corresponding to the triggering of the potential leakage area marker as a moisture monitoring signal.

[0051] The risk assessment unit, based on model parameters trained using extracted dynamic features and historical fault data, outputs a leakage risk level after Bayesian network fusion analysis, including: G1: The risk assessment unit obtains the real-time dynamic features extracted by the edge data processing unit, as well as the parameters of the Bayesian network model trained based on historical fault data; the Bayesian network model parameters include the conditional probability table of the Bayesian network, wherein the Bayesian network model is the prior art in this field and is not an inventive solution of this application, and will not be described in detail here. G2: The risk assessment unit inputs real-time dynamic features as evidence into the corresponding input nodes of the Bayesian network model, and initiates the probabilistic inference process in conjunction with the built-in conditional probability table. G3: The Bayesian network performs probabilistic inference based on the real-time dynamic features of the input and the conditional probability table, and updates the posterior probability of all nodes in the Bayesian network. G4: Calculate the probability of each risk level corresponding to the output node of the Bayesian network based on the inference results, and take the level with the highest probability as the final output result; the level includes low risk, medium risk and high risk.

[0052] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments under the guidance of the present invention without departing from the spirit and scope of the present invention. All of these variations are within the protection scope of the present invention.

Claims

1. A real-time detection system for the grease chamber of a shield tail seal, characterized in that, include: Multiphysics monitoring module and intelligent early warning analysis module; The multiphysics monitoring module includes a pressure difference monitoring unit and a moisture monitoring unit; The intelligent early warning analysis module includes an edge data processing unit and a risk assessment unit; The pressure difference monitoring unit collects oil pressure data in the oil cavity and water pressure data in the external environment, constructs a dynamic model of pressure difference and generates a spatiotemporal distribution map of pressure difference. Then, it identifies pressure imbalance points based on the circumferential center of the oil cavity, calculates the pressure difference between the pressure imbalance point and the circumferentially symmetrical position point, generates an abnormal distribution map of pressure difference, and locates pressure weak areas based on fluid mechanics principles. Based on a preset threshold, it marks high-risk leakage areas and outputs high-risk leakage area data and corresponding pressure monitoring signals. The moisture monitoring unit collects moisture data at the bottom of the oil chamber, constructs a moisture diffusion model using a support vector machine algorithm, and outputs a three-dimensional moisture distribution map. Then, a three-dimensional coordinate system is established with the center of the bottom of the oil chamber as the origin, the diffusion rate gradient of each monitoring point is calculated, and a moisture intrusion path map is generated. Based on the moisture intrusion path map, potential leakage areas are marked according to a preset moisture rise rate threshold, and potential leakage area data and corresponding moisture monitoring signals are output. Based on data from high-risk leakage areas, pressure monitoring signals, potential leakage areas, and moisture monitoring signals, the edge data processing unit uses the isolated forest algorithm to extract dynamic features; these dynamic features include the rate of change of pressure difference and the rate of moisture rise. The risk assessment unit outputs the leakage risk level based on the extracted dynamic features and model parameters trained by combining historical fault data through Bayesian network fusion analysis. When triple evidence of a sudden drop in grease viscosity, a sudden increase in moisture, and an excessive pressure difference is detected during the fusion analysis, a level-two warning is triggered.

2. The real-time detection system for the grease chamber of a shield tail seal as described in claim 1, characterized in that, The pressure difference monitoring unit includes two types of sensors; The first type of sensor is a pressure sensor array that is uniformly arranged in at least two rings on the inner wall of the grease chamber. The ring closest to the tail brush is the first pressure array, which is used to collect grease pressure data in the grease chamber. The other ring is the second pressure array, which is used to assist in verifying the uniformity of pressure distribution. The second type of sensor is an external water pressure sensor group embedded on the outside of the shield tail shell, corresponding to the grease chamber, used to collect external environmental water pressure data; Both the first type of sensor and the second type of sensor are connected to a pressure data acquisition device; the pressure data acquisition device is used to convert the pressure signals collected by the sensors into digital pressure signals and to synchronize the uploading of intracavity pressure and external water pressure data.

3. The real-time detection system for the grease chamber of a shield tail seal as described in claim 2, characterized in that, The process of constructing a dynamic model of the pressure difference and generating a spatiotemporal distribution map of the pressure difference is as follows: The pressure sensor array of the pressure differential monitoring unit and the external water pressure sensor group synchronously collect data to generate grease pressure data in the grease chamber and external environmental water pressure data, respectively, and transmit them to the pressure data acquisition instrument. The pressure data acquisition instrument timestamps the oil pressure data in the oil chamber and the external environmental water pressure data, and then converts them into digital pressure signals before transmitting them to the pressure difference analysis processor. The pressure difference analysis processor receives the synchronized digital pressure signal through the data interface, analyzes the measured grease pressure value of each pressure sensor node in the cavity and the corresponding external measured water pressure value, and calculates the measured pressure difference at each spatial location; the measured pressure difference is the difference between the external measured water pressure value and the measured grease pressure value. Based on the fluid dynamics pressure transmission theory, the measured pressure difference data of time series is used as input parameters. The measured pressure difference data of multiple consecutive time moments are associated with the corresponding spatial location information. The time series analysis algorithm is used to fit the dynamic change trend of the measured pressure difference, establish a dynamic pressure difference model, and output the theoretical pressure difference value of each point in the grease cavity at different time moments. Meanwhile, the measured pressure difference of each node is compared one by one with the theoretical pressure difference value of the corresponding position under the same real-time pressure data calculated by the pressure difference dynamic model to obtain the deviation between the measured and theoretical pressure difference of each node. Using the spatial coordinates of all sensor nodes and the corresponding measured and theoretical pressure difference deviation data as input, a three-dimensional interpolation algorithm is used for data fusion and spatial estimation to deduce the pressure difference deviation at all locations on the inner wall of the grease cavity. Combined with the changing trend in the time dimension, the spatial distribution of the pressure difference deviation is combined with the dynamic changes in time to transform it into a pressure difference spatiotemporal distribution map. The pressure difference spatiotemporal distribution map reflects the abnormal distribution and evolution trend of pressure difference in the entire grease cavity at different times.

4. The real-time detection system for the grease chamber of the shield tail seal as described in claim 3, characterized in that, The process of identifying pressure imbalance points using the circumferential center of the grease cavity as a reference, calculating the pressure difference between the pressure imbalance point and a circumferentially symmetrical point, and generating a pressure difference anomalous distribution map includes: The pressure difference analysis processor uses the circumferential center of the grease chamber as a reference to perform a 360-degree circumferential scan of the pressure difference spatiotemporal distribution map, and obtains the position information of all sensor nodes in the pressure difference spatiotemporal distribution map and the deviation between the measured and theoretical pressure difference. If the absolute error between the measured and theoretical pressure difference of any sensor node is greater than the first preset deviation threshold, then the node is marked as a primary pressure imbalance point. For the marked primary pressure imbalance point, the pressure difference analysis processor determines its symmetrical position point about the circumferential center of the grease cavity, obtains the measured pressure difference at the symmetrical position point, and then calculates the absolute error between the measured pressure difference between the primary pressure imbalance point and the symmetrical position point to obtain the measured symmetrical pressure difference value. If the measured symmetrical pressure difference is greater than the second preset symmetrical deviation threshold, then the primary pressure imbalance point is confirmed as the final pressure imbalance point. The pressure difference analysis processor collects all final pressure imbalance points and their corresponding measured symmetrical pressure difference values, calculates the pressure difference anomalous coefficient for each final pressure imbalance point, and generates the pressure difference anomalous distribution map using the contour line method.

5. The real-time detection system for the grease chamber of the shield tail seal as described in claim 4, characterized in that, The method of locating pressure-weak areas using fluid dynamics principles and marking high-risk leakage areas based on preset thresholds includes: The pressure differential analysis processor, based on fluid mechanics principles and combined with the generated pressure differential distribution map, analyzes the correlation between the pressure differential coefficient and fluid permeation risk, and locates the pressure-weak areas in the grease cavity; the pressure-weak areas are defined as continuous regions with a pressure differential coefficient greater than 1.2, that is, regions where the external water pressure is significantly higher than the grease pressure inside the cavity; The identified pressure-weak areas are analyzed, and areas with pressure difference coefficient values ​​greater than a preset pressure difference threshold in the pressure difference distribution map, or areas with measured symmetrical pressure difference values ​​greater than a preset pressure difference threshold, are marked as high-risk leakage areas. High-risk leakage area data, including location, area, maximum pressure difference coefficient value, and maximum measured symmetrical pressure difference value, is output. Simultaneously, the real-time digital pressure signal corresponding to the triggering of the high-risk leakage area marking is extracted and output as a pressure monitoring signal. The preset pressure difference threshold is determined based on the anti-permeability grade of the shield tail grease.

6. The real-time detection system for the grease chamber of a shield tail seal as described in claim 5, characterized in that, The moisture monitoring unit includes multiple high-frequency capacitive moisture sensors deployed in the sedimentation zone at the bottom of the oil chamber; the high-frequency capacitive moisture sensors are connected to a data acquisition unit, which receives and uploads the digital signals of moisture content collected by the sensors.

7. The real-time detection system for the grease chamber of a shield tail seal as described in claim 6, characterized in that, The moisture monitoring unit collects moisture data from the bottom of the oil chamber, constructs a moisture diffusion model using a support vector machine algorithm, and outputs a three-dimensional moisture distribution map, including: The sensors in the moisture monitoring unit collect moisture content data from multiple monitoring points at the bottom of the oil chamber, convert it into a digital signal of moisture content, and then transmit it to the data acquisition unit. The data acquisition unit transmits the digital signal of moisture content to the moisture analysis processor; The moisture analysis processor uses the received digital signal of moisture content as the training sample, takes the coordinates of each monitoring point in the three-dimensional coordinate system established with the center of the bottom of the oil cavity as the origin as the input feature, takes the moisture content value of the corresponding monitoring point as the output label, and uses the support vector machine regression algorithm for training and learning to construct a moisture diffusion model that predicts the moisture content of any point in the oil cavity. The moisture analysis processor is based on a moisture diffusion model. It divides the oil cavity space into a three-dimensional mesh and predicts the moisture content at each point within the mesh using the moisture diffusion model, thereby generating the three-dimensional moisture distribution map.

8. The real-time detection system for the grease chamber of the shield tail seal as described in claim 7, characterized in that, The process of establishing a three-dimensional coordinate system with the center of the bottom of the oil cavity as the origin, calculating the diffusion rate gradient at each monitoring point, and generating a water intrusion path map includes: The moisture analysis processor invokes the established three-dimensional coordinate system with the center of the bottom of the oil chamber as the origin; Based on the three-dimensional distribution map of moisture from multiple consecutive time series, the moisture content data of each monitoring point at different time points are extracted, and the rate of change of moisture content at each monitoring point per unit time, i.e., the diffusion rate, is calculated. By combining the spatial coordinates of each monitoring point and the corresponding diffusion rate, the diffusion rate gradient of each monitoring point on its spatial coordinates is calculated to obtain the gradient vector. Based on the gradient vectors of all monitoring points, a streamline tracing algorithm is used to simulate the direction and path of water movement, generating the water intrusion path map; the water intrusion path map indicates the trend and channels of water spreading from the bottom of the cavity upwards and around the cavity.

9. The real-time detection system for the grease chamber of a shield tail seal as described in claim 8, characterized in that, The step of marking potential leakage areas based on a water intrusion path map and a preset water rise rate threshold includes: The moisture analysis processor analyzes the generated moisture intrusion path map and identifies one or more core intrusion paths with the largest diffusion rate gradient modulus in the map. Along the identified core intrusion path, extract the diffusion rate data of each segment on the core intrusion path; The diffusion rate of each segment on the core intrusion path is compared with a preset moisture rise rate threshold. If the diffusion rate of any segment of the core intrusion path is greater than the preset moisture rise rate threshold, the area corresponding to the core intrusion path is marked as the potential leakage area. The moisture analysis processor outputs potential leakage area data, including the location, range, and maximum diffusion rate of the potential leakage area. It also extracts the real-time moisture content signal corresponding to the triggering of the potential leakage area marker and uses it as a moisture monitoring signal.

10. The real-time detection system for the grease chamber of a shield tail seal as described in claim 9, characterized in that, The risk assessment unit, based on model parameters trained using extracted dynamic features and historical fault data, outputs a leakage risk level after Bayesian network fusion analysis, including: The risk assessment unit acquires the real-time dynamic features extracted by the edge data processing unit, as well as the parameters of the Bayesian network model trained based on historical fault data; the Bayesian network model parameters include the conditional probability table of the Bayesian network. The risk assessment unit inputs real-time dynamic features as evidence into the corresponding input nodes of the Bayesian network model, and initiates the probabilistic reasoning process in conjunction with the built-in conditional probability table. The Bayesian network performs probabilistic inference based on the real-time dynamic features of the input and the conditional probability table, and updates the posterior probability of all nodes in the Bayesian network. The probability of each risk level corresponding to the output node of the Bayesian network is calculated based on the inference results, and the level with the highest probability is taken as the final output result; the level includes low risk, medium risk and high risk.

Images