Shield tunneling machine cutterhead mud cake monitoring system arrangement method, mud cake prediction method and shield tunneling machine

Abstract

The embodiment of the invention provides a shield cutter head mud cake monitoring system arrangement method, a mud cake prediction method and a shield tunneling machine, and relates to the technical field of shield tunnel construction monitoring and intelligent early warning. S2, sensor type selection and arrangement; s3, construction of a monitoring system: a sensing layer-transmission layer-processing layer three-level system is constructed, the sensing layer is the sensor, the transmission layer adopts a wireless + wired hybrid transmission mode, the processing layer comprises edge computing nodes and a cloud server, the edge nodes are responsible for data preprocessing, and the cloud server is responsible for data storage and prediction operation; the monitoring system is scientific in arrangement, multiple types of sensors are arranged in a partitioned mode based on the cutterhead mud cake accumulation rule, key signals are accurately captured through encryption arrangement of a first-stage monitoring area, multi-dimensional data complementation of a third-stage monitoring area is achieved, the data collection integrity is effectively improved, and the problems that multiple blind areas exist and data are scattered in traditional arrangement are solved.

Description

Technical Field

[0001] This application relates to the field of shield tunnel construction monitoring and intelligent early warning technology, and in particular to the layout method of shield cutterhead mud cake monitoring system, mud cake prediction method and shield machine. Background Technology

[0002] During tunnel boring machine (TBM) construction, when the cutterhead cuts through the soil, cohesive soil or water-rich soft soil adheres and accumulates on the cutterhead surface, cutterhead holder, and inner wall of the soil chamber, forming a "mud cake." This can lead to increased cutterhead torque, decreased tunneling efficiency, and accelerated equipment wear, and in severe cases, cause safety accidents such as cutterhead jamming and ground subsidence. Therefore, achieving early monitoring and accurate prediction of mud cake is one of the core requirements for intelligent TBM construction.

[0003] In existing technologies, mud cake monitoring often uses a single sensor (such as a temperature sensor or a torque sensor) with a single-point arrangement, which has the following drawbacks: (1) The monitoring dimension is singular and cannot capture the coordinated change pattern of "temperature-vibration-pressure-soil properties" during the mud cake formation process; (2) The sensor layout lacks specificity. The characteristics of mud cake accumulation vary greatly in different areas such as the edge and center of the cutterhead. A uniform layout leads to data redundancy or missing key information. (3) The prediction models are mostly based on traditional LSTM or SVM algorithms, which do not take into account the strong temporal and spatial correlation of construction data, and have poor robustness to interference signals such as geological changes and equipment vibrations, and the prediction lag is obvious (usually only 5-8 minutes in advance).

[0004] For example, Chinese patent CN114544229A discloses a method for monitoring mud cake on a tunnel boring machine cutterhead. It collects temperature data by arranging temperature sensors on the cutterhead and combines them with a BP neural network to predict mud cake. However, this method only relies on temperature parameters and does not consider key influencing factors such as cutterhead vibration and earth pressure, resulting in a prediction accuracy of less than 75%. At the same time, the sensors are arranged uniformly, but not densely in the edge area where the cutterhead cutting load is the greatest, which leads to untimely capture of the initial mud cake signal. Summary of the Invention

[0005] To address one of the aforementioned technical deficiencies, this application provides a method for arranging a shield cutterhead mud cake monitoring system, a mud cake prediction method, and a shield machine.

[0006] According to a first aspect of the embodiments of this application, a method for arranging a shield tunnel cutterhead mud cake monitoring system is provided, comprising: S1. Monitoring Area Division: Based on the shield cutterhead structure and mud cake accumulation pattern, the monitoring area is divided into a primary monitoring area, a secondary monitoring area, and a tertiary monitoring area. The primary monitoring area is the cutterhead edge area; the secondary monitoring area is the cutterhead transition area; and the tertiary monitoring area is the cutterhead center area. S2. Sensor selection and layout: In the primary monitoring area, a set of "vibration-temperature-stress" triple sensors are arranged at 15° intervals, and earth pressure sensors are densely arranged at the tool holder connection area; in the secondary monitoring area, "vibration-temperature" dual sensors are arranged at 30° intervals; in the tertiary monitoring area, laser displacement sensors, high-definition industrial cameras, LED supplementary lighting modules and moisture content sensors are arranged. S3. Monitoring System Construction: A three-level system of "perception layer - transmission layer - processing layer" is established. The perception layer is the aforementioned sensor, the transmission layer adopts a hybrid transmission method of "wireless + wired", and the processing layer includes edge computing nodes and cloud servers. The edge nodes are responsible for data preprocessing, and the cloud servers are responsible for data storage and predictive calculations.

[0007] Furthermore, the primary monitoring zone is 70%-100% of the cutterhead radius, the secondary monitoring zone is 30%-70% of the cutterhead radius, and the tertiary monitoring zone is 0%-30% of the cutterhead radius and the inner wall of the soil chamber.

[0008] Furthermore, the vibration sensor is an IEPE type accelerometer, and the temperature sensor is a PT1000 platinum resistance sensor.

[0009] Furthermore, the wireless transmission uses a 5G industrial module; the wired transmission uses industrial Ethernet.

[0010] According to a second aspect of the embodiments of this application, a method for predicting mud cake on a tunnel boring machine cutterhead is provided, comprising: A1. Multi-source data preprocessing: Sensor data is processed using an improved wavelet threshold denoising algorithm, data synchronization is achieved through timestamp alignment, and missing data is filled in using linear interpolation. A2. Feature Enhancement and Fusion: Extract statistical features from sensor data and image features from mud cake images, assign weights to the fused features through a channel-spatial dual attention mechanism, and output enhanced features; A3. Spatiotemporal Fusion Prediction: The enhanced features are input into the "CNN-LSTM-ELM" fusion model. The CNN extracts spatial correlation features, the improved LSTM extracts temporal correlation features, and the ELM optimized by the chaotic particle swarm algorithm outputs the preliminary prediction results. A4. Error Correction and Early Warning: A rolling window error correction mechanism is used to correct the prediction results, and graded early warning information is output based on the corrected probability of mud cake formation.

[0011] Furthermore, the improved wavelet thresholding denoising algorithm employs a continuously differentiable threshold function, using an exponential term to achieve a smooth transition at the threshold, thereby reducing signal distortion.

[0012] Furthermore, the specific strategy for feature enhancement and fusion in A2 is as follows: A21. Statistical Feature Extraction; A22. Image feature extraction; A23. Improved attention mechanisms.

[0013] Furthermore, the improved LSTM model enhances the ability to filter redundant temporal information by introducing a nonlinear transformation of the hidden layer from the previous time step to optimize the forget gate.

[0014] Furthermore, the chaotic particle swarm optimization algorithm enhances particle diversity through Logistic mapping and optimizes the input layer weights and biases of ELM.

[0015] Furthermore, the graded early warning includes Level 1 early warning: probability of mud cake formation 0.7-0.85, warning time 15-20 min; Level 2 early warning: probability 0.85-0.95, warning time 10-15 min; Level 3 early warning: probability ≥0.95, warning time 5-10 min.

[0016] According to a third aspect of the embodiments of this application, a water jet cleaning control method is provided, comprising: Based on the cutter head cleaning grid parameter set, the cutter head surface is cut into strips radially according to structural units and then subdivided according to tangential angles. The center point of the sub-region is corrected with the test frame positioning reference. Then, it is reordered and numbered according to the direction of the cutter head rotation area to obtain the cutter head cleaning grid parameter set. Based on the cutter head cleaning grid parameter set, the installation point and direction of the water jet nozzle are read, and the distance and direction between the center point of the sub-area and the pressure node of the test water supply pipeline are combined to form an energy value. Then, the effective coverage range is determined by a threshold to obtain the jet coverage energy distribution set. Based on the spray coverage energy distribution set, the Hungarian allocation algorithm is used to calculate the cleaning time according to the thickness of the sub-region adhesion layer and the pressure value. The time values ​​are formed into a matrix and multiple values ​​are compared in the same row to determine the nozzle correspondence. All matching items are combined into a unified sequence to obtain the nozzle cleaning assignment result set. Based on the nozzle cleaning assignment result set, the particle swarm optimization algorithm is used to add fine-tuning amount to the nozzle installation point, and then the distance is compared with the center point of the cutter head structure unit. The reachability is judged by the angle between the nozzle direction and the normal vector of the cleaning area. The reachable units are recorded to obtain the nozzle arrangement adjustment parameter set. Based on the nozzle arrangement adjustment parameter set, the nozzle start-up time is read according to the test, the assigned sub-region time is arranged into the time axis and aligned with the cutter head rotation area reference, and then integrated into a continuous time period and uniformly sorted to obtain the nozzle start-up and stop sequence table.

[0017] According to a fourth aspect of the embodiments of this application, a tunnel boring machine is provided, including: a control system, the control system being used to execute the method described above.

[0018] Compared with the prior art, the technical solution provided in this embodiment has the following advantages: The monitoring system is scientifically deployed: based on the accumulation pattern of the cutterhead cake, multiple types of sensors are arranged in zones. The primary monitoring zone is densely deployed to achieve accurate capture of key signals. The three-level monitoring zones provide multi-dimensional data complementarity, which effectively improves the integrity of data collection and solves the problems of "many blind spots and scattered data" in traditional deployment. The prediction model has high accuracy: improved wavelet threshold denoising reduces signal distortion, CSAM dual attention mechanism highlights key features, and chaotic optimization ELM solves the local optimum problem, thus improving the accuracy of mud cake prediction; the early warning time is advanced, allowing sufficient time for construction adjustments; Strong anti-interference capability: The multi-source data fusion and rolling error correction mechanism enable the model to control the fluctuation of prediction error to a certain extent under complex working conditions such as geological changes (such as alternation of clay and sand layers) and equipment vibration, thus improving the anti-interference capability of the existing model. Highly practical: The monitoring system adopts "wireless + wired" transmission to adapt to the shield tunneling rotation condition. The prediction model supports real-time output of edge computing and can be directly integrated into the shield tunneling construction control system to reduce the cost of mud cake disposal. Attached Figure Description

[0019] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart illustrating the arrangement method of the shield cutterhead mud cake monitoring system of this application; Figure 2 This is a flowchart illustrating the shield cutterhead mud cake prediction method of this application; Figure 3 This is a flowchart illustrating the specific strategy for feature enhancement and fusion in the shield cutterhead mud cake prediction method of this application; Figure 4 This is a schematic diagram of the structure of the tunnel boring machine cutterhead provided in an embodiment of this application; Figure 5 This is a structural schematic diagram of the central region of the tunnel boring machine cutterhead provided in an embodiment of this application; Figure 6 This is a partial sectional view of the cutterhead of a tunnel boring machine provided in an embodiment of this application; Figure 7 This is a schematic diagram of the tunnel boring machine provided in an embodiment of this application. Detailed Implementation

[0020] To make the technical solutions and advantages of the embodiments of this application clearer, the exemplary embodiments of this application will be described in further detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not an exhaustive list of all embodiments. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other.

[0021] like Figure 1 As shown, the method for arranging the shield tunnel cutterhead mud cake monitoring system includes the following steps: S1. Monitoring Area Division: Based on the shield cutterhead structure and mud cake accumulation pattern, the monitoring area is divided into a primary monitoring area, a secondary monitoring area, and a tertiary monitoring area. The primary monitoring area is the cutterhead edge area; the secondary monitoring area is the cutterhead transition area; and the tertiary monitoring area is the cutterhead center area. The primary monitoring area is the range of 70%-100% of the cutterhead radius, where the cutting linear velocity is the highest and the mud cake accumulation rate is the fastest. The secondary monitoring area is the range of 30%-70% of the cutterhead radius, and the tertiary monitoring area is the range of 0%-30% of the cutterhead radius and the inner wall of the soil chamber. S2. Sensor Selection and Layout: In the primary monitoring zone, a set of three-unit "vibration-temperature-stress" sensors is arranged at 15° intervals. Earth pressure sensors are densely arranged at the cutterhead connection. The vibration sensor is an IEPE type accelerometer, and the temperature sensor is a PT1000 platinum resistance sensor. Each set includes one IEPE type accelerometer, one PT1000 platinum resistance temperature sensor, and one strain gauge stress sensor. At the same time, two sets of earth pressure sensors are densely arranged at the connection between the cutterhead and the cutter head to capture pressure changes caused by mud cake. In the secondary monitoring zone, a "vibration-temperature" dual-unit sensor is arranged at 30° intervals to reduce the redundancy of stress sensors. In the tertiary monitoring zone, a laser displacement sensor, a high-definition industrial camera, an LED supplementary lighting module, and a moisture content sensor are arranged. One laser displacement sensor is arranged at the center of the cutter head to monitor the thickness of mud cake accumulation in the central area. Three high-definition industrial cameras are evenly arranged along the circumference of the inner wall of the soil chamber, and the LED supplementary lighting module is used to acquire mud cake images. One moisture content sensor is arranged at the inlet of the screw conveyor to obtain soil physical property parameters. S3. Monitoring System Construction: A three-level system of "perception layer - transmission layer - processing layer" is established. The perception layer is the aforementioned sensor, the transmission layer adopts a hybrid transmission method of "wireless + wired", and the processing layer includes edge computing nodes and cloud servers. The edge nodes are responsible for data preprocessing, and the cloud servers are responsible for data storage and predictive calculations.

[0022] Wireless transmission uses 5G industrial modules; wired transmission uses industrial Ethernet.

[0023] like Figure 2 and Figure 3As shown, this embodiment also provides a method for predicting mud cake on a tunnel boring machine cutterhead, including the following steps: A1. Multi-source data preprocessing: Sensor data is processed using an improved wavelet threshold denoising algorithm, data synchronization is achieved through timestamp alignment, and missing data is filled in using linear interpolation. The improved wavelet thresholding denoising algorithm employs a continuously differentiable threshold function, using an exponential term to achieve a smooth transition at the threshold, thereby reducing signal distortion. The specific strategy is as follows: Improved Wavelet Thresholding Denoising: Traditional wavelet thresholding functions suffer from distortion at signal abrupt changes. This invention improves the hard thresholding function into a continuously differentiable thresholding function, as shown in the following formula: , in, The coefficients are those obtained after wavelet decomposition; These are the denoised coefficients; Threshold and , For data length, The standard deviation of noise. ; This is an adjustment factor (ranging from 0.8 to 1.2, adaptively adjusted according to data noise intensity). This improved function achieves a smooth transition at the threshold through an exponential term, reducing signal distortion. Data synchronization and completion: Multi-sensor data synchronization is achieved using timestamp alignment, with time error controlled within 50ms; missing data is completed using linear interpolation, and a backup sensor data switching mechanism is triggered when the missing rate exceeds 10%.

[0024] A2. Feature Enhancement and Fusion: Extract statistical features from sensor data and image features from mud cake images, assign weights to the fused features through a channel-spatial dual attention mechanism, and output enhanced features; The specific strategies for feature enhancement and fusion in A2 are as follows: A21. Statistical Feature Extraction: Extract time-domain features (mean, variance, peak value, kurtosis) and frequency-domain features (dominant frequency, spectral energy) from sensor time-series data (vibration, temperature, pressure, etc.), for a total of 12 statistical features; extract 2-dimensional features, namely rate of change and cumulative deviation, from moisture content data; A22. Image Feature Extraction: An improved CNN model (with a residual module) is used to extract features from the mud pie image. The model structure includes three convolutional layers (kernel sizes of 3×3, 3×3, and 5×5, all with a stride of 1), two pooling layers (max pooling, with a 2×2 kernel), and one fully connected layer, outputting 64-dimensional image features. The introduction of the residual module solves the gradient vanishing problem in deep CNNs, as shown in the following formula: ,in: For the input feature map, , For convolution weights, , For bias terms, For ReLU activation function, Output for the residual module; A23. Improved Attention Mechanism Enhancement: A channel-spatial dual attention mechanism (CSAM) is introduced to assign weights to the fused statistical features and image features, highlighting key features of mud cake formation (such as the vibration peak at the cutterhead edge, temperature abrupt change rate, and mud cake grayscale features in the image). The channel attention weights are calculated as follows: ; Spatial attention weights are calculated as follows: ; in, As a feature of fusion, For global average pooling, It is a two-dimensional convolution. This is element-wise multiplication. , The output features are channel and spatial attention features, respectively, and the final output is an 80-dimensional enhanced feature.

[0025] A3. Spatiotemporal Fusion Prediction: The enhanced features are input into the "CNN-LSTM-ELM" fusion model. CNN extracts spatial correlation features, LSTM is improved to extract temporal correlation features, ELM optimizes the accuracy of the output layer, and the ELM optimized by the chaotic particle swarm algorithm outputs the preliminary prediction results. The improved LSTM model optimizes the forget gate by introducing a nonlinear transformation of the hidden layer from the previous time step, thereby enhancing the ability to filter redundant temporal information. The chaotic particle swarm optimization algorithm enhances particle diversity through logistic mapping and optimizes the input layer weights and biases of ELM. Specifically: CNN spatial feature extraction: The 80-dimensional enhanced features are reshaped into an 8×10 feature matrix, which is then input into two convolutional layers (3×3 kernels, stride 1) and outputs a 32-dimensional spatial feature vector. LSTM temporal feature extraction: 32-dimensional spatial features are input into an improved LSTM model according to the time series (time window size is 20, i.e., the 21st data group is predicted from the first 20 data groups). A forget gate optimization mechanism is introduced, and the forget gate output formula is as follows: ,in, This is the output of the hidden layer from the previous time step. Input for the current time. , This is the weight matrix. For bias terms, This is an adjustment coefficient (with a value of 0.1-0.3). The sigmoid activation function is used. This improvement enhances the forget gate's ability to filter redundant temporal information by introducing a nonlinear transformation of the hidden layer from the previous time step. The LSTM outputs a 16-dimensional temporal feature vector. ELM Output Optimization: The 16-dimensional temporal features are input into the Extreme Learning Machine (ELM). The input layer weights and biases of the ELM are optimized using the Chaotic Particle Swarm Optimization (CPSO) algorithm to avoid local optima caused by random initialization. The fitness function of CPSO is the sum of squared prediction errors of the ELM. Particle diversity is enhanced through a chaotic mapping (Logistic mapping). The formula for the Logistic mapping is as follows: ,in, For the first Substitute chaotic variables, For control parameters, when When the system is in a completely chaotic state, the optimized ELM outputs the probability of cake formation P (0≤P≤1) and the prediction time T (unit: min).

[0026] A4. Error Correction and Early Warning: A rolling window error correction mechanism is used to correct the prediction results, and graded early warning information is output based on the corrected probability of mud cake formation.

[0027] A rolling window error correction mechanism is adopted, using the errors of the previous 5 predictions as feedback to correct the current prediction result. The correction formula is as follows: , ,in, , For correction factor, , The first The probability and time of actual mud cake formation; when When, trigger an alert, according to Values ​​are used for tiered early warning.

[0028] The tiered early warning system includes: Level 1: probability of mud cake formation 0.7-0.85, warning time 15-20 min; Level 2: probability 0.85-0.95, warning time 10-15 min; Level 3: probability ≥0.95, warning time 5-10 min.

[0029] Based on the above technical solutions, after setting and predicting the mud cake layout, the following water jet cleaning control method for the tunnel boring machine cutterhead can also be adopted, including the following steps: S1: Based on the cutter head cleaning grid parameter set, the cutter head surface is cut into strips radially according to the structural unit and then subdivided according to the tangential angle. The center point of the sub-region is corrected with the test frame positioning reference. Then, it is reordered and numbered according to the direction of the cutter head rotation area to obtain the cutter head cleaning grid parameter set. S2: Based on the cutter head cleaning grid parameter set, read the water jet nozzle installation point and direction, synthesize the distance and direction between the sub-area center point and the water supply pipeline pressure node into an energy value, and then determine the effective coverage range with a threshold to obtain the jet coverage energy distribution set; S3: Based on the spray coverage energy distribution set, the Hungarian allocation algorithm is used to calculate the cleaning time according to the thickness of the sub-region adhesion layer and the pressure value. The time quantities are formed into a matrix and multiple values ​​are compared in the same row to determine the nozzle correspondence. All matching items are combined into a unified sequence to obtain the nozzle cleaning assignment result set. S4: Based on the nozzle cleaning assignment result set, the particle swarm optimization algorithm is used to add fine-tuning amount to the nozzle installation point, and then compare the distance with the center point of the cutter head structure unit. The reachability is judged by the angle between the nozzle direction and the normal vector of the cleaning area. The reachable units are recorded to obtain the nozzle arrangement adjustment parameter set. S5: Based on the nozzle arrangement adjustment parameter set, read the nozzle start time according to the test, arrange the assigned sub-area time into the time axis and align it with the cutter head rotation area reference, then integrate it into a continuous time period and organize it uniformly to obtain the nozzle start-stop timing table.

[0030] The cutterhead cleaning grid parameter set includes sub-region number, sub-region coordinate points, and sub-region position sequence; the spray coverage energy distribution set includes sub-region energy value, coverage mark, and direction superposition value; the nozzle cleaning assignment result set includes nozzle number, sub-region number, and cleaning time; the nozzle arrangement adjustment parameter set includes position adjustment amount, direction correction amount, and reachability mark; and the nozzle start-stop sequence table includes start time, stop time, and nozzle sequence number.

[0031] The specific steps for generating the tool shroud cleaning mesh parameter set are as follows: Based on the cutter head cleaning grid parameter set, the cutter head surface is divided into several radial strips according to the change from the outer edge to the center. Then, tangential partitions are drawn according to a fixed angle. The actual position of the center point of each sub-region is re-determined using the test frame positioning point to generate a sub-region positioning dataset. Based on the sub-region positioning dataset, the numbering order of all sub-regions is rearranged according to the direction of the tool head rotation area. The rearranged numbers are arranged in a matrix to form a unified index, thus obtaining the tool head cleaning grid parameter set. Based on the cutter head cleaning grid parameter set, a radial slicing numerical partitioning algorithm is used to continuously take points along a fixed step distance from the outer edge of the cutter head to the center, and the areas between each point are sequentially marked as radial strips. Then, on each radial strip, tangential regions are generated by rotating segment by segment with a fixed angular step distance, and the center point angle of each tangential region is recorded. Then, the three-axis coordinates of the test frame positioning points are used to perform three-way translation on each center point, and the displacement is superimposed sequentially to form the corrected center point position. All corrected positions are arranged in the generation order to generate a sub-region positioning dataset. Based on the sub-region positioning dataset, the rotation zone numbering rearrangement algorithm is used to perform angle conversion on the center point coordinates of each sub-region, and the converted angles are arranged in order according to the starting direction of the cutter head rotation zone. The arranged sub-regions are then renumbered. Then, a numbering matrix is ​​established with the number of radial strips as the row structure basis and the number of tangential partitions as the column structure basis. The corresponding numbers are filled into the matrix positions one by one, and all row and column information is sorted to obtain the cutter head cleaning mesh parameter set.

[0032] The specific steps for generating the jet coverage energy distribution set are as follows: Based on the cutterhead cleaning grid parameter set, the installation point position and spray direction of the water jet nozzle are extracted, and then the energy quantity is combined with the straight distance and direction between the center point of each sub-area and the pressure node of the water supply pipeline, and recorded as the same sequence to generate the energy base set. Based on the energy base set, the energy in the sequence is compared with a set threshold in turn. Sub-regions that reach the threshold are included in the effective coverage range. The coverage results are then classified into separate sequences according to the numbering order to obtain the jet coverage energy distribution set. Based on the cutterhead cleaning mesh parameter set, an energy combination operation algorithm is used to execute a three-axis coordinate value retrieval command on the water jet nozzle installation point position and a direction vector analysis command on the spray direction. Then, a three-axis difference command is executed on the coordinates of the center point of each sub-zone and the coordinates of the pressure node of the water supply pipeline. The difference components are combined into a distance vector in a fixed order, and the distance vector length is used as the distance quantity. The distance quantity and the spray direction vector are extracted one by one. After extraction, a multiplication and addition command is executed with the distance quantity as the first factor and the angle quantity as the second factor to form the energy quantity value of a single sub-zone. Then, all energy quantity values ​​are added to the sequence buffer in the order of sub-zone number, and a one-time read command is executed on the buffer to form a continuous energy quantity sequence, generating an energy base set. Based on the energy base set, the energy threshold comparison algorithm is used to perform the comparison command on the energy in the sequence in numerical order, and a fixed threshold is used as the comparison benchmark. For the energy greater than the threshold, the valid identifier is written and the corresponding sub-region number is written to the coverage buffer. Then, the contents of the buffer are rearranged in ascending order of sub-region number. After rearrangement, all valid numbers are written into the coverage record sequence in a single column structure to form an independent data group, thus obtaining the jet coverage energy distribution set.

[0033] The specific steps for generating the nozzle cleaning assignment result set are as follows: Based on the energy distribution set of the spray coverage, the Hungarian allocation algorithm is used to extract the thickness of the attachment layer of each sub-region from the record table one by one, and the pressure value under the same number is sequentially substituted into the thickness amount to convert it into time amount. Then, the converted time amount is arranged into a continuous sequence according to the order of the sub-region to generate a time amount sequence set. Based on the time sequence set, the sequence is divided into multiple rows according to the number of nozzles. Then, the position of the identifier is determined in each row according to the time magnitude. These identifiers are grouped together according to the nozzle arrangement order to generate the nozzle corresponding identifier set. Based on the nozzle corresponding identifier set, the nozzle number and sub-area number are retrieved in pairs according to the actual landing point of the identifier. Then, the corresponding time quantity is inserted into the same sequence to form a continuous record. The records are then merged into the set in the overall order to obtain the nozzle cleaning assignment result set. Based on the spray coverage energy distribution set, the Hungarian allocation algorithm is used to execute a read command for the thickness of the adhesion layer in each sub-region of the record table according to the sub-region number. The read thickness is written into the thickness sequence in numerical order. Then, the pressure value retrieval command is called to write the pressure values ​​with the same number into the pressure sequence in sequence. Next, using the thickness sequence as input, a conversion command is executed item by item according to the number. The corresponding value in the pressure sequence is used as the conversion factor to form the time quantity, and the time quantity is written into the time block buffer in an indexed manner. Then, in the Hungarian allocation algorithm, a row minimum value subtraction command is executed on the time block buffer. The minimum value is subtracted from each item in the corresponding row to form a row difference matrix, and then the process is continued. The minimum value subtraction command is used to subtract the minimum value from each item in the corresponding column to form a column difference matrix. The position of zero value is found in the difference matrix and written into the zero value mark list according to the row and column of the zero value. Then, the zero value mark list is covered by the cover command, and the cover index is recorded in the row direction and column direction respectively. If the coverage is insufficient, the minimum difference is recorded in the uncovered position and the minimum difference subtraction and cover position addition commands are executed to update the difference matrix. Then, the zero value search and row and column coverage actions are repeated until the coverage meets the assignment scale. The row and column numbers in the difference matrix are written into the index sequence according to the zero value distribution to generate the time quantity sequence set. Based on the time sequence set, the Hungarian allocation algorithm is used to split the sequence according to the number of nozzles, and the generated row structure is sorted according to the time size. The original index of each item during the sorting process is written into the identifier buffer. Then, the position registration command is executed on the identifier buffer row by row, and the identifiers in each row are merged into identifier groups according to the nozzle number order. After merging, all identifier groups are written into the identifier set in order, and an identifier mapping table is constructed. After the entire mapping table is written, the identifier corresponding sequence is formed, and the nozzle corresponding identifier set is generated. Based on the nozzle corresponding identifier set, the Hungarian allocation algorithm is used to execute the read command on the row and column numbers in the identifier, retrieve the corresponding nozzle number according to the read number, retrieve the corresponding sub-area number according to the same number, execute the pairing command after retrieval, write the pairing number into the pairing list, execute the insertion command on the time quantity with the same row number in the pairing list, and form a continuous time record sequence after insertion. Then, all record sequences are merged into a unified set in sequence, and the set alignment command is executed to form the final pairing result group, thus obtaining the nozzle cleaning assignment result set.

[0034] The Hungarian allocation algorithm first extracts the valley value of each row of the cleaning time quantity arranged in matrix form and subtracts it from the corresponding items in the row, so that each row forms a difference distribution with zero as the base. Then, it extracts the valley value of each column of the processed matrix and subtracts it from the corresponding items in the column, so that the values ​​in the column synchronously form a new difference matrix. Then, it searches for all positions with zero values ​​in the matrix and covers the rows and columns containing zero values ​​in a straight line. The covered rows and columns are recorded as a cover set. When the number of rows and columns in the cover set is insufficient to cover all zero values, the valley value is extracted from the uncovered positions in the matrix and subtracted from all uncovered positions. At the same time, it is added to all covered positions to form a new difference matrix. The zero value search and row and column covering actions are repeated on the new difference matrix. When the number of rows and columns in the cover set reaches the number of nozzles, the uncovered zero value positions are collected into pairs according to the row and column coordinates and the pairs are used as the pairing indexes corresponding to the nozzles and sub-regions. The Hungarian allocation algorithm, according to the formula:

[0035] in: Indicates the number is The thickness of the target adhesion layer in the blade plate area. Indicates the number is The water jet injection pressure value in the cutter head area, This represents the water jet deposition removal efficiency coefficient. Indicates the number is The duration of water jet action in the blade plate area Indicates the number is The uniformity index of water jet energy coverage in the blade plate area. Indicates the number is The overlap rate index of water jet trajectory coverage in the blade plate area. Indicates the number is The nozzle wear index in the cutter head area. This represents the weighting coefficient for energy coverage uniformity correction. This represents the weighting coefficient for trajectory coverage overlap correction. This indicates the nozzle wear correction weighting factor; Execution process: First, read the target adhesion layer thickness of the sub-region numbered in the order of sub-region numbering on the surface of the cutter head. And retrieve the water jet injection pressure value that matches the sub-zone number from the injection control unit. The system then retrieves the water jet deposition removal efficiency coefficient, which has been determined experimentally, from the calibration parameter set. Then, based on the water jet energy distribution sampling results, the number is calculated as follows: Subregion energy coverage uniformity index And calculate the number based on the water jet trajectory generation model. Sub-region trajectory coverage overlap index At the same time, the number is calculated by accumulating the nozzle operation records. Sub-region nozzle wear index The system then retrieves the energy coverage uniformity correction weighting coefficients obtained through least-squares fitting from the controller parameter area. And trajectory coverage overlap rate correction weight coefficient and nozzle wear correction weighting factor The system will , , , , Substituting all the values ​​into the formula, and applying a joint correction factor that includes three correction factors—energy uniformity, trajectory overlap rate, and nozzle wear—to the original time calculation results, we obtain the result numbered as follows: Water jet action time in the blade plate area Finally, the above process is executed sequentially on all cutter disc areas according to their numbering order to form a complete set of water jet action time sequence.

[0036] The specific steps for generating the nozzle arrangement adjustment parameter set are as follows: Based on the nozzle cleaning assignment result set, the particle swarm optimization algorithm is used to move the coordinates of the nozzle installation point by a certain distance in a fixed direction to form a new set of coordinates. Then, the corresponding nozzle number is appended to these new coordinates to make them a list that can be directly indexed, thus generating a set of nozzle adjustment coordinates. Based on the nozzle adjustment coordinate set, the adjusted coordinates of each group are compared with the coordinates of the center point of the cutter head structure unit one by one, the distance between the two points is measured, and the distances are arranged into a record column according to the nozzle number to generate a nozzle distance comparison set. Based on the nozzle distance reference set, the nozzle direction and the normal vector of the cleaning area are taken out in numerical order, the angle between the directions is calculated, and the angle and distance are recorded together as a judgment item. Finally, the numbers that meet the attainable requirements are collected into a list to obtain the nozzle layout adjustment parameter set. Based on the nozzle cleaning assignment result set, the particle swarm optimization algorithm is used to execute the position initialization command for the nozzle installation point coordinates. The three-axis coordinates of each nozzle are written into the particle position vector, the corresponding nozzle number is written into the index vector, the three-axis components of the velocity vector are initialized to zero, and after initialization, the direction offset command is executed for each position vector. The offset distance parameter is used as the three-axis increment to write the offset coordinates into the candidate coordinate list. Then, the nozzle number is appended to the candidate coordinate list to form an indexable record, and the record is written into the particle current coordinate set. After the coordinate set is established, the velocity update command is executed for each particle. The inertia factor is multiplied by the three-axis components of the previous velocity vector, the difference component between the particle's current coordinate and the particle's historical coordinate is multiplied by coefficient one and added to the velocity vector, the difference component between the particle's current coordinate and the group's historical coordinate is multiplied by coefficient two and added to the velocity vector, and the updated velocity vector is added to the current position component by component to obtain the new position coordinates. The new position is written into the coordinate set and the update steps are repeated until the iteration number requirement is reached to generate the nozzle adjustment coordinate set. Based on the nozzle adjustment coordinate set, the particle swarm optimization algorithm is used to execute the coordinate reading command for each set of adjustment coordinates according to the nozzle number. The three-axis components of the adjustment coordinates are calculated item by item with the three-axis components of the center point coordinates of the cutter head structure unit. The difference components are written into the difference buffer. Then, the three-axis quantity accumulation command is executed on the difference buffer to form the distance quantity. The distance quantity is written into the distance record column in the order of nozzle number. During the writing process, the number and the distance quantity are kept in a one-to-one correspondence to form a distance sequence. After all the writing is completed, the distance sequence is used as the input data group of the particle swarm optimization algorithm fitness value to generate the nozzle distance comparison set. Based on the nozzle distance reference set, the particle swarm optimization algorithm is used to execute the direction reading command according to the nozzle number. The three-axis components of the direction vector of each nozzle are written into the direction buffer, and the three-axis components of the normal vector of the cleaning zone are written into the normal buffer. Then, the multiply-add command is executed on the two sets of vectors to form the angle combination value. The combination value is input into the direction difference command to obtain the direction angle. The angle and distance are written into the decision sequence in parallel according to the number. After the decision sequence is written, the number filtering command is executed on the sequence. The numbers that meet the attainable requirements are written into the output list one by one to obtain the nozzle layout adjustment parameter set.

[0037] The particle swarm optimization algorithm first randomly generates position vectors and sets corresponding velocity vectors for each nozzle arrangement scheme within the search space. The position vectors are recorded as the initial coordinate set of the particles, and the velocity vectors as the initial velocity set. Then, based on data such as nozzle installation point coordinates, injection direction angle, cutter head mesh cell coverage, and cleaning time, the cost value of each particle is calculated. This cost value is compared with the particle's historical cost to update the particle's historical adaptive position set. Simultaneously, the particle with the lowest cost value is selected from all particles and recorded as the swarm's adaptive position set. Finally, for each particle, the inertial weight is multiplied by the previous generation's velocity, and the particle's historical position and current position are added. The velocity vector is updated by weighting the difference between the current position and the difference between the current position and the group adaptation position. The updated velocity vector is then added to the current position coordinates dimension by dimension to obtain the new position coordinates. Coordinates that are outside the installation range are truncated by boundary values ​​or corrected by reverse foldback to form a new generation of particle position set. The cost value corresponding to the new position is recalculated and the particle history adaptation position set and the group adaptation position set are updated according to the same rules. The velocity update, position update and cost calculation steps are repeated within the preset number of iterations. When the iteration ends, the nozzle coordinates and injection direction angle in the group adaptation position set are recorded as the output results of the particle swarm optimization algorithm. Particle swarm optimization algorithm, according to the formula:

[0038] in: Indicates the first The iteration number is The nozzle particle velocity vector Indicates the first The iteration number is The nozzle particle velocity vector Indicates the inertia weighting coefficient. Represents individual learning factors. Represents a random number in the interval between zero and one. Indicates the first The iteration number is The individual historical optimal position vector of the nozzle particle. Indicates the first The iteration number is The current position vector of the nozzle particle. This represents the radial position correction weighting coefficient. Indicates the number is The nozzle radial position index, This indicates the tool occlusion correction weight coefficient. Indicates the number is The nozzle-cutter obstruction impact index, Represents the group learning factor. Represents a random number in the interval between zero and one. Indicates the first The global optimal position vector of the nozzle particle swarm in the next iteration. This represents the weighting factor for correcting historical scour intensity deviation. Indicates the number is The nozzle's historical erosion intensity deviation index. This represents the weighting coefficient for correcting jet crosstalk. Indicates the number is The nozzle jet cross-interference index; Execution process: First, initial coordinates are set for each of the multiple nozzles and used as the first... The current position vector of the next iteration At the same time, set the initial velocity vector of the corresponding nozzle. And load the inertia weight coefficient Individual learning factors With group learning factor The controller then calculates the cleaning coverage evaluation value based on the current nozzle coordinates and determines the number as follows. The individual historical best position vector of the nozzle and the global optimal position vector corresponding to all nozzle combinations In the same iteration, the controller obtains the radial position index based on the ratio of the nozzle radial distance to the cutter head radius. The tool obstruction influence index is obtained based on the obstruction angle and minimum distance between the nozzle and the adjacent tool. The historical scouring intensity deviation index was obtained by normalizing the residual thickness deviation of the nozzles in previous cleaning tests. The jet cross-interference index is obtained based on the proportion of overlapping areas between the water jet trajectories of the nozzles. Simultaneously, the radial position correction weighting coefficient is read from the calibration parameter set. Tool occlusion correction weight coefficient Historical scour intensity deviation correction weighting coefficient and the weighting coefficient for jet cross-interference correction The controller then generates a random number for each nozzle. and And and Substituting the individual terms into the formula, , , , , and Substitute the group terms into the formula, and then combine them with the inertia term. Add them together to get the speed update result. Then according to The displacement is decomposed into a unit vector along a fixed direction and superimposed with the current coordinates to form the nozzle coordinates for the next iteration. The controller repeats the above speed update and position update process for all nozzles until the iteration count and fitness convergence conditions are met. Finally, the multiple nozzle coordinates at the termination iteration time are used as the nozzle adjustment coordinate set for the cutterhead water jet cleaning path planning and execution.

[0039] The specific steps for generating the nozzle start-stop timing table are as follows: Based on the nozzle arrangement adjustment parameter set, the nozzle start time is extracted according to the original test record, and the time corresponding to the assigned sub-area is added to the same time axis according to the start order, thereby forming a continuous time arrangement sequence and generating a time axis arrangement set. Based on the time axis arrangement set, the position of the sequence is checked segment by segment according to the time reference of the cutter head rotation area, and the continuous time periods are re-connected into a unified record format according to the nozzle number to obtain the nozzle start and stop sequence table. Based on the nozzle arrangement adjustment parameter set, a time series construction algorithm is used to execute a read command for each nozzle start time in the original test record. The read time values ​​are written into the start buffer according to the nozzle number. Then, the time quantity corresponding to the assigned sub-area is executed with a value retrieval command and written into the sub-area time buffer in numerical order. Then, the time axis establishment command is executed with the time in the start buffer as the reference and the minimum start value is used as the start point of the time axis. The time superposition command is executed for the time quantity of each nozzle. The superimposed time points are written into a unified time axis list in sequence. During the writing process, the continuity marking command is executed on the time axis list. The interval between adjacent time points is written into the interval sequence with a fixed step size. Then, the time points and the interval sequence are combined and written into the arrangement buffer to form a continuous time arrangement sequence. The numbering and labeling command is executed on the arrangement sequence to generate a time axis arrangement set. Based on the time axis arrangement set, a time calibration and reconstruction algorithm is used to execute a read command for each time period in the sequence and use the time reference value set in the cutter head rotation area as the calibration reference for the read time period. The difference between the start time of each time period and the reference value is calculated and the difference is written into the offset list as the calibration offset. The offset list is then applied to all time periods in sequence to execute calibration commands and the calibrated time periods are written into the reconstruction sequence. After the reconstruction sequence is established, a serial command is executed according to the nozzle number to merge all time periods of the same nozzle number into a single continuous structure in sequence. The start time, end time, and nozzle number in each structure are written into a data column of a unified record format in a fixed field order and all data columns are written into the output set in sequence to obtain the nozzle start and stop timing table.

[0040] A unified recording format is used to create a collection of time segment records for each nozzle, arranged in a fixed field order, including the start time, end time, sub-area number, and nozzle number. Each record is based on a continuous time period and forms an independent entry according to the nozzle number. The time segments within each entry are arranged in order of start time first and end time last. All entries form a data group with the same field structure, enabling the time segments of different nozzles to be read and compared in the same set according to the fixed field positions.

[0041] Based on the above solution, this embodiment also provides a tunnel boring machine, including: a control system, which is used to execute the method described above.

[0042] The tunnel boring machine also includes: the cutterhead and casing. Among them, such as... Figure 4 and Figure 5 As shown, the tunnel boring machine cutterhead includes: cutterhead body 1, first water jet nozzle 13 and second water jet nozzle 14.

[0043] The outer end face of the cutter head body 1 is divided into a central region 11 and a peripheral region, and multiple cutter heads 12 are provided in both the central region and the peripheral region. The cutter heads in the central region are irregularly positioned, while the cutter heads in the peripheral region are arranged radially outward from the central region.

[0044] In this embodiment, six rows of cutter heads are arranged outward from the central region. Each row of cutter heads extends radially outward along the cutter head body 1 and is spaced apart. The six rows of cutter heads are evenly distributed circumferentially. Each cutter head 12 includes two coaxially arranged cutter rings, which are axially symmetrical.

[0045] The first water jet nozzle 13 is located at the intersection of the extension line of the axis of symmetry between the two cutter rings in one cutter head 12 and the extension line of the axis of symmetry between the two cutter rings in an adjacent cutter head. Specifically, as follows... Figure 5 As shown, taking the cutter head 12 located at the lower left and the cutter head 12 located at the upper left as examples: the first water jet nozzle 13 is set at the intersection of the extension line L of the two cutter ring symmetry axes in the lower left cutter head 12 and the extension line L of the two cutter ring symmetry axes in the upper left cutter head 12. Similarly, a similar layout is adopted in the remaining cutter heads.

[0046] The second water jet nozzle 14 is located at the intersection of the line connecting the center of one cutter ring and the edge of the other cutter ring in a cutter head 12, and the extended line connecting the center of one cutter ring and the edge of the other cutter ring in a neighboring cutter head. Specifically, as follows... Figure 5As shown, taking the cutter head 12 located at the lower left and the cutter head 12 located at the upper left as examples: In the cutter head 12 located at the lower left, a second water jet nozzle 14 is installed at the intersection of the extension line K of the line connecting the center of one cutter ring to the edge of the other cutter ring and the extension line K of the line connecting the center of one cutter ring to the edge of the other cutter ring in the cutter head 12 located at the upper left. Similarly, the other two adjacent cutter heads 12 adopt a similar scheme.

[0047] The cutter head 12 includes two types. One type has a gap between two cutter rings, in which case the extensions L and K of the two cutter rings do not coincide, and they form an acute angle. Figure 5 The blade 12 is located in the upper left and lower left corners.

[0048] Another scenario is where there is no gap or a very small gap between the two cutter rings in cutter head 12. In this case, the extension lines L and K of the two cutter rings are considered to coincide. Figure 5 The cutter head 12 is located at the top and the cutter head 12 is located on the right. In this design, for the top cutter head 12, the intersection of its overlapping extension line L (K) with the extension line L of the right cutter head 12 is provided with a first water jet nozzle 13, and the intersection of its extension line K with the extension line K of the upper left cutter head 12 is provided with a second water jet nozzle 14.

[0049] In the two cutter heads 12, the intersection of the extended line L of the cutter ring symmetry axis and the intersection of the extended line K of the line connecting the center of one cutter ring and the edge of the other cutter ring are prone to the accumulation and deposition of slag and soil, which are high-incidence areas for mud cake formation. By setting the first water jet nozzle 13 and the second water jet nozzle 14 at these two locations, high-pressure water jets can be used to flush away the mud cake as soon as it begins to form, effectively preventing further accumulation of mud cake.

[0050] Based on the above scheme, in this embodiment, the first water jet nozzle 13 and the second water jet nozzle 14 are set according to the position of the cutter head 12, which can accurately match the formation pattern of mud cake near each cutter head 12. Compared with the scheme of uniformly arranging water jet nozzles in the prior art, the high-pressure water jets sprayed by the first water jet nozzle 13 and the second water jet nozzle 14 in this embodiment can remove mud cake more efficiently, reduce the obstruction of mud cake to the normal operation of the cutter head body 1 and the cutter head 12, significantly improve the mud cake removal efficiency and effect, and ensure the smoothness and reliability of cutting the soil.

[0051] The technical solution provided in this embodiment includes a cutter head body with multiple cutter heads in the central region of its outer end face. Each cutter head includes two axially symmetrically arranged cutter rings. A first water jet nozzle is located at the intersection of the extension line of the axis of symmetry between the two cutter rings in one cutter head and the extension line of the axis of symmetry between the two cutter rings in an adjacent cutter head. A second water jet nozzle is located at the intersection of the line connecting the center of one cutter ring and the edge of another cutter ring in one cutter head and the extension line connecting the center of one cutter ring and the edge of another cutter ring in an adjacent cutter head. This solution can remove mud cake more efficiently, reduce the obstruction of mud cake to the normal operation of the cutter head body and cutter heads, significantly improve the efficiency and effect of mud cake removal, and ensure the smoothness and reliability of cutting the soil.

[0052] Based on the above technical solution, a third water jet nozzle 15 is also adopted. An outer tangent circle C is drawn with the outer peripheral edge of a cutter head 12, and the outer tangent circle C is internally tangent to the circular envelope of the central region 11 of the cutter head body. If there is no first water jet nozzle 13 or second water jet nozzle 14 within the outer tangent circle C, the third water jet nozzle 15 is disposed at the center of the outer tangent circle C.

[0053] The setting can be based on the position of the cutter head 12 in the central region 11. Figure 5 Taking the cutter head layout shown as an example, the lower right blank area is relatively large. The outer periphery of the cutter head 12 is circumscribed by a circle C, and the circumscribed circle C is internally tangent to the circular envelope of the central area 11 of the cutter head body. Figure 5 A third water jet nozzle 15 can be installed at the center of each of the three circumscribed circles C.

[0054] The third water jet nozzle 15 effectively fills the gap in mud cake prevention that may exist after the arrangement of the first two types of water jet nozzles. It can assist the first water jet nozzle 13 and the second water jet nozzle 14 in providing all-round, no-dead-angle mud cake prevention coverage to the central area, avoiding the local accumulation of mud cake on the central area 11, which would cause the center of gravity of the cutter head body 1 to shift, and improving the stability and safety of the cutter head body 1 when it rotates.

[0055] The aforementioned first water jet nozzle 13, second water jet nozzle 14, and third water jet nozzle 15 can all be columnar structures, vertically mounted on the cutterhead body 1. All three nozzles spray water radially along the cutterhead body 1 to ensure that the high-pressure water jet can vertically impact the mud cake, improving the scouring effect. All nozzles are positioned in the central region 11 where the cutter head 12 is not located, thus not affecting the normal operation and design of the cutter head 12, ensuring that the cutting function and mud cake prevention function do not interfere with each other and work synergistically.

[0056] In this embodiment, the cutter head 12 is a double-edged hob, which is a structure containing two cutter rings.

[0057] Reference Figure 6 Considering that the depth of mud cake accumulation is not completely uniform, in this embodiment, the number of the first water jet nozzle 13, the second water jet nozzle 14, and the third water jet nozzle 15 is at least one. In this embodiment, the distance h between the nozzle 19 of the different first water jet nozzles 13, the second water jet nozzles 14, and the third water jet nozzles 15 and the panel of the cutter head body 1 is different, that is, the length of the first water jet nozzles 13, the second water jet nozzles 14, and the third water jet nozzles 15 extending out of the end face of the cutter head is different. They are arranged in a stepped manner on the cross-section of the cutter head, so that the high-pressure water jets ejected by the first water jet nozzles 13, the second water jet nozzles 14, and the third water jet nozzles 15 at different positions can form a multi-layered flushing effect in front of the cutter head body 1, effectively cleaning mud cakes of different accumulation depths. Whether it is shallow mud cake or mud cake that has accumulated to a relatively thick depth, it can be effectively cleaned, effectively enhancing the comprehensiveness and effectiveness of mud cake prevention and control.

[0058] Reference Figure 6 The implementation methods of the first water jet nozzle 13, the second water jet nozzle 14, and the third water jet nozzle 15 in this embodiment will be described. The first water jet nozzle 13, the second water jet nozzle 14, and the third water jet nozzle 15 can adopt the same structure and are collectively referred to as water jet nozzles. The water jet nozzle is provided with a water inlet channel 16, which runs through the water jet nozzle along the axial direction. The other end of the water jet nozzle is provided with a water outlet channel 17, which is perpendicular to the axial direction of the water jet nozzle. Both ends of the water outlet channel 17 pass through the outer peripheral surface of the water jet nozzle. The water outlet channel 17 is connected to the water inlet channel 16, so the water jet nozzle can spray two jets of high-pressure water in opposite directions to achieve bidirectional scouring.

[0059] In practical work, the spray direction at both ends of the water outlet channel 17 can be flexibly adjusted according to the distribution of mud cake on the cutter head 12. For example, both ends can be aimed at the area around the cutter head 12 where mud cake is easy to accumulate, or at least one end can be aimed at one cutter head 12, thereby increasing the coverage area of ​​the high-pressure water jet on the area around the cutter head 12 and the central area 11, enhancing the scouring effect and improving the scouring efficiency.

[0060] In addition, a pulse solenoid valve 18 is installed on the water inlet channel 16 of the water jet nozzle, enabling the water jet nozzle to perform pulse water spraying when needed, thereby improving the efficiency of mud cake removal. Compared with continuous water spraying, pulse water spraying can generate a stronger impact force, and with the same water consumption, the effect of breaking up and removing mud cake is more significant, effectively reducing water consumption and improving the efficiency of mud cake control.

[0061] To achieve intelligent mud cake prevention, this embodiment also includes a control system. The control system monitors the torque of the cutterhead body 1 and adjusts the water pressure, flow rate, and spraying time of each water jet nozzle based on the torque. Torque is a key parameter reflecting the working state of the cutterhead body 1 and the degree of mud cake accumulation. When the torque of the cutterhead body 1 increases, indicating severe mud cake accumulation, the control system can increase the water pressure, flow rate, and spraying time of each water jet nozzle to quickly remove the mud cake and reduce the load on the cutterhead body 1. Conversely, it can decrease these parameters to save energy and extend the equipment's service life.

[0062] In this embodiment, the control system monitors the torque of the cutter head body 1 in several ways, including but not limited to the following: (1) Using a torque sensor, the torque sensor is installed on the drive unit of the drive motor 6 that can drive the cutter head body 1 to rotate. The torque sensor works based on the strain principle or the magnetoelectric principle and can directly measure the torque transmitted by the drive unit. It has the characteristics of high measurement accuracy and fast response speed, and can obtain the torque information of the cutter head body in real time and accurately.

[0063] (2) When the drive motor 6 that drives the cutter head body 1 to rotate is running, the current magnitude and the output torque have a certain correspondence. According to the characteristic curve of the drive motor 6 and the relevant mathematical model, the torque of the cutter head body 1 can be indirectly calculated based on the current of the drive motor 6 by detecting the working current of the drive motor 6. There is no need to install a special torque sensor, and the cost is relatively low. While ensuring a certain measurement accuracy, it effectively reduces the cost investment of the equipment and is suitable for engineering projects with strict cost control.

[0064] (3) If the cutter head body 1 is hydraulically driven, a pressure sensor can be installed in the hydraulic drive system. For example, the pressure sensor can be installed on the oil inlet or return pipe of the hydraulic motor. When the cutter head body 1 is subjected to torque, the pressure in the hydraulic system will change. The pressure sensor can measure this pressure change and obtain the torque value through the corresponding conversion relationship. The pressure sensor is easy to install and requires little modification to the hydraulic drive system.

[0065] On the panel of the cutter body 1, the mud cake usually forms gradually in the central area 11 first, and then accumulates thicker and thicker, gradually expanding from the central area 11 to the edge of the cutter body 1.

[0066] Reference Figure 7In this embodiment of the tunnel boring machine, the rear of the cutterhead body 1 is rotatably connected to the shield shell 2. The shield shell 2 is a cylindrical hollow structure. The inner circumferential surface of the shield shell 2 is connected to a shield partition 3 that is parallel to the cutterhead body 1. An installation ring 4 is installed on the side of the shield partition 3 near the cutterhead body 1. The end of the installation ring 4 near the cutterhead body 1 is connected to the outer ring of the shield bearing 5. The inner ring of the shield bearing 5 is connected to the drive part of the drive motor 6. The drive motor 6 is installed on the shield partition 3. The drive part of the drive motor 6 is connected to the cutterhead support arm 8 through a connecting plate 7. The cutterhead support arm 8 is connected to the cutterhead body 1.

[0067] The shield diaphragm 3 has a central hole in the center. The outer cylinder 9 passes through the central hole and is fixedly connected to the shield diaphragm 3. The outer circumferential surface of the outer cylinder 9 has a high-pressure water inlet 011. The inner cylinder 010 is coaxially sleeved inside the outer cylinder 9. The inner cylinder 010 is rotatably connected to the outer cylinder 9. The end of the inner cylinder 010 near the cutterhead body 1 is connected to the connecting plate 7, and the inner cylinder 010 penetrates the connecting plate 7.

[0068] The first water jet nozzle 13, the second water jet nozzle 14, and the third water jet nozzle 15 are all connected to one end of the high-pressure water inlet pipe 1 012 via a rotary joint. The other end of the high-pressure water inlet pipe 1 012 passes through the connecting plate 7 and extends into the gap between the outer cylinder 9 and the inner cylinder 010. This gap is connected to the high-pressure water inlet 011. The high-pressure water inlet 011 is connected to one end of the high-pressure water inlet pipe 2 013. The other end of the high-pressure water inlet pipe 2 013 is connected to the water storage tank 014. The water storage tank 014 is connected to the air pressure pump 015 via the high-pressure water inlet pipe 3 016. The high-pressure water inlet pipe 2 013 is equipped with a high-pressure water valve 017, and the high-pressure water inlet pipe 3 016 is equipped with a pressure gauge 018.

[0069] The first water jet nozzle 13, the second water jet nozzle 14, and the third water jet nozzle 15 on the cutterhead body 1 can spray high-pressure water in the radial direction of the cutterhead body 1. At the same time, the cutterhead body 1 can rotate under the drive of the drive motor 6, excavating and removing mud cake at the same time. This improves the construction efficiency of the tunnel boring machine in complex geological conditions such as rich mud soil layers, effectively reduces problems such as cutterhead wear and equipment failure caused by mud cake accumulation, reduces construction costs, shortens the construction cycle, and significantly improves the overall performance and engineering applicability of the tunnel boring machine.

[0070] After the tunnel boring machine (TBM) is started, the drive motor 6 begins to work. The drive unit of the drive motor 6 is connected to the cutterhead support arm 8 via the connecting plate 7. The cutterhead support arm 8 is in turn connected to the cutterhead body 1. Therefore, the power of the drive motor 6 can be transmitted to the cutterhead body 1, causing the cutterhead body 1 to rotate around the central axis of the shield bearing 5. Several cutter heads 12 set in the central area 11 of the cutterhead body 1 rotate accordingly, excavating clay strata, clayey sandy soil strata, etc., cutting the soil into fine sand particles and debris.

[0071] The air pump 015 is connected to the water storage tank 014 via the high-pressure water inlet pipe 3 016, pressurizing the water storage tank 014. The pressure gauge 018 on the high-pressure water inlet pipe 3 016 monitors the pressure provided by the air pump 015 in real time, ensuring that the water supply pressure is stable within a suitable range, thus guaranteeing the stable spraying of the subsequent high-pressure water jet. High-pressure water flows out from the water storage tank 014, passing through the high-pressure water inlet pipe 2 013. The high-pressure water valve 017 on the high-pressure water inlet pipe 2 013 controls the on / off state and flow rate of the high-pressure water. The high-pressure water enters the high-pressure water inlet 011 of the outer cylinder 9. The inner cylinder 010 is coaxially sleeved inside the outer cylinder 9, and the inner cylinder 010 is rotatably connected to the outer cylinder 9. After entering the gap between the outer cylinder 9 and the inner cylinder 010, the high-pressure water is delivered to the first water jet nozzle 13, the second water jet nozzle 14, and the third water jet nozzle 15 through the high-pressure water inlet pipe 1 012. As the cutter head body 1 rotates continuously, the high-pressure water inlet pipe 1012 is connected to the first water jet nozzle 13, the second water jet nozzle 14, and the third water jet nozzle 15 through a rotary joint, ensuring a continuous supply of high-pressure water during the rotation of the cutter head body 1.

[0072] The first water jet nozzle 13, the second water jet nozzle 14, and the third water jet nozzle 15 can spray high-pressure water along the radial direction of the cutterhead body 1. These nozzles are designed based on the position of the cutter head 12, and the high-pressure water jets they spray precisely impact areas on the cutterhead body 1 where mud cake easily forms, such as the intersection of the extended lines L of the cutter head 12's cutter ring symmetry axis, the intersection of the extended line K connecting the center of the outer cutter ring and any corner of the inner cutter ring, and the center of a specific circumscribed circle C. All three nozzles can achieve pulsed water spraying, enhancing the breaking and removal of mud cake and reducing water consumption. Through the high-pressure water jet flushing, the mud cake is broken and dispersed, preventing it from covering the cutter head and clogging the cutter cylinder, reducing cutter wear, ensuring normal cutter rotation and cutting function, and maintaining stable tunneling of the cutterhead body 1.

[0073] During the excavation of soil by the cutterhead body 1, the control system monitors the torque of the cutterhead body 1 in real time. During normal excavation, the torque of the cutterhead body 1 remains within a relatively stable range. However, as mud cake gradually accumulates on the cutterhead body 1, the resistance it experiences increases, and the torque also rises accordingly. Based on the monitored torque changes, the control system automatically adjusts the water pressure, flow rate, and spraying time of the first water jet nozzle 13, the second water jet nozzle 14, and the third water jet nozzle 15. For example, when the increased torque indicates heavier mud cake accumulation, the control system increases the water pressure and flow rate of the first water jet nozzle 13, the second water jet nozzle 14, and the third water jet nozzle 15, and appropriately extends the spraying time to enhance the flushing and breaking up of the mud cake; conversely, when the torque returns to normal, the corresponding parameters are reduced to save energy and water resources.

Claims

1. A method for arranging a shield tunnel cutterhead mud cake monitoring system, characterized in that, include: S1. Monitoring Area Division: Based on the shield cutterhead structure and mud cake accumulation pattern, the monitoring area is divided into a primary monitoring area, a secondary monitoring area, and a tertiary monitoring area. The primary monitoring area is the cutterhead edge area; the secondary monitoring area is the cutterhead transition area; and the tertiary monitoring area is the cutterhead center area. S2. Sensor selection and layout: In the primary monitoring area, a set of "vibration-temperature-stress" triple sensors are arranged at 15° intervals, and earth pressure sensors are densely arranged at the tool holder connection area; in the secondary monitoring area, "vibration-temperature" dual sensors are arranged at 30° intervals; in the tertiary monitoring area, laser displacement sensors, high-definition industrial cameras, LED supplementary lighting modules and moisture content sensors are arranged. S3. Monitoring System Construction: A three-level system of "perception layer - transmission layer - processing layer" is established. The perception layer is the aforementioned sensor, the transmission layer adopts a hybrid "wireless + wired" transmission method, and the processing layer includes edge computing nodes and cloud servers. The edge nodes are responsible for data preprocessing, and the cloud servers are responsible for data storage and predictive calculations.

2. The method for arranging the shield cutterhead mud cake monitoring system according to claim 1, characterized in that: The primary monitoring zone is 70%-100% of the cutterhead radius, the secondary monitoring zone is 30%-70% of the cutterhead radius, and the tertiary monitoring zone is 0%-30% of the cutterhead radius and the inner wall of the soil chamber.

3. The method for arranging the shield cutterhead mud cake monitoring system according to claim 1, characterized in that: The vibration sensor is an IEPE type accelerometer, and the temperature sensor is a PT1000 platinum resistance sensor.

4. The method for arranging the shield cutterhead mud cake monitoring system according to claim 1, characterized in that: The wireless transmission uses a 5G industrial module; the wired transmission uses an industrial Ethernet.

5. A method for predicting shield cutterhead mud cake based on the arrangement method described in any one of claims 1-4, characterized in that, Includes the following steps: A1. Multi-source data preprocessing: Sensor data is processed using an improved wavelet threshold denoising algorithm, data synchronization is achieved through timestamp alignment, and missing data is filled in using linear interpolation. A2. Feature Enhancement and Fusion: Extract statistical features from sensor data and image features from mud cake images, assign weights to the fused features through a channel-spatial dual attention mechanism, and output enhanced features; A3. Spatiotemporal Fusion Prediction: The enhanced features are input into the "CNN-LSTM-ELM" fusion model. The CNN extracts spatial correlation features, the improved LSTM extracts temporal correlation features, and the ELM optimized by the chaotic particle swarm algorithm outputs the preliminary prediction results. A4. Error Correction and Early Warning: A rolling window error correction mechanism is used to correct the prediction results, and graded early warning information is output based on the corrected probability of mud cake formation.

6. The method for predicting mud cake on a tunnel boring machine cutterhead according to claim 5, characterized in that: The improved wavelet thresholding denoising algorithm uses a continuously differentiable threshold function, and achieves a smooth transition at the threshold through an exponential term, thereby reducing signal distortion.

7. The method for predicting mud cake on a tunnel boring machine cutterhead according to claim 5, characterized in that: The specific strategies for feature enhancement and fusion in A2 are as follows: A21. Statistical Feature Extraction; A22. Image feature extraction; A23. Improved attention mechanisms.

8. The method for predicting mud cake on a tunnel boring machine cutterhead according to claim 5, characterized in that: The improved LSTM model enhances the ability to filter redundant temporal information by introducing a nonlinear transformation of the hidden layer from the previous time step to optimize the forget gate.

9. A water jet cleaning control method for the cutterhead of a tunnel boring machine based on any one of claims 1-8, comprising: Based on the cutter head cleaning grid parameter set, the cutter head surface is cut into strips radially according to structural units and then subdivided according to tangential angles. The center point of the sub-region is corrected with the test frame positioning reference. Then, it is reordered and numbered according to the direction of the cutter head rotation area to obtain the cutter head cleaning grid parameter set. Based on the cutter head cleaning grid parameter set, the installation point and direction of the water jet nozzle are read, and the distance and direction between the center point of the sub-area and the pressure node of the test water supply pipeline are combined to form an energy value. Then, the effective coverage range is determined by a threshold to obtain the jet coverage energy distribution set. Based on the spray coverage energy distribution set, the Hungarian allocation algorithm is used to calculate the cleaning time according to the thickness of the sub-region adhesion layer and the pressure value. The time values ​​are formed into a matrix and multiple values ​​are compared in the same row to determine the nozzle correspondence. All matching items are combined into a unified sequence to obtain the nozzle cleaning assignment result set. Based on the nozzle cleaning assignment result set, the particle swarm optimization algorithm is used to add fine-tuning amount to the nozzle installation point, and then the distance is compared with the center point of the cutter head structure unit. The reachability is judged by the angle between the nozzle direction and the normal vector of the cleaning area. The reachable units are recorded to obtain the nozzle arrangement adjustment parameter set. Based on the nozzle arrangement adjustment parameter set, the nozzle start-up time is read according to the test, the assigned sub-region time is arranged into the time axis and aligned with the cutter head rotation area reference, and then integrated into a continuous time period and uniformly sorted to obtain the nozzle start-up and stop sequence table.

10. A tunnel boring machine, characterized in that, Includes a control system for performing the method according to any one of claims 1-9.

Images