Shield tunneling adaptive control method and system based on stratum identification

Abstract

The application relates to the technical field of tunnel construction, and discloses a shield tunneling self-adaptive control method and system based on stratum identification, the method comprising the following steps: in the shield tunneling process, collecting a shield cutter cutting vibration signal and a tunneling state parameter in real time; monitoring a shield advancing distance in real time, obtaining a spatial slice every fixed advancing distance, for the current spatial slice, constructing a vibration signal feature and a tunneling load feature data pair; taking the vibration signal feature and the tunneling load feature data pair as input, adopting a pre-trained stratum identification model to obtain a current stratum type; and optimizing real-time tunneling parameters according to the current stratum type. The application can improve the stratum identification and tunneling control precision based on multi-source data fusion.

Description

Technical Field

[0001] This invention belongs to the field of tunnel construction technology, specifically relating to an adaptive control method and system for tunnel boring machine excavation based on stratum identification. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] In urban rail transit construction, shield tunneling faces complex geological environments, including spheroidal weathered bodies (isolated boulders), karst cavities, and composite strata with soft upper layers and hard lower layers. The uncertainty of geological conditions can easily lead to adverse consequences such as cutterhead mud cake formation, abnormal cutter wear, and even breakage. Current technologies mostly rely solely on the amplitude of vibration signals to determine rock hardness. However, the vibrations of hard rock cutting at low penetration depths and the vibrations of soft soil cutting at high penetration depths, which are extremely similar in amplitude in the time domain, can easily cause misjudgments of stratum hardness, affecting construction accuracy. To solve the problem of misjudgment from a single data source, current technologies are gradually adopting multi-source data, such as vibration signals and shield tunneling data, to identify stratum hardness or type. However, shield tunneling data is a low-frequency scalar, while vibration signals are high-frequency waveforms. Current technologies typically use time alignment for data preprocessing, but shield tunneling machines experience various operating conditions, such as stopping, speed changes, and alternating constant speeds. This can lead to inaccurate matching of high-frequency vibration signals and low-frequency shield tunneling data on the time axis, thus affecting the perception of geological conditions and making it difficult to effectively avoid adverse construction consequences. Summary of the Invention

[0004] In view of this, the present invention provides an adaptive control method and system for tunnel boring machines based on stratum identification. This improves the accuracy of stratum identification and tunneling control based on multi-source data fusion.

[0005] The first aspect of the present invention provides an adaptive control method for tunnel boring machine excavation based on stratum identification, characterized by comprising the following steps: During the tunnel boring machine's excavation process, the cutting vibration signal of the tunnel boring machine's cutterhead and the tunneling status parameters are collected in real time. Real-time monitoring of the tunnel boring machine's advance distance; a spatial slice is obtained after each fixed advance distance; for the current spatial slice, a data pair of vibration signal characteristics and tunneling load characteristics is constructed. Using vibration signal characteristics and tunneling load characteristics as input, a pre-trained stratum identification model is used to obtain the current stratum category; Optimize real-time tunneling parameters based on the current geological formation type.

[0006] In some embodiments, the vibration signal features are time-frequency feature spectrum tensors, which are obtained by performing a short-time Fourier transform on the vibration signals within the current spatial slice; the tunneling load features are state feature time series, which are obtained by normalizing the tunneling state parameters at different times within the current spatial slice and splicing them in chronological order.

[0007] In some embodiments, the training method for the stratigraphic identification model includes: The machine collects cutterhead cutting vibration signals and tunneling status parameters of the tunnel boring machine under different geological conditions. At the same time, it obtains the actual geological category. Taking each fixed distance advanced by the tunnel boring machine as a spatial slice, the machine extracts vibration signal features and tunneling load features of each spatial slice and associates them with the actual geological category to obtain a training dataset. Based on a pre-built neural network model architecture, a stratum identification model is obtained by training a training dataset. The neural network model architecture includes a convolutional neural network branch and a long short-term memory network branch, which take vibration signal features and tunneling load features as inputs, respectively. The outputs of both branches are connected to a cross-attention fusion module.

[0008] In some embodiments, the stratum identification model extracts features from vibration signals using a convolutional neural network branch to obtain a rock mass strength feature vector, extracts features from tunneling load features using a long short-term memory network branch to obtain a load trend feature vector, and fuses these features using a cross-attention fusion module to obtain a comprehensive geological feature vector. The fusion method is as follows: using the load trend feature vector as the query benchmark and the rock mass strength feature vector as the matching basis, the corresponding attention weights are obtained based on the correlation between each dimension of the rock mass strength feature and the load trend feature; the dimensions of the rock mass strength feature vector are then weighted and fused according to the attention weights; the fused rock mass strength feature vector is then concatenated and fused with the load trend feature vector to obtain the comprehensive geological feature vector.

[0009] In some embodiments, the strata categories include homogeneous hard rock, boulders / objects, cohesive soft soil, and composite strata with a soft upper layer and a hard lower layer.

[0010] In some embodiments, if the formation is identified as homogeneous hard rock, the propulsion speed is adjusted to keep the instantaneous penetration within a set range; the rotation speed is disturbed every set time interval, and the mechanical rock breaking specific energy is calculated in real time. If the acceleration causes the mechanical rock breaking specific energy to decrease, the rotation speed is increased; otherwise, the rotation speed is decreased, so that the tunnel boring machine operates at the point of lowest energy consumption. If a boulder / foreign object is identified, the cutter head speed will be forcibly reduced and the total thrust will be limited. If the soil is identified as cohesive soft soil, the relationship between the mechanical rock breaking specific energy and the vibration energy is monitored in real time. When the mechanical rock breaking specific energy increases beyond the set specific gravity and the vibration signal attenuates beyond the set specific gravity, the cutter head speed is increased and the cutter head flushing function is activated. If the geological formation is identified as a composite stratum with a soft upper layer and a hard lower layer, the advance speed is limited, and the thrust difference between the upper and lower section hydraulic cylinders of the tunnel boring machine is dynamically adjusted to ensure the stability of the tunnel boring machine's construction posture.

[0011] A second aspect of the present invention provides an adaptive control device for tunnel boring machines based on stratum identification, comprising: The real-time data acquisition module is configured to collect the cutting vibration signal of the shield machine cutterhead and the tunneling status parameters in real time during the tunneling process; The grid map building module is configured to monitor the tunnel boring machine's advance distance in real time. A spatial slice is obtained for each fixed advance distance. For the current spatial slice, a data pair of vibration signal characteristics and tunneling load characteristics is constructed. The state vector construction module is configured to take vibration signal feature and tunneling load feature data pairs as input, and use a pre-trained stratum identification model to obtain the current stratum category; The dynamic path planning module is configured to optimize the cutterhead's real-time tunneling parameters based on the current geological formation type.

[0012] A third aspect of the present invention provides an electronic device including a processor and a memory, wherein the memory stores computer instructions that, when executed by the processor, cause the electronic device to perform the method described thereon.

[0013] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method.

[0014] A fifth aspect of the present invention provides a computer program product comprising a computer program that, when executed by a processor, implements the method described herein.

[0015] The above one or more technical solutions achieve spatial alignment of vibration signals and tunneling load characteristics by using a fixed advance distance as a spatial slice. This effectively solves the problem of time alignment misalignment caused by the influence of construction conditions on the two types of heterogeneous data, ensuring that the vibration signal characteristics and tunneling load characteristics of the same spatial slice correspond to the same geological conditions, thus improving the accuracy of subsequent geological identification. Based on this, using the aligned feature data pair as input, the geological identification model can accurately obtain the current geological category and clarify the current tunneling geological conditions. Then, based on the identified geological category, the real-time tunneling parameters of the cutterhead can be optimized in a targeted manner to achieve the adaptation of the tunnel boring machine's tunneling to the geological conditions. Attached Figure Description

[0016] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0017] Figure 1 A flowchart of the adaptive control method for tunnel boring machine excavation based on stratum identification provided in an embodiment of this application is shown; Figure 2 A schematic diagram of the stratigraphic identification model in an embodiment of this application is shown; Figure 3 The diagram shows the program module architecture of the tunnel boring machine adaptive control device based on stratum identification provided in this application embodiment. Detailed Implementation

[0018] Embodiments of this application will now be described in more detail with reference to the accompanying drawings. While some embodiments of this application are shown in the drawings, it should be understood that this application can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this application. It should be understood that the drawings and embodiments of this application are for illustrative purposes only and are not intended to limit the scope of protection of this application.

[0019] In the description of the embodiments of this application, the term "comprising" and similar terms should be understood as open-ended inclusion, i.e., "including but not limited to". The term "based on" should be understood as "at least partially based on".

[0020] To address the issues of misalignment in time between high-frequency vibration signals and low-frequency PLC tunneling parameters, and inaccurate stratum identification in existing technologies, one or more embodiments of this invention achieve spatial alignment of two types of heterogeneous data by using a fixed distance of shield machine advancement as a spatial slice, eliminating interference from construction conditions on data matching. Then, using the aligned vibration signal characteristics and tunneling load characteristics data pairs as input, a pre-trained stratum identification model is used to accurately identify the current stratum type. Finally, based on the identification results, the cutterhead tunneling parameters are specifically optimized, achieving adaptation of shield machine tunneling to complex stratum conditions and improving construction safety and accuracy.

[0021] In one or more of the following embodiments, the non-pressure-bearing side of the tunnel boring machine's earth chamber wall has a pre-set high-frequency vibration sensor installation interface, and also has an industrial Ethernet data transmission module, which can synchronously read all tunneling status parameters such as the total thrust, cutterhead torque, and cutterhead speed of the PLC system, and supports the synchronous acquisition and transmission of construction status signals (cutterhead start / stop, rate of change of propulsion speed, etc.); the cutterhead has an adjustable speed function and supports dynamic fine-tuning of the speed, the cutterhead drive system has a thrust limiting function, and can quickly adjust the total thrust output according to control commands, and the cutterhead needs to be equipped with a flushing system to support automatic adjustment of flushing flow; in terms of propulsion system structure, the propulsion cylinders need to adopt a partitioned design (at least including upper and lower partition cylinders), support independent dynamic adjustment of the thrust of each partition cylinder to control the attitude, and the propulsion system has a propulsion speed control function; in addition, the tunnel boring machine also has a control module that can execute the adaptive control method provided in one or more of the following embodiments.

[0022] Figure 1 A flowchart of an example method 100 for adaptive control of tunnel boring machine excavation based on stratum identification, provided by one or more embodiments of the present invention, is shown, including steps S101-S104. It should be understood that method 100 may also include additional actions not shown. Method 100 is described in detail below.

[0023] S101. During the tunneling process of the tunnel boring machine, the cutting vibration signal of the tunnel boring machine cutterhead and the tunneling status parameters are collected in real time. S102. Real-time monitoring of the tunnel boring machine's advance distance. A spatial slice is obtained after each fixed advance distance. For the current spatial slice, a data pair of vibration signal characteristics and tunneling load characteristics is constructed. S103. Using vibration signal characteristics and tunneling load characteristics data pairs as input, a pre-trained stratum identification model is used to obtain the current stratum category; S104. Optimize the real-time tunneling parameters of the cutterhead according to the current geological formation type.

[0024] By using a fixed advance distance as a spatial slice, spatial alignment of vibration signals and tunneling load characteristics is achieved, effectively solving the problem of temporal alignment misalignment caused by the influence of construction conditions on the two types of heterogeneous data. This ensures that the vibration signal characteristics and tunneling load characteristics of the same spatial slice correspond to the same geological conditions, improving the accuracy of subsequent geological identification. Based on this, using the aligned feature data pair as input, the geological identification model can accurately obtain the current geological category and clarify the current tunneling geological conditions. Then, based on the identified geological category, the real-time tunneling parameters of the cutterhead are optimized in a targeted manner to achieve the adaptation of the tunnel boring machine's tunneling to the geological conditions.

[0025] In step S101, high-frequency vibration sensors are installed on the non-pressure-bearing side of the tunnel boring machine's soil chamber wall to collect vibration signals in real time during the cutterhead cutting process. Simultaneously, the tunneling status parameters of the tunnel boring machine's PLC system are read synchronously via industrial Ethernet. The tunneling status parameters include at least total thrust, cutterhead torque, cutterhead rotation speed, propulsion speed, and screw conveyor pressure. The vibration signals and PLC tunneling status parameters acquired in this step serve as the raw foundation data for all subsequent data processing, feature extraction, and control calculations.

[0026] Specifically, the high-frequency vibration sensor uses an IEPE piezoelectric accelerometer, which is installed at the 3, 6, 9, and 12 o'clock positions on the back of the tunnel boring machine's soil chamber partition (atmospheric pressure side) using a strong magnetic base and epoxy resin adhesive, directly sensing the cutting vibration transmitted from within the soil chamber. The PLC acquisition subsystem reads data from the tunnel boring machine's PLC registers via industrial Ethernet, with a sampling rate of 1Hz. Acquired parameters include: total thrust. (kN), cutter head torque (kN·m), propulsion speed (mm / min), cutter head speed (rpm).

[0027] In step S102, the tunnel boring machine's (TBM) advance distance is monitored in real time, and the continuous data in the time dimension is converted into slice data in the spatial dimension. The vibration signal is then aligned with the PLC's tunneling status parameters in the spatial dimension. Specifically, the TBM's advance distance is monitored in real time, and the vibration signal and tunneling status parameters are recorded using the advance distance as an index. A fixed distance is set for each advance of the TBM. For a spatial slice, exemplarily, this fixed distance It can be set to 10mm. Based on this slice, the vibration signal and PLC tunneling status parameters collected in the corresponding time period are extracted, and the continuous data in the time dimension is converted into slice data in the spatial dimension, so as to achieve the initial alignment of the two types of heterogeneous data.

[0028] For each spatial slice, feature extraction is performed based on the vibration signal and tunneling load parameters. Specifically, a short-time Fourier transform (STFT) is performed on the vibration signal within a single spatial slice to generate a time-frequency feature map tensor. This tensor can clearly preserve the frequency domain energy distribution and impact characteristics of vibration signals (such as the high-frequency components of hard rock and the low-frequency components of soft soil), intuitively reflecting the hardness and brittleness of the strata. The PLC tunneling state parameters within a single spatial slice are normalized. Features such as thrust, torque, rotational speed, penetration depth, and cutterhead power acquired at different times within this spatial slice are then spliced ​​together in chronological order to generate a state feature time series. This vector corresponds to the load change of the tunnel boring machine when it is tunneling within this spatial slice, and can reflect the magnitude and stability of the stratum resistance.

[0029] For each spatial slice, a corresponding time-frequency feature map tensor and state feature time series are matched to construct one-to-one heterogeneous data pairs. .

[0030] High-frequency vibration signals reflect the instantaneous rock-breaking intensity of the cutterhead, while low-frequency PLC parameters reflect the macroscopic load trend of the tunnel boring machine (TBM). Traditional time alignment only splices the data in time sequence. However, due to the TBM's operating conditions such as shutdown, speed change, and alternating constant speed, the high-frequency vibration signal (12.8kHz) and the low-frequency PLC parameters (1Hz) cannot be precisely matched on the time axis. This results in misalignment issues, where the same vibration signal corresponds to multiple sets of PLC parameters or one set of PLC parameters corresponds to multiple vibration signal segments, leading to distortion in subsequent feature extraction. Spatial dimension alignment, on the other hand, binds the two types of data to the same spatial slice (the same advance mileage), associating instantaneous vibration characteristics with the load characteristics of the corresponding mileage. This achieves physical adaptation of heterogeneous data and fundamentally avoids the misalignment risks of time alignment.

[0031] In addition, spatial dimension alignment is based on the advance distance, which matches the actual working conditions of shield tunneling (advance by mileage). This ensures that the vibration signal characteristics and tunneling load characteristics of the same spatial slice correspond to the same geological conditions, providing data support for subsequent geological identification and indirectly improving the accuracy of geological identification.

[0032] In step S103, the training method for the stratigraphic identification model includes: (1) Collect raw data and perform preprocessing. Collect the cutterhead cutting vibration signal and PLC tunneling status parameters (total thrust, cutterhead torque, etc.) of the tunnel boring machine under different geological conditions. At the same time, obtain the actual geological category data corresponding to the tunneling mileage. This data can be obtained through geological survey reports, on-site borehole sampling, and real-time observation records during construction to ensure that the geological category is true and accurate, covering four target geological categories: homogeneous hard rock, cohesive soft soil, upper soft and lower hard composite strata, and isolated boulders / foreign objects. Then, perform standardized preprocessing on the collected raw data, remove abnormal data, denoise the vibration signal, and normalize the PLC tunneling status parameters to ensure the validity and consistency of the data. At the same time, referring to step S102, take each fixed distance advanced by the tunnel boring machine (exemplary 10mm) as a spatial slice, and extract the vibration signal and PLC tunneling status parameters corresponding to each spatial slice.

[0033] In addition to acquiring the PLC tunneling status parameters, construction status signals are also collected simultaneously, including cutterhead start / stop status, propulsion speed change rate, and rotational speed fluctuation value. Combined with propulsion mileage data, the construction status (effective tunneling, shutdown, speed change) corresponding to each spatial slice is determined. An effective tunneling state is defined as a state where the cutterhead rotates normally and the propulsion speed is stable. The criteria for determining stable propulsion speed include, for example, a fluctuation value ≤ 0.5 mm / min and a rotational speed fluctuation ≤ 0.1 rpm. Based on the criteria for determining the effective tunneling state, the relevant data for spatial slices corresponding to invalid data are discarded.

[0034] (2) Constructing the training dataset. First, heterogeneous data pairs are constructed: short-time Fourier transform is performed on the vibration signal of each spatial slice to generate a time-frequency feature spectrum tensor (vibration signal features); the PLC tunneling state parameters of each spatial slice are processed to generate a state feature vector time series (tunneling load features); a set of corresponding time-frequency feature spectrum tensors and state feature vectors are matched for each spatial slice to construct a one-to-one heterogeneous data pair, ensuring that each data pair corresponds to the vibration and load features of the same spatial slice. Then, the heterogeneous data pairs are associated with the corresponding actual stratum categories to obtain the training dataset. For example, the stratum categories include: homogeneous hard rock, boulders / foreign objects, cohesive soft soil, and composite strata with soft upper and hard lower layers.

[0035] (3) Training the stratigraphic identification model. The stratigraphic identification model is based on a CNN-LSTM dual-stream network and a cross-attention mechanism, combined with fully connected layers, a loss function, and an optimizer. Specifically, the model architecture includes a convolutional neural network (CNN) branch, a long short-term memory network (LSTM) branch, a cross-attention fusion module, fully connected layers, and an activation function, while also setting a loss function and an optimizer. The outputs of both branches are connected to the cross-attention fusion module, and the output of the cross-attention fusion module is connected to 1-2 fully connected layers. An activation function is set at the output of the fully connected layers. Figure 2 As shown, the CNN branch uses a 3-layer convolutional network, with the input being... The STFT pattern. The first layer uses... Convolution kernels extract frequency distribution features (identifying the dominant frequency of rocks); the second layer uses... Convolutional kernels extract texture features (to identify impact pulses). The output is a 128-dimensional rock mass strength feature vector. A larger vector magnitude indicates higher rock strength; a higher proportion of high-frequency components indicates greater rock brittleness. The LSTM branch uses a two-layer Bi-LSTM structure, with the input being a time series of state vectors within a spatial cell, and the output being a 128-dimensional load trend feature vector. It is used to capture the rate of change of tunneling parameters. For example, a linear increase in torque often indicates an increase in the clay content of the formation.

[0036] Among them, the CNN branch uses vibration signal characteristics, i.e., the time-frequency feature map tensor. Using this as input, the frequency domain energy distribution and impact characteristics are extracted, and the rock mass strength feature vector is output. It is used to reflect the hardness and brittleness of the current strata; the LSTM branch uses tunneling load characteristics, i.e., state characteristic time series. As input, extract the trend of tunneling parameters with the advance mileage, and output the load trend feature vector. This method is used to determine the stability of formation resistance. Both the rock mass strength feature vector and the load trend feature vector are input into the cross-attention fusion module. The load trend feature vector output by the LSTM is used as the query benchmark, and the rock mass strength feature output by the CNN is used as the matching basis. Based on the correlation between each dimension of the rock mass strength feature and the load trend feature, corresponding attention weights are obtained. Each dimension of the rock mass strength feature vector is weighted according to these attention weights, filtering and strengthening feature dimensions with high correlation to load trend and formation type, while weakening irrelevant feature interference. Finally, the weighted rock mass strength feature vector is concatenated and fused with the load trend feature vector used as the query benchmark to obtain a comprehensive geological feature vector that reflects the formation characteristics. This comprehensive geological feature vector retains the instantaneous rock breaking strength features (formation hardness, brittleness) captured by the CNN branch and integrates the macroscopic load trend features (formation resistance, stability) captured by the LSTM branch, achieving complementary fusion of instantaneous and trend features.

[0037] For example, the cross-attention fusion module is configured to: use the load trend feature vector As a query vector, the rock mass strength feature vector As key and value vectors, a linear transformation is performed using a learnable weight matrix to generate the query, key, and value required for attention computation: , ; in, , It is a learnable weight matrix used to map features to a unified attention space.

[0038] Subsequently, attention weights are calculated using a scaling dot product attention mechanism, and the rock mass strength features are then weighted and fused. ; in, For feature dimension, As a scaling factor, it is used to prevent the softmax function from saturating due to excessively large inner product values, thus ensuring training stability. In the formula, The weights, used to measure the correlation between each rock mass strength characteristic and the current load trend characteristic, are normalized using Softmax to obtain a weight distribution between 0 and 1. These weights directly correspond to the importance of each rock mass strength characteristic. Also generated from rock mass strength characteristics, the aforementioned attention weight matrix is ​​combined with... Multiplication is a dynamic weighting of the various dimensions of rock mass strength characteristics according to their importance. The higher the weight of a rock mass strength characteristic, the higher its proportion in the final output, thereby achieving weighted fusion of rock mass strength characteristics.

[0039] Finally, the original load trend characteristics are concatenated with the weighted rock mass strength characteristics to obtain the final geological comprehensive feature vector:

[0040] This cross-attention fusion mechanism allows the model to focus on key frequency band features in the vibration spectrum by utilizing macroscopic tunneling conditions (such as whether the torque changes abruptly). For example, when the LSTM detects stable torque (soft soil feature) but the CNN detects high-frequency vibration (suspected hard rock feature), the attention mechanism automatically reduces the weight of this vibration feature, classifying it as mechanical interference, thereby avoiding false alarms. Before a sudden torque change, the mechanism can keenly capture the impact features of isolated boulders in the vibration signal, achieving early warning.

[0041] Finally, a fully connected layer maps the high-dimensional geological feature vector to an output vector corresponding to the number of stratigraphic categories. A Softmax activation function is applied at the output of the fully connected layer to normalize the 4-dimensional output vector, ensuring that the value of each dimension falls between 0 and 1, and the sum of the four values ​​is 1. Each value represents the confidence level for determining the stratigraphic category of the current excavation face. Furthermore, a cross-entropy loss function is used to measure the deviation between the model's predictions and the actual stratigraphic category labels. Based on the Adam optimizer and the loss function calculation results, the learnable parameters of each layer of the model are adjusted to further improve the model's robustness.

[0042] During training, the aforementioned model can learn the patterns between a combination of vibration and load characteristics and a certain type of stratum, thereby providing support for real-time stratum type identification. For example, high-frequency vibration and stable load correspond to homogeneous hard rock, low-frequency vibration with no obvious impact and stable load correspond to cohesive soft soil, and vibration signals with sudden impact peaks and large load fluctuations correspond to boulders / foreign objects.

[0043] In step S104, the total thrust, cutterhead torque, cutterhead speed, and propulsion speed collected in real time in step S101 are combined with the PLC tunneling status parameters such as total thrust, cutterhead torque, cutterhead speed, and propulsion speed. These parameters are then substituted into the mechanical rock breaking energy calculation formula to calculate the mechanical rock breaking energy under the current construction conditions in real time. The excavation cross-sectional area is used as a fixed parameter of the tunnel boring machine in the calculation. The system has a built-in "geology-specific energy" adaptive control strategy library. Based on different geological formations and combined with real-time calculated mechanical rock-breaking specific energy and vibration signal characteristic indicators, it executes differentiated closed-loop tunneling control strategies, specifically: (1) Homogeneous hard rock formations: Implement a specific energy optimization control strategy. While maintaining a constant penetration depth, dynamically fine-tune the cutterhead rotation speed to find the optimal energy efficiency operating point with the goal of minimizing the rock-breaking specific energy. Specifically, if the current formation is a homogeneous hard rock formation and the signal vibration kurtosis is... Adjust the propulsion speed to maintain the instantaneous penetration within a set range, for example... The rotation speed is disturbed at set intervals, such as 30 seconds. At the same time, the mechanical rock-breaking specific energy is calculated in real time. If the acceleration causes the mechanical rock-breaking specific energy to decrease, the rotation speed is increased; otherwise, the rotation speed is decreased, so that the tunnel boring machine works at the point of lowest energy consumption.

[0044] The formula for calculating instantaneous penetration is as follows: The formula for calculating the mechanical rock-breaking specific energy is: ,in, F For total thrust, v To accelerate, T The torque of the cutter head. n The rotational speed of the cutter head. A The cross-sectional area of ​​the excavation.

[0045] (2) Isolated boulders and abrupt hard rock formations: Implement an impact response protection strategy. When the impact component in the vibration signal exceeds the limit or the formation identification result is an isolated boulder / foreign object, or the vibration signal kurtosis is continuous for 3 points, the protection strategy will be implemented. At this time, the cutterhead speed is forcibly reduced and the total thrust is limited to prioritize construction safety protection. Specifically, the controller bypasses PID regulation and forcibly reduces the cutterhead speed setpoint by a step within 500ms. The following limits the total thrust to: Within this range, prevent the hob from cracking under high pressure; maintain the grinding state until the vibration and impact characteristics disappear and the advance distance exceeds 1.5 meters.

[0046] (3) Cohesive soft soil strata: Implement a coordinated strategy to prevent mud cake formation, and monitor the relationship between the mechanical rock breaking specific energy and vibration energy in real time. When there is a deviation between the characteristics of abnormally high specific energy and decreased vibration energy, automatically increase the cutterhead speed and increase the flushing flow to prevent mud cake formation on the cutterhead. Specifically, when the specific energy... SEThe increase exceeded 20%, but the vibrational energy... The decline exceeded When this occurs, it indicates that the tool is encased and cutting has failed. The automatic tool turret speed is then increased to... The above utilizes centrifugal force to detach the mud cake; simultaneously, it is linked to the foam system to increase the foam injection flow rate. Then, start the central flushing pump. The calculation method for mechanical rock breaking specific energy is the same as above.

[0047] (4) Composite strata with soft upper layer and hard lower layer: Implement a differentiated thrust strategy for attitude stabilization, limit the advance speed to reduce uneven impact, and dynamically adjust the thrust difference between the upper and lower zone cylinders of the tunnel boring machine according to the distribution of the soft and hard interfaces of the strata to counteract the upward attitude trend caused by the hard rock at the bottom and ensure the stability of the tunnel boring machine's attitude during construction. Specifically, limit the advance speed to This reduces impact; simultaneously, it monitors the tunnel boring machine's pitch angle in real time: if a pitching trend is detected (pitch angle > 0.5°), it automatically increases the pressure of the upper hydraulic cylinder group and decreases the pressure of the lower hydraulic cylinder group, creating a reverse torque to forcibly lower the shield head's posture and maintain the tunnel axis deviation within a certain range. Within.

[0048] Figure 3 A schematic diagram of an apparatus provided by one or more embodiments of the present invention is shown. The apparatus 200 includes: a real-time data acquisition module 201, configured to acquire in real-time vibration signals of the shield machine cutterhead and tunneling status parameters during the tunneling process of the shield machine; a spatial slice construction module 202, configured to monitor the shield machine's advance distance in real time, obtaining a spatial slice after each fixed advance distance, and constructing a vibration signal feature and tunneling load feature data pair for the current spatial slice; a stratum type identification module 203, configured to use the vibration signal feature and tunneling load feature data pair as input, and employ a pre-trained stratum identification model to obtain the current stratum type; and a tunneling parameter optimization module 204, configured to optimize the real-time tunneling parameters of the cutterhead according to the current stratum type.

[0049] Furthermore, one or more embodiments of the present invention also provide an electronic device that can be used to implement the tunnel boring machine adaptive control method based on stratum identification in the above embodiments. The electronic device includes one or more processors, one or more memories coupled to the processors, and a communication module coupled to the processors.

[0050] The memory may include one or more non-volatile memories and one or more volatile memories. Examples of non-volatile memories include, but are not limited to, at least one of the following: read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, hard disk, compact disc (CD), digital video disc (DVD), or other magnetic and / or optical storage. Examples of volatile memories include, but are not limited to, at least one of the following: random access memory (RAM), or other volatile memories that do not persist during the duration of a power outage. The computer program may be stored in the ROM. When the processor executes the computer program, it implements the above-described adaptive control method for tunnel boring machine excavation based on stratum identification.

[0051] In some embodiments, the program may be tangibly contained in a computer-readable medium, which may include a device (such as a memory) or other storage device accessible by the device. The program may be loaded from the computer-readable medium into RAM for execution. The computer-readable medium may include any type of tangible non-volatile memory, such as ROM, EPROM, flash memory, hard disk, where a computer program is stored that, when executed by a processor, implements the aforementioned adaptive control method for tunnel boring machine excavation based on strata identification.

[0052] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When these computer program instructions are loaded and executed on a server or terminal, they generate, in whole or in part, the processes or functions described in the embodiments of this application. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic cable, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to the server or terminal, or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, and magnetic tape), an optical medium (e.g., digital video disk (DVD), etc.), or a semiconductor medium (e.g., solid-state drive).

[0053] Furthermore, although the operations are described in a specific order, this should be understood as requiring that such operations be performed in the specific order shown or in sequential order, or requiring that all illustrated operations be performed to achieve the desired result. In certain environments, multitasking and parallel processing may be advantageous. Similarly, although several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of this application. Certain features described in the context of individual embodiments may also be implemented in combination in a single implementation. Conversely, various features described in the context of a single implementation may also be implemented individually or in any suitable sub-combination in multiple implementations.

[0054] Although the subject matter has been described using language specific to structural features and / or methodological logic, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. Rather, the specific features and actions described above are merely illustrative examples of implementing the claims.

Claims

1. A shield tunneling adaptive control method based on stratum identification, characterized in that, Includes the following steps: During the tunnel boring machine's excavation process, the cutting vibration signal of the tunnel boring machine's cutterhead and the tunneling status parameters are collected in real time. Real-time monitoring of the tunnel boring machine's advance distance; a spatial slice is obtained after each fixed advance distance; for the current spatial slice, a data pair of vibration signal characteristics and tunneling load characteristics is constructed. Using vibration signal characteristics and tunneling load characteristics as input, a pre-trained stratum identification model is used to obtain the current stratum category; Optimize real-time tunneling parameters based on the current geological formation type.

2. The adaptive control method for tunnel boring machine excavation based on stratum identification as described in claim 1, characterized in that, The vibration signal features are time-frequency feature spectrum tensors, which are obtained by performing short-time Fourier transform on the vibration signals within the current spatial slice; the tunneling load features are state feature time series, which are obtained by normalizing the tunneling state parameters at different times within the current spatial slice and splicing them in chronological order.

3. The adaptive control method for tunnel boring machine excavation based on stratum identification as described in claim 1 or 2, characterized in that, The training method for the stratigraphic identification model includes: The machine collects cutterhead cutting vibration signals and tunneling status parameters of the tunnel boring machine under different geological conditions. At the same time, it obtains the actual geological category. Taking each fixed distance advanced by the tunnel boring machine as a spatial slice, the machine extracts vibration signal features and tunneling load features of each spatial slice and associates them with the actual geological category to obtain a training dataset. Based on a pre-built neural network model architecture, a stratum identification model is obtained by training a training dataset. The neural network model architecture includes a convolutional neural network branch and a long short-term memory network branch, which take vibration signal features and tunneling load features as inputs, respectively. The outputs of both branches are connected to a cross-attention fusion module.

4. The adaptive control method for tunnel boring machine excavation based on stratum identification as described in claim 3, characterized in that, The geological formation identification model extracts features from vibration signals using a convolutional neural network branch to obtain a rock mass strength feature vector, and extracts features from tunneling load characteristics using a long short-term memory network branch to obtain a load trend feature vector. These features are then fused using a cross-attention fusion module to obtain a comprehensive geological feature vector. The fusion method is as follows: using the load trend feature vector as the query benchmark and the rock mass strength feature vector as the matching basis, attention weights are obtained based on the correlation between each dimension of the rock mass strength feature and the load trend feature. The dimensions of the rock mass strength feature vector are then weighted and fused according to these attention weights. Finally, the fused rock mass strength feature vector is concatenated and fused with the load trend feature vector to obtain the comprehensive geological feature vector.

5. The adaptive control method for tunnel boring machine excavation based on stratum identification as described in claim 1, characterized in that, The stratigraphic categories include homogeneous hard rock, boulders / objects, cohesive soft soil, and composite strata with soft upper layers and hard lower layers.

6. The adaptive control method for tunnel boring machine excavation based on stratum identification as described in claim 5, characterized in that, If the formation is identified as homogeneous hard rock, the advance speed is adjusted to keep the instantaneous penetration within the set range; the rotation speed is disturbed every set time interval, and the mechanical rock breaking specific energy is calculated in real time. If the acceleration causes the mechanical rock breaking specific energy to decrease, the rotation speed is increased; otherwise, the rotation speed is decreased, so that the tunnel boring machine works at the point of lowest energy consumption. If a boulder / foreign object is identified, the cutter head speed will be forcibly reduced and the total thrust will be limited. If the soil is identified as cohesive soft soil, the relationship between the mechanical rock breaking specific energy and the vibration energy is monitored in real time. When the mechanical rock breaking specific energy increases beyond the set specific gravity and the vibration signal attenuates beyond the set specific gravity, the cutter head speed is increased and the cutter head flushing function is activated. If the geological formation is identified as a composite stratum with a soft upper layer and a hard lower layer, the advance speed is limited, and the thrust difference between the upper and lower section hydraulic cylinders of the tunnel boring machine is dynamically adjusted to ensure the stability of the tunnel boring machine's construction posture.

7. A shield tunneling adaptive control device based on stratum identification, characterized in that, include: The real-time data acquisition module is configured to collect the cutting vibration signal of the shield machine cutterhead and the tunneling status parameters in real time during the tunneling process; The spatial slice construction module is configured to monitor the tunnel boring machine's advance distance in real time, and obtain a spatial slice for each fixed advance distance. For the current spatial slice, a data pair of vibration signal characteristics and tunneling load characteristics is constructed. The stratum category identification module is configured to take vibration signal feature and tunneling load feature data pairs as input, and use a pre-trained stratum identification model to obtain the current stratum category; The tunneling parameter optimization module is configured to optimize the cutterhead's real-time tunneling parameters based on the current geological formation type.

8. An electronic device, characterized in that, It includes a processor and a memory, the memory storing computer instructions that, when executed by the processor, cause the electronic device to perform the method of any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method of any one of claims 1 to 6.

10. A computer program product, the computer program product comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 6.

Images