A Dynamic Optimization Method for Tunneling Parameters of Shield Machines in Karst Strata Based on Industrial Big Data

By analyzing the time-frequency data of the tunnel boring machine and optimizing the model, the problems of structural resonance and earth pressure oscillation during tunnel boring in karst strata were solved, and dynamic optimization of the tunnel boring machine parameters was achieved, reducing the risk and probability of instability.

CN123115778APending Publication Date: 2026-11-17CCCC SECOND HIGHWAY ENG CO LTD
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202611242269.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-17
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

When tunneling in karst formations, the difference in the mechanical properties of the strata on both sides of the cutterhead leads to a load phase difference, which can cause structural resonance risks and bistable oscillations in the earth pressure balance system. Existing technologies make it difficult to achieve dynamic optimization.

Method used

By collecting industrial data during the tunnel boring machine's excavation process, performing time-domain and frequency-domain analysis, extracting the peak torque frequency and instantaneous phase difference sequence on the high-torque side, calculating the phase-speed coupling characteristic quantity, and combining the time-domain segmentation of the earth pressure time series signal, generating earth pressure bistable oscillation characteristic values, inputting the tunneling parameter dynamic optimization model, and outputting the cutterhead speed and earth pressure correction values.

Benefits of technology

It effectively reduced the risk of cutterhead structure resonance, identified and suppressed minute-level low-frequency oscillations of the earth pressure balance system, reduced the risk of local instability at the tunnel face, and achieved dynamic optimization of tunnel boring parameters.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123115778A_ABST
    Figure CN123115778A_ABST
Patent Text Reader

Abstract

This invention discloses a dynamic optimization method for tunneling parameters of shield tunneling machines in karst strata based on industrial big data, relating to the field of intelligent control technology for shield tunnel construction. The method includes collecting time-series signals of torque in the left and right zones of the cutterhead, soil pressure, and cutterhead rotation speed; performing time-domain and frequency-domain analysis to extract the peak torque frequency on the high-torque side and calculate the instantaneous phase difference sequence, thereby calculating the phase-rotation coupling characteristic quantity and matching the cutterhead rotation speed adjustment amount to avoid resonance; performing time-domain segmentation on the soil pressure signal to extract the high-pressure steady-state period and the low-pressure steady-state period, calculating the alternation period and transition slope to generate earth pressure bistable oscillation characteristic values; constructing a feature vector, inputting it into the dynamic optimization model of tunneling parameters, and outputting the corrected values ​​of cutterhead rotation speed and soil pressure setpoint. This invention can reduce the risk of resonance caused by alternating eccentric load torque in the cutterhead structure and suppress low-frequency bistable oscillations of the earth pressure balance system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for shield tunnel construction, and in particular to a method for dynamic optimization of shield machine tunneling parameters in karst formations based on industrial big data. Background Technology

[0002] As urban rail transit construction extends into karst-developed areas, the application of shield tunneling in karst formations is increasing. Karst formations are highly heterogeneous, with randomly distributed caves, dissolution fissures, and weak infill materials, posing significant challenges to shield tunneling. Grouting reinforcement is a common method for treating karst formations, improving the strength and stability of the formation by injecting grout into caves and fissures. However, due to the highly random diffusion paths of dissolution channels in karst formations, the actual diffusion range of the grout is difficult to control precisely, often resulting in a special "semi-reinforced, semi-original" interface at the edge of the grouting reinforcement. When the shield tunnels to this interface, one side of the cutterhead enters the grouting reinforcement zone, while the other side remains in the original karst formation. The difference in the mechanical properties of the two sides causes the cutterhead to bear asymmetrical loads.

[0003] Chinese invention patent application CN119917808A, published on May 2, 2025, discloses a method and system for predicting tunneling parameters in karst formations based on federated learning. It establishes a three-dimensional distribution model of karst caves by constructing a deep learning classification model combined with multimodal data. Microscopic feature indicators are input into a parameter mapping model to establish the intrinsic relationship between microscopic features and tunneling parameters, thereby outputting tunneling parameter adjustment commands. This invention patent application primarily focuses on tunneling parameter prediction based on geological survey data. Its parameter mapping model uses stratum microscopic features as input and outputs static tunneling parameter adjustments. It does not model or analyze the phase difference of the loads on both sides of the cutterhead under the "semi-reinforced, semi-original" interface condition and the resulting dynamic response problem. It also lacks a mechanism for identifying and processing the bistable oscillation phenomenon that occurs in the earth pressure balance system under this condition, and cannot dynamically correct the cutterhead speed and earth chamber pressure during tunneling based on real-time acquired torque and earth pressure time-series signals. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a dynamic optimization method for tunneling parameters of shield tunneling machines in karst strata based on industrial big data. It aims to solve the problems of structural resonance risk caused by load phase difference due to the difference in the mechanical properties of the strata on both sides of the cutterhead when the shield tunneling machine reaches the edge of the grouting reinforcement zone, and the risk of face instability caused by the bistable oscillation of the earth pressure balance system. The method achieves dynamic optimization of tunneling parameters through industrial big data.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a dynamic optimization method for tunneling parameters of shield tunneling machines in karst strata based on industrial big data, comprising the following steps:

[0006] S1. Collect industrial data during the tunneling process of the tunnel boring machine. The industrial data includes the torque timing signal of the left zone of the cutterhead, the torque timing signal of the right zone of the cutterhead, the soil chamber pressure timing signal, and the cutterhead rotation speed timing signal.

[0007] S2. Perform time-domain and frequency-domain analysis on the torque timing signals of the left and right regions of the cutterhead, extract the peak torque frequency on the high-torque side, and calculate the instantaneous phase difference sequence. Calculate the phase-speed coupling characteristic quantity based on the instantaneous phase difference sequence and the cutterhead speed timing signal. Use the phase-speed coupling characteristic quantity as the optimization target to reverse match and avoid the cutterhead structure resonance by adjusting the cutterhead speed.

[0008] S3. Perform time-domain segmentation on the earth chamber pressure time sequence signal to obtain pressure segment time periods. Extract the time periods from each pressure segment time period where the pressure value is continuously higher than the preset first pressure threshold within the preset steady state discrimination time window as the high pressure steady state time period and the low pressure steady state time period where the pressure value is continuously lower than the preset second pressure threshold.

[0009] S4. Calculate the alternation period and transition slope between the high-pressure steady-state period and the low-pressure steady-state period. Multiply the alternation period and transition slope to generate the earth pressure bistable oscillation characteristic value. Based on the cutterhead speed adjustment, the earth pressure bistable oscillation characteristic value, and the peak torque frequency on the high torque side, construct the characteristic vector of the interface tunneling condition.

[0010] S5. Input the feature vector into the pre-constructed dynamic optimization model of tunneling parameters, and output the correction value of the cutterhead speed and the correction value of the soil chamber pressure setting under the current interface tunneling conditions.

[0011] S6. The cutter head speed adjustment is updated with the cutter head speed correction value to obtain the optimized cutter head speed, and the soil chamber pressure setting value is updated with the soil chamber pressure setting value correction value to obtain the optimized soil chamber pressure setting value.

[0012] Furthermore, the industrial data collected during the tunnel boring machine's excavation process includes the torque timing signal of the left cutterhead zone, the torque timing signal of the right cutterhead zone, the soil chamber pressure timing signal, and the cutterhead rotation speed timing signal, including:

[0013] The timing signal of the cutterhead rotation speed is acquired by the encoder of the cutterhead drive motor of the tunnel boring machine;

[0014] The original torque voltage signal of the left zone during the preset tunneling cycle is collected by the torque sensor in the left zone of the tunnel boring machine cutterhead.

[0015] The original torque voltage signal of the right zone during the preset tunneling cycle is collected by the torque sensor in the right zone of the tunnel boring machine cutterhead.

[0016] Obtain the pre-calibrated calibration coefficients of the left-zone torque sensor and the right-zone torque sensor;

[0017] The original torque voltage signal of the left zone is linearly converted by the calibration coefficient of the left zone torque sensor to generate the initial torque timing signal of the left zone.

[0018] The original torque voltage signal of the right zone is linearly converted by the calibration coefficient of the right zone torque sensor to generate the initial torque timing signal of the right zone.

[0019] Digital low-pass filtering is performed on the initial torque timing signal of the left zone and the initial torque timing signal of the right zone to obtain the torque timing signal of the left zone and the torque timing signal of the right zone of the cutterhead.

[0020] The original analog current signal of the soil chamber pressure is acquired by the pressure transmitter on the upper part of the soil chamber partition.

[0021] The original analog current signal of the earth chamber pressure is converted from analog to digital to generate the original digital sequence of the earth chamber pressure.

[0022] Median filtering is applied to the original digital sequence of earth pressure to generate the earth pressure time sequence signal.

[0023] Furthermore, the step of performing time-domain and frequency-domain analysis on the torque timing signals of the left and right regions of the cutterhead, extracting the peak torque frequency on the high-torque side, and calculating the instantaneous phase difference sequence includes:

[0024] A moving average operation is performed on the torque timing signals of the left and right regions of the cutter head to generate smoothed torque sequences for the left and right regions.

[0025] The arithmetic mean of the left-zone torque smoothing sequence and the right-zone torque smoothing sequence is calculated to generate the mean torque of the left zone and the mean torque of the right zone.

[0026] Compare the average torque values ​​of the left and right zones, and mark the side with the larger value as the high torque side and the side with the smaller value as the low torque side.

[0027] Fast Fourier transform is performed on the torque time-series signal corresponding to the high torque side and the torque time-series signal corresponding to the low torque side to generate the torque spectrum of the high torque side and the torque spectrum of the low torque side.

[0028] Extract the frequency corresponding to the maximum amplitude from the torque spectrum on the high torque side to obtain the peak torque frequency on the high torque side;

[0029] Perform Hilbert transform on the torque timing signals corresponding to the high torque side and the low torque side to generate analytical signals for the high torque side and the low torque side.

[0030] Instantaneous phases are extracted from the high-torque side analytical signal and the low-torque side analytical signal to generate high-torque side instantaneous phase sequences and low-torque side instantaneous phase sequences;

[0031] The instantaneous phase difference sequence is generated by performing a difference operation between the instantaneous phase sequence on the high torque side and the instantaneous phase sequence on the low torque side.

[0032] Furthermore, the step of calculating the phase-speed coupling characteristic quantity based on the instantaneous phase difference sequence and the cutter head speed timing signal, and using the phase-speed coupling characteristic quantity as the optimization target to reverse match and avoid cutter head structural resonance, includes:

[0033] Perform a difference operation on the instantaneous phase difference sequence to generate an instantaneous phase difference rate of change sequence;

[0034] Perform absolute value calculation on the instantaneous phase difference change rate sequence to generate an instantaneous phase difference change rate amplitude sequence;

[0035] The unit conversion of the cutter head speed timing signal is performed to generate the cutter head speed frequency signal. The division operation between the high torque side torque peak frequency and the cutter head speed frequency signal is performed to generate the torque-speed frequency ratio sequence.

[0036] The arithmetic mean of the corresponding point product operation is performed on the instantaneous phase difference change rate amplitude sequence and the torque-speed frequency ratio sequence to generate the phase-speed coupling characteristic quantity;

[0037] Obtain the set of natural frequencies from the finite element modal analysis results of the tunnel boring machine cutterhead structure;

[0038] The current sampled value of the cutter head speed timing signal is used as the initial speed. Within the allowable adjustment range of the cutter head speed, a candidate speed sequence is generated with a preset speed step size. Candidate speed values ​​in the candidate speed sequence whose absolute value of the difference between the converted candidate speed frequency value and the natural frequency set is less than a preset neighborhood radius are removed to generate a feasible speed sequence.

[0039] For each feasible speed value in the feasible speed sequence, calculate the torque-speed frequency ratio corresponding to the feasible speed value, and calculate the arithmetic mean after performing corresponding point product operation on the torque-speed frequency ratio and the instantaneous phase difference change rate amplitude sequence, which is used as the estimated value of the phase-speed coupling characteristic quantity.

[0040] The feasible speed value with the smallest estimated value of the phase-speed coupling characteristic is selected as the target speed value.

[0041] The difference between the target speed and the initial speed is calculated to generate the cutter head speed adjustment amount.

[0042] Furthermore, the process of performing time-domain segmentation on the earth pressure time-series signal to obtain pressure segment time periods, and extracting the time periods from each pressure segment time period where the pressure value is continuously higher than a preset first pressure threshold within a preset steady-state discrimination time window as high-pressure steady-state time periods and low-pressure steady-state time periods where the pressure value is continuously lower than a preset second pressure threshold, includes:

[0043] The soil pressure timing signal is low-pass filtered to generate a soil pressure filtered signal.

[0044] Median filtering is applied to the soil pressure filter signal to generate a denoised soil pressure signal.

[0045] A first-order difference operation is performed on the denoised earth pressure signal to generate an earth pressure difference sequence.

[0046] Zero-crossing detection is performed on the differential pressure sequence of the earth chamber, and the indexes of sampling points in the differential pressure sequence of the earth chamber where the sign changes are extracted to generate a set of zero-crossing indexes;

[0047] Using the index position in the zero-crossing index set as the dividing boundary, the soil pressure denoising signal is segmented to generate pressure segment time periods;

[0048] Remove pressure segments whose duration is less than the preset steady-state discrimination time window from the pressure segment time periods, and generate effective pressure segment time periods;

[0049] For each effective pressure segment, extract the pressure values ​​of all sampling points within the effective pressure segment, and mark the continuous intervals where the pressure value is continuously higher than the preset first pressure threshold and the duration exceeds the preset steady-state discrimination time window as high-pressure steady-state periods;

[0050] For each effective pressure segment, extract the pressure values ​​of all sampling points within the effective pressure segment, and mark the continuous interval where the pressure value is continuously lower than the preset second pressure threshold and the duration exceeds the preset steady-state discrimination time window as the low-pressure steady-state period.

[0051] Furthermore, the alternation period and transition slope of the high-pressure steady-state period and the low-pressure steady-state period are calculated, and the alternation period and transition slope are multiplied to generate the earth pressure bistable oscillation characteristic value; the feature vector of the interface tunneling condition is constructed based on the cutterhead speed adjustment, the earth pressure bistable oscillation characteristic value, and the peak torque frequency on the high torque side, including:

[0052] Arrange all high-pressure steady-state periods in ascending order of their start times to generate a high-pressure steady-state period sequence.

[0053] Arrange all low-pressure steady-state periods in ascending order of their start times to generate a low-pressure steady-state period sequence.

[0054] The time interval between the start times of two adjacent high-pressure steady-state periods in the high-pressure steady-state period sequence is taken as a complete alternation cycle. The difference calculation is performed on the start times of all adjacent high-pressure steady-state periods in the high-pressure steady-state period sequence to generate an alternation cycle sequence.

[0055] For each high-pressure steady-state period in the high-pressure steady-state period sequence, the first low-pressure steady-state period that is temporally adjacent to the high-pressure steady-state period and whose start time is later than the end time of the high-pressure steady-state period is extracted. The denoised earth pressure signal between the end time of the high-pressure steady-state period and the start time of the low-pressure steady-state period is used as the transition interval signal. The transition interval signal is subjected to least squares linear fitting, and the absolute value of the slope of the fitted line is used as the transition slope to generate a transition slope sequence.

[0056] The arithmetic mean of all alternating periods in the alternating period sequence is calculated to generate the average alternating period.

[0057] The average transition slope is generated by averaging all transition slopes in the transition slope sequence.

[0058] The average alternation period and the average transition slope are multiplied to generate the earth pressure bistable oscillation characteristic value;

[0059] The cutterhead speed adjustment, the earth pressure bistable oscillation characteristic value, and the peak torque frequency of the high torque side are combined into a three-dimensional vector to generate the characteristic vector of the interface tunneling condition.

[0060] Furthermore, the step of inputting the feature vector into the pre-constructed dynamic optimization model of tunneling parameters and outputting the correction values ​​of the cutterhead rotation speed and the soil chamber pressure setpoint under the current interface tunneling conditions includes:

[0061] The dynamic optimization model for tunneling parameters is obtained. The input feature dimension of the dynamic optimization model for tunneling parameters is the same as the dimension of the feature vector of the interface tunneling working condition. The output of the dynamic optimization model for tunneling parameters is the correction value of the cutterhead speed and the correction value of the soil chamber pressure setpoint.

[0062] Input the interface tunneling condition feature vector into the tunneling parameter dynamic optimization model to generate the cutterhead speed correction value and the soil chamber pressure set value correction value.

[0063] Furthermore, the pre-construction process of the dynamic optimization model for tunneling parameters includes:

[0064] A tunneling performance cost function is constructed with cutterhead speed and soil chamber pressure as decision variables. The tunneling performance cost function is composed of a weighted sum of tunneling speed, cutterhead torque fluctuation, and soil chamber pressure deviation after normalization. The weights are determined by a combination of the analytic hierarchy process and the entropy weight method.

[0065] Obtain feature vector samples of interface tunneling conditions from historical tunneling cycles;

[0066] For each interface tunneling condition feature vector sample, with the goal of minimizing the tunneling performance cost function, the Bayesian optimization algorithm is used to search for the optimal cutterhead speed adjustment and the optimal soil chamber pressure adjustment within the range of cutterhead speed adjustment and soil chamber pressure adjustment.

[0067] Using the interface tunneling working condition feature vector samples as input features, and the optimal cutterhead speed adjustment and optimal soil chamber pressure adjustment obtained by the search as output labels, a training sample set is constructed.

[0068] Construct a fully connected neural network consisting of an input layer, a hidden layer, and an output layer. The hidden layer uses a linear rectified function as the activation function, and the output layer has two output nodes.

[0069] The training sample set is input into a fully connected neural network for training. The mean square error between the output values ​​of two nodes in the output layer and their corresponding labels is used as the loss function. The network weights are updated through backpropagation and iterated until the loss function converges, generating a dynamic optimization model for the tunneling parameters.

[0070] Furthermore, the step of updating the cutter head speed adjustment amount with the cutter head speed correction value to obtain the optimized cutter head speed, and updating the soil chamber pressure setting value with the soil chamber pressure setting value correction value to obtain the optimized soil chamber pressure setting value, includes:

[0071] Obtain the current sample value of the tool turret speed timing signal;

[0072] The current sampled value of the cutter head speed timing signal is added to the cutter head speed adjustment amount to generate the reference cutter head speed;

[0073] The base cutter head speed and the cutter head speed correction value are added together to generate the optimized cutter head speed.

[0074] Obtain the current sample value of the earth chamber pressure timing signal;

[0075] The current sampled value of the earth chamber pressure timing signal is added to the correction value of the earth chamber pressure setpoint to generate an optimized earth chamber pressure setpoint.

[0076] The method for dynamic optimization of tunneling parameters of shield tunneling machines in karst strata based on industrial big data provided by this invention has the following beneficial effects:

[0077] 1. This invention performs time-domain and frequency-domain analysis on the torque timing signals of the left and right regions of the cutterhead, extracts the peak torque frequency on the high-torque side, calculates the instantaneous phase difference sequence, and then calculates the phase-speed coupling characteristic quantity. This allows for reverse matching to avoid cutterhead structural resonance by adjusting the cutterhead speed. This method incorporates load phase differences into the optimization of tunneling parameters, rather than focusing solely on load amplitude differences. This effectively reduces the risk of alternating eccentric load torque triggering modal resonance in the cutterhead structure, thus overcoming the shortcomings of existing research in load dynamic response analysis.

[0078] 2. This invention performs time-domain segmentation on the earth pressure time-series signal, extracts the high-pressure steady-state period and the low-pressure steady-state period, calculates the alternation period and transition slope, and generates earth pressure bistable oscillation characteristic values. These, along with the cutterhead speed adjustment and the peak torque frequency on the high-torque side, constitute the feature vector of the interface tunneling condition. After inputting into the dynamic optimization model of tunneling parameters, the corrected values ​​for the cutterhead speed and earth pressure setpoint are output. This method can identify minute-level low-frequency oscillations in the earth pressure balance system under the "semi-reinforced, semi-original" interface condition, avoiding misjudging them as normal fluctuations. By dynamically correcting the earth pressure setpoint, bistable oscillations are suppressed, reducing the risk of local instability at the tunnel face. Attached Figure Description

[0079] Figure 1 This is a flowchart illustrating the method for dynamic optimization of tunneling parameters of shield tunneling machines in karst formations based on industrial big data, as described in this invention.

[0080] Figure 2 This is a flowchart of the earth pressure steady-state time period extraction process of the present invention;

[0081] Figure 3 This is a flowchart illustrating the construction process of earth pressure characteristics and feature vectors in this invention.

[0082] Figure 4 This is a flowchart illustrating the construction process of the tunneling parameter optimization model for this invention. Detailed Implementation

[0083] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0084] Please see Figures 1-4 This invention provides a method for dynamic optimization of tunneling parameters of tunnel boring machines in karst strata based on industrial big data, comprising the following steps:

[0085] S1. Collect industrial data during the tunneling process of the tunnel boring machine. The industrial data includes the torque timing signal of the left zone of the cutterhead, the torque timing signal of the right zone of the cutterhead, the soil chamber pressure timing signal, and the cutterhead rotation speed timing signal.

[0086] In this embodiment, the industrial data collected during the tunnel boring machine's excavation process includes the torque timing signal of the left cutterhead zone, the torque timing signal of the right cutterhead zone, the soil chamber pressure timing signal, and the cutterhead rotation speed timing signal, including:

[0087] The timing signal of the cutterhead rotation speed is collected by the encoder of the cutterhead drive motor of the tunnel boring machine.

[0088] Specifically, the cutter head drive motor encoder refers to the incremental photoelectric encoder installed on the shaft of the cutter head drive motor. Its internal code disk outputs two pulse sequences with a 90° phase difference when it rotates with the shaft. The pulse count is proportional to the angular displacement of the shaft. The cutter head speed timing signal refers to the speed value sequence arranged in the order of sampling time at equal intervals after proportional conversion of the encoder pulse count per unit time. The unit is revolutions per minute.

[0089] The original torque voltage signal of the left zone during the preset tunneling cycle is collected by the torque sensor in the left zone of the tunnel boring machine cutterhead.

[0090] Specifically, the left-side torque sensor of the cutterhead refers to a resistance strain gauge torque sensor installed on the input shaft end of the drive reducer on the left side of the cutterhead. It contains a Wheatstone bridge composed of four strain gauges attached to the elastic shaft section. When the shaft section is subjected to torque and generates shear strain, the bridge outputs a differential voltage signal that is proportional to the torque. The range is determined according to the rated torque of the tunnel boring machine.

[0091] The preset tunneling cycle refers to a complete propulsion stroke from the start of tunneling by the shield machine until it stops tunneling. Its length is determined by the stroke of the propulsion cylinder and the tunneling speed. The original torque voltage signal in the left zone refers to the unprocessed differential voltage amplitude sequence that the sensor continuously collects and records at a fixed sampling frequency within the preset tunneling cycle.

[0092] The original torque voltage signal of the right zone during the preset tunneling cycle is collected by the torque sensor in the right zone of the tunnel boring machine cutterhead.

[0093] Specifically, the right-side torque sensor of the cutterhead refers to the resistance strain gauge torque sensor installed on the input shaft end of the drive reducer on the right side of the cutterhead. Its internal structure and working principle are the same as those of the left-side torque sensor. Both are based on the strain bridge to convert the torque on the shaft section into a differential voltage output. The original torque voltage signal of the right-side zone refers to the sequence of original voltage amplitude values ​​continuously collected and recorded by the sensor at the same sampling frequency within the same preset tunneling cycle as the left zone.

[0094] Obtain the pre-calibrated calibration coefficients of the left-zone torque sensor and the right-zone torque sensor.

[0095] Specifically, pre-calibration refers to installing the torque sensor on a standard torque calibration platform, applying several levels of known standard torque values ​​to the sensor, and simultaneously recording the sensor's output voltage at each standard torque value. The least squares method is used to perform univariate linear regression on the paired voltage-torque data to obtain the linear relationship equation between the output voltage and the input torque. This equation includes two parameters: slope coefficient and intercept coefficient. The calibration coefficients of the left and right torque sensors refer to the slope coefficient and intercept coefficient of the left and right sensors, respectively, and are used to convert the voltage signals of the two sensors into torque physical quantities with consistent dimensions.

[0096] The original torque voltage signal in the left zone is linearly converted by the calibration coefficient of the left zone torque sensor to generate the initial torque timing signal in the left zone.

[0097] Specifically, linear conversion refers to multiplying each voltage sample value in the original torque voltage sequence of the left zone by the slope coefficient in the sensor calibration coefficient and then adding the intercept coefficient to complete the dimensional conversion from voltage quantity to torque physical quantity unit - kilonewton-meter. The initial torque time sequence signal of the left zone refers to the left torque value sequence output after the above calibration conversion, and each value is aligned with the corresponding original voltage sampling time.

[0098] The original torque voltage signal of the right zone is linearly converted by the calibration coefficient of the right zone torque sensor to generate the initial torque timing signal of the right zone.

[0099] Specifically, linear conversion refers to multiplying each voltage sample value in the original torque voltage sequence of the right zone by the slope coefficient in the sensor calibration coefficient and then adding the intercept coefficient to convert the voltage amplitude into a torque value in kilonewton-meters. The initial torque timing signal of the right zone refers to the right torque value sequence output after the above calibration conversion, and its time base is consistent with that of the initial torque timing signal of the left zone.

[0100] The initial torque timing signals of the left and right regions are digitally low-pass filtered to obtain the torque timing signals of the left and right regions of the cutterhead.

[0101] Specifically, digital low-pass filtering employs a finite impulse response filter, using a cutoff frequency ten times the rotational frequency corresponding to the cutterhead's rated speed. This attenuates high-frequency components of mechanical vibration and sensor electronic noise above the cutoff frequency in the initial torque timing signal, while retaining effective low-frequency components containing the characteristics of the formation cutting load. The resulting left and right torque timing signals, after filtering, eliminate high-frequency interference and accurately reflect the time-varying waveforms of the instantaneous cutting resistance torque experienced by the left and right sides of the cutterhead.

[0102] The original analog current signal of the soil chamber pressure is collected by the pressure transmitter on the upper part of the soil chamber partition.

[0103] Specifically, the soil chamber partition refers to the steel partition structure between the soil chamber of the tunnel boring machine and the outlet of the screw conveyor. The upper pressure transmitter refers to the piezoresistive pressure transmitter installed at the pressure measuring hole at the top of the partition. Its internal diffused silicon sensing element converts the fluid pressure transmitted by the slag in the soil chamber into a standard analog current signal of 4-20mA. The range is selected according to the maximum design pressure of the soil chamber. The original analog current signal of the soil chamber pressure refers to the sequence of current values ​​directly output by the transmitter without any processing. The current amplitude at each moment has a linear relationship with the pressure in the soil chamber at that moment.

[0104] The original analog current signal of the earth chamber pressure is converted from analog to digital to generate the original digital sequence of the earth chamber pressure.

[0105] Specifically, analog-to-digital conversion refers to using a successive approximation analog-to-digital converter to perform equally spaced discrete sampling of the analog current signal at a sampling frequency of 250Hz, quantizing the continuous amplitude into a 12-bit binary digital quantity. The sampling frequency is set according to the effective bandwidth of the earth pressure signal in combination with the Nyquist sampling theorem to avoid frequency domain aliasing. The original digital sequence of earth pressure refers to the unfiltered digital quantized value sequence output by the analog-to-digital converter.

[0106] Median filtering is applied to the original digital sequence of earth pressure to generate the earth pressure time sequence signal.

[0107] Specifically, median filtering involves using a sliding window with a window length of 11 sampling points. The median of the sampled values ​​arranged in ascending order within the window is taken as the reconstructed output center. The window length is pre-set to an odd number based on the fluctuation period of the earth pressure signal and the tunneling control response time to ensure that there is a unique sampling point at the center of the window. The earth pressure time series signal refers to the pressure value sequence output after median filtering. This sequence completely preserves the macroscopic trend and steady-state amplitude of the earth pressure while eliminating pulse spikes and random noise in the original signal, providing reliable input data for subsequent segmentation and identification of the steady-state earth pressure period.

[0108] S2. Perform time-domain and frequency-domain analysis on the torque timing signals of the left and right regions of the cutterhead, extract the peak torque frequency on the high-torque side, and calculate the instantaneous phase difference sequence. Calculate the phase-speed coupling characteristic quantity based on the instantaneous phase difference sequence and the cutterhead speed timing signal. Use the phase-speed coupling characteristic quantity as the optimization target to reverse match and avoid the cutterhead structure resonance by adjusting the cutterhead speed.

[0109] In this embodiment, the step of performing time-domain and frequency-domain analysis on the torque timing signals of the left and right regions of the cutterhead, extracting the peak torque frequency on the high-torque side, and calculating the instantaneous phase difference sequence includes:

[0110] A moving average operation is performed on the torque timing signals of the left and right regions of the cutterhead to generate smoothed torque sequences for the left and right regions.

[0111] Specifically, the moving average operation uses a rectangular sliding window of fixed length. The torque values ​​of all sampling points within the window are summed and divided by the window length. The resulting arithmetic mean is used as the reconstructed output value at the center of the window. The window length is the number of sampling points corresponding to one revolution of the cutterhead, calculated from the measured cutterhead speed and signal sampling frequency at the current moment. The number of sampling points is rounded to ensure that the window length is odd, thus ensuring that there is a unique sampling point at the center of the window. The left and right torque smoothing sequences are numerical sequences of the left and right regions, respectively, after moving average processing and arranged according to the original sampling time sequence. This sequence reduces random fluctuations and short-term spikes in the signal while preserving the overall trend of torque change.

[0112] The arithmetic mean of the left-zone torque smoothing sequence and the right-zone torque smoothing sequence is calculated to generate the mean torque of the left zone and the mean torque of the right zone.

[0113] Specifically, the arithmetic mean operation refers to summing the torque values ​​of all sampling points in the left zone torque smoothing sequence and dividing by the total number of sampling points to obtain the average torque amplitude of the left zone in the entire preset tunneling cycle. The same summation and averaging operation is used in the right zone. The average torque values ​​of the left zone and the right zone respectively represent the central trend of the torque borne by the left and right sides of the cutterhead in the entire tunneling cycle.

[0114] The average torque value in the left zone is compared with the average torque value in the right zone. The side with the larger value is marked as the high torque side, and the side with the smaller value is marked as the low torque side.

[0115] Specifically, the comparison involves determining the magnitude of the average torque values ​​in the left and right zones. If the left zone average is greater than the right zone average, the left side of the cutterhead is marked as the high-torque side and the right side as the low-torque side. Conversely, if the right zone average is greater than the left zone average, the markings are reversed. If they are equal, the left side is marked as the high-torque side by default. The high-torque side refers to the cutterhead region experiencing greater average cutting resistance during the tunneling cycle, while the low-torque side refers to the cutterhead region experiencing less average cutting resistance. This marking result guides the differentiation of signals from both sides in subsequent frequency domain analysis and instantaneous phase difference calculations.

[0116] Fast Fourier transform is performed on the torque timing signals corresponding to the high torque side and the low torque side to generate the torque spectrum of the high torque side and the torque spectrum of the low torque side.

[0117] Specifically, the torque timing signal corresponding to the high torque side refers to the original torque sampling sequence selected based on the marking results for that side, and the torque timing signal corresponding to the low torque side is similarly selected from the original torque sampling sequence for that side. The original sequence is used instead of the smoothed sequence to ensure that high-frequency harmonic components are not lost in the frequency domain analysis.

[0118] Fast Fourier Transform (FFT) refers to the transformation of a discrete time-domain signal into a discrete frequency-domain spectrum using a radix-2 time-division decimation algorithm. The output is a complex sequence, where each complex number corresponds to a frequency component. The torque spectrum on the high-torque side and the torque spectrum on the low-torque side refer to the amplitude-frequency characteristic curves obtained by taking the modulus value after the original torque signals on both sides have undergone FFT, respectively. These curves are used to identify the amplitude distribution of each frequency component in torque fluctuations.

[0119] The frequency corresponding to the maximum amplitude is extracted from the torque spectrum of the high torque side to obtain the peak torque frequency of the high torque side.

[0120] Specifically, the maximum amplitude refers to the frequency value corresponding to the largest amplitude obtained by traversing all spectral lines within the effective frequency band of the torque spectrum on the high torque side. The peak frequency of the high torque side represents the frequency component with the most concentrated energy in the torque signal on the high torque side, reflecting the rhythm of the periodic change of the cutting resistance of the formation on the high torque side.

[0121] Hilbert transform is performed on the torque timing signals corresponding to the high torque side and the low torque side to generate analytical signals for the high torque side and the low torque side.

[0122] Specifically, the Hilbert transform refers to performing a linear convolution operation on a real signal to transform it into its orthogonal imaginary part signal. The high-torque side analytic signal refers to a complex signal sequence constructed using the original torque sequence on the high-torque side as the real part and the Hilbert transform result of the real part as the imaginary part. The low-torque side analytic signal is constructed in the same way. Both retain the amplitude and phase information of the original signal. The construction of the analytic signal makes it possible to extract the instantaneous phase, providing a mathematical basis for calculating the phase difference of the torque on both sides.

[0123] Instantaneous phases are extracted from the high-torque side analytical signal and the low-torque side analytical signal to generate high-torque side instantaneous phase sequences and low-torque side instantaneous phase sequences.

[0124] Specifically, instantaneous phase extraction refers to calculating the arctangent function value of the ratio of the imaginary part to the real part of the complex number at each sampling moment in the analytical signal, and obtaining the instantaneous phase angle at that moment. The high-torque side instantaneous phase sequence and the low-torque side instantaneous phase sequence are numerical sequences composed of the instantaneous phase angles of the high-torque side and low-torque side analytical signals at each sampling moment arranged in their original time sequence, respectively.

[0125] The instantaneous phase difference sequence is generated by performing a difference operation between the instantaneous phase sequence on the high torque side and the instantaneous phase sequence on the low torque side.

[0126] Specifically, the difference operation refers to the subtraction operation performed on two instantaneous phase values ​​with the same time index in the instantaneous phase sequence on the high torque side and the instantaneous phase sequence on the low torque side. The instantaneous phase difference sequence refers to the sequence formed by arranging the phase difference values ​​at all sampling times in the original time sequence. This sequence reflects the relative misalignment of the torque fluctuations on the left and right sides of the cutterhead on the time axis.

[0127] In this embodiment, the step of calculating the phase-speed coupling characteristic quantity based on the instantaneous phase difference sequence and the cutter head speed timing signal, and using the phase-speed coupling characteristic quantity as the optimization target to reverse match and avoid cutter head structural resonance, includes:

[0128] Perform a difference operation on the instantaneous phase difference sequence to generate an instantaneous phase difference rate of change sequence.

[0129] Specifically, the instantaneous phase difference sequence refers to the time series of instantaneous phase difference values ​​of the left and right torque regions, generated in the previous step and arranged in the order of sampling times. The difference operation refers to subtracting the instantaneous phase difference value of the previous sampling time from the instantaneous phase difference value of each sampling time using the method of subtracting the previous term from the subsequent term, resulting in the change in phase difference at that moment. The instantaneous phase difference change rate sequence refers to the sequence of phase difference changes at each sampling time arranged in the original time sequence. This sequence reflects the rate of change of the torque phase difference between the left and right sides of the cutterhead over time.

[0130] Perform absolute value calculation on the instantaneous phase difference rate of change sequence to generate an instantaneous phase difference rate of change amplitude sequence.

[0131] The unit conversion of the cutter head speed timing signal is performed to generate the cutter head speed frequency signal. The torque peak frequency on the high torque side is divided by the cutter head speed frequency signal to generate the torque-speed frequency ratio sequence.

[0132] Specifically, unit conversion refers to converting the cutter head speed timing signal, expressed in revolutions per minute (RPM), into a rotational frequency value expressed in Hertz (Hz). Division operation refers to performing point-by-point division using the peak torque frequency on the high-torque side as the dividend and the sampled values ​​of the cutter head speed frequency signal at each moment as the divisor. The torque-speed-frequency ratio sequence is a sequence composed of the division results at each sampling moment arranged in their original time sequence. This sequence reflects the multiple relationship between the torque fluctuation frequency and the instantaneous rotational frequency of the cutter head.

[0133] The arithmetic mean of the corresponding point product operation is performed on the instantaneous phase difference change rate amplitude sequence and the torque-speed frequency ratio sequence to generate the phase-speed coupling characteristic quantity.

[0134] Specifically, the corresponding point product operation refers to multiplying the values ​​with the same sampling time index in two sequences, that is, multiplying the amplitude of the instantaneous phase difference change rate at the same moment by the torque-speed frequency ratio. The arithmetic mean is obtained by summing all the corresponding point product results and dividing by the total number of product results. The resulting scalar value is the phase-speed coupling characteristic quantity, which comprehensively characterizes the degree of change of the left and right torque phase difference and the ratio of the torque fluctuation frequency to the cutterhead rotation frequency.

[0135] Obtain the set of natural frequencies from the finite element modal analysis results of the tunnel boring machine cutterhead structure.

[0136] Specifically, finite element modal analysis refers to the process during the design phase of a tunnel boring machine (TBM) cutterhead, where the design unit establishes a three-dimensional finite element model of the cutterhead structure, applies constraint boundary conditions consistent with the actual installation state, and calculates the natural frequencies of the cutterhead structure under each mode shape through modal analysis. The natural frequency set refers to an array composed of the aforementioned natural frequency values. This set is pre-stored in the TBM equipment database for online retrieval. When the excitation frequency of an external alternating load approaches any natural frequency value in the set, the cutterhead structure will resonate. This set is used to determine whether a candidate rotational speed falls into the resonance danger band.

[0137] The current sampled value of the cutter head speed timing signal is used as the initial speed. Within the allowable adjustment range of the cutter head speed, a candidate speed sequence is generated with a preset speed step size. Candidate speed values ​​in the candidate speed sequence whose absolute difference between the converted candidate speed frequency value and the natural frequency set is less than a preset neighborhood radius are removed, thus generating a feasible speed sequence.

[0138] Specifically, the allowable adjustment range of the cutterhead speed refers to the adjustable speed range determined by the lower and upper limits of the rated speed of the main drive motor of the tunnel boring machine. The preset speed step size refers to the interval value between adjacent candidate speeds set according to the frequency resolution of the main drive speed regulating device. This step size determines the fineness of the speed adjustment. The candidate speed frequency value refers to the value of each candidate speed in revolutions per minute converted into a frequency value in Hertz.

[0139] The preset neighborhood radius refers to the half-width of the frequency band that is prohibited from entering on both sides of each natural frequency determined by the modal damping ratio of the cutterhead structure. This width is equal to the natural frequency value multiplied by the damping ratio and then multiplied by the safety factor. When the absolute value of the difference between the excitation frequency and the natural frequency is less than this radius, it falls into the resonance danger zone. After eliminating the above candidate speeds, the remaining candidate speed values ​​constitute a feasible speed sequence. No speed value in this sequence will trigger resonance of the cutterhead structure on its own.

[0140] For each feasible speed value in the feasible speed sequence, calculate the torque-speed frequency ratio corresponding to the feasible speed value. Then, perform a point-to-point product operation on the torque-speed frequency ratio and the instantaneous phase difference change rate amplitude sequence, and obtain the arithmetic mean as the estimated value of the phase-speed coupling characteristic quantity.

[0141] Specifically, the torque-speed frequency ratio corresponding to a feasible speed value is obtained by performing a division operation with the rotational frequency obtained after unit conversion of the feasible speed value as the divisor and the peak torque frequency on the high torque side as the dividend, resulting in a torque-speed frequency ratio sequence at each sampling time under that speed. This sequence is then multiplied at corresponding points with the instantaneous phase difference change rate amplitude sequence, i.e., the values ​​at the same sampling time are multiplied together, and all product results are summed and divided by the total number of sampling points. The resulting scalar value is the estimated value of the phase-speed coupling characteristic quantity under that feasible speed.

[0142] The feasible speed value with the smallest estimated value of the phase-speed coupling characteristic is selected as the target speed value.

[0143] Specifically, each feasible speed value and its corresponding phase-speed coupling feature value in the feasible speed sequence are selected, and the minimum value is found through numerical comparison. The feasible speed value corresponding to the minimum value is then determined as the target speed value.

[0144] The difference between the target speed and the initial speed is calculated to generate the cutter head speed adjustment amount.

[0145] Specifically, the difference operation refers to subtracting the initial speed value from the target speed value, and the resulting algebraic difference is the tool head speed adjustment amount. When the adjustment amount is positive, it means that the tool head speed needs to be increased, and when it is negative, it means that the tool head speed needs to be decreased.

[0146] S3. Perform time-domain segmentation on the earth chamber pressure time sequence signal to obtain pressure segment time periods. Extract the time periods from each pressure segment time period where the pressure value is continuously higher than the preset first pressure threshold within the preset steady state discrimination time window as the high pressure steady state time period and the low pressure steady state time period where the pressure value is continuously lower than the preset second pressure threshold.

[0147] In this embodiment, the step of performing time-domain segmentation on the earth pressure time-series signal to obtain pressure segment time periods, and extracting the time periods from each pressure segment time period where the pressure value is continuously higher than a preset first pressure threshold within a preset steady-state discrimination time window as high-pressure steady-state time periods and low-pressure steady-state time periods where the pressure value is continuously lower than a preset second pressure threshold, includes:

[0148] The soil pressure timing signal is low-pass filtered to generate a soil pressure filtered signal.

[0149] Specifically, low-pass filtering refers to processing the soil pressure time-series signal using a finite impulse response digital filter. The cutoff frequency is set to the main frequency bandwidth of the effective soil pressure signal, which is determined by the time constants of the extension and retraction of the propulsion cylinder and the speed adjustment of the screw conveyor. High-frequency components of mechanical vibration above the cutoff frequency are filtered out, and the generated soil pressure filtered signal retains the low-frequency trend components of the actual changes in soil pressure.

[0150] Median filtering is applied to the soil pressure filter signal to generate a denoised soil pressure signal.

[0151] Specifically, median filtering refers to using a sliding window with a fixed number of sampling points. The filtered signal samples within the window are sorted in ascending order of value, and the median is taken as the reconstructed output value at the center of the window. The number of sampling points in the window is the number of sampling points corresponding to the duration of the noise pulse; if this value is even, it is incremented by one to ensure a unique sampling point at the center of the window. The soil pressure denoising signal refers to the pressure value sequence output after median filtering. This sequence removes pulse spikes and isolated noise points from the original signal while preserving the overall trend of soil pressure variation.

[0152] A first-order difference operation is performed on the denoised earth pressure signal to generate an earth pressure difference sequence.

[0153] Zero-crossing detection is performed on the differential pressure sequence of the earth chamber, and the indexes of sampling points in the differential pressure sequence where the sign changes are extracted to generate a set of zero-crossing indexes.

[0154] Specifically, zero-crossing detection involves checking the numerical signs of adjacent sampling points in the differential pressure sequence of the earth chamber point. When the zero-crossing point does not fall on a sampling point, if the difference value of the preceding sampling point is positive and the difference value of the following sampling point is negative, then the index of the preceding sampling point is the positive to negative zero-crossing index; if the difference value of the preceding sampling point is negative and the difference value of the following sampling point is positive, then the index of the preceding sampling point is the negative to positive zero-crossing index. When there is a sampling point in the differential sequence with a value exactly zero, the index of that sampling point is directly marked as the zero-crossing index. All detected zero-crossing indices are arranged in ascending order to form a zero-crossing index set.

[0155] Using the index position in the zero-crossing index set as the dividing boundary, the pressure denoising signal of the earth chamber is segmented to generate pressure segment time periods.

[0156] Specifically, segmentation refers to using each index value in the zero-crossing index set as a dividing point. On the time axis of the soil pressure denoising signal, starting from the first sampling point, each zero-crossing index is used as the end point of the current time period and the start point of the next time period. The zero-crossing index itself belongs to the time period before that index. The continuous interval between two adjacent zero-crossing indices is used as a pressure segment time period. The time periods are arranged in chronological order to form a pressure segment time period sequence.

[0157] Pressure segments with a duration shorter than the preset steady-state discrimination time window are removed from the pressure segment time period to generate effective pressure segment time periods.

[0158] Specifically, the preset steady-state discrimination time window refers to a pre-set duration parameter based on the response time of a single pressure adjustment made by the earth pressure balance control system to changes in earth chamber pressure. The value of this time window is equal to three times the response time. The duration refers to the physical time length obtained by multiplying the difference between the end sampling time and the start sampling time of the pressure segment by the sampling interval. Segments with a duration shorter than the preset time window are discarded, and the remaining segments constitute the effective pressure segment sequence.

[0159] For each effective pressure segment, the pressure values ​​of all sampling points within the effective pressure segment are extracted, and the continuous intervals in which the pressure values ​​are continuously higher than the preset first pressure threshold and the duration exceeds the preset steady-state discrimination time window are marked as high-pressure steady-state periods.

[0160] Specifically, the preset first pressure threshold refers to a high steady-state judgment benchmark value pre-set based on the maximum working pressure of the earthwork chamber. This threshold is taken as 80% of the maximum working pressure of the earthwork chamber. The marking refers to checking the pressure sampling values ​​within each effective pressure segment point by point. When the pressure values ​​of multiple consecutive sampling points are all higher than the preset first pressure threshold and the physical time length covered by these sampling points is greater than the preset steady-state judgment time window, the continuous interval is marked as a high-pressure steady-state period.

[0161] For each effective pressure segment, extract the pressure values ​​of all sampling points within the effective pressure segment, and mark the continuous interval where the pressure value is continuously lower than the preset second pressure threshold and the duration exceeds the preset steady-state discrimination time window as the low-pressure steady-state period.

[0162] Specifically, the preset second pressure threshold refers to a low steady-state judgment benchmark value pre-set based on the static soil and water pressure of the stratum. This threshold is taken as 110% of the static soil and water pressure value of the stratum. The marking refers to checking the pressure sampling values ​​within each effective pressure segment time period point by point. When the pressure values ​​of multiple consecutive sampling points are all lower than the preset second pressure threshold and the physical time length covered by these sampling points is greater than the preset steady-state judgment time window, the continuous interval is marked as a low-pressure steady-state period.

[0163] S4. Calculate the alternation period and transition slope between the high-pressure steady-state period and the low-pressure steady-state period. Multiply the alternation period and transition slope to generate the earth pressure bistable oscillation characteristic value. Based on the cutterhead speed adjustment, the earth pressure bistable oscillation characteristic value, and the peak torque frequency on the high torque side, construct the characteristic vector of the interface tunneling condition.

[0164] In this embodiment, the alternation period and transition slope of the high-pressure steady-state period and the low-pressure steady-state period are calculated, and the alternation period and transition slope are multiplied to generate the earth pressure bistable oscillation characteristic value; the feature vector of the interface tunneling condition is constructed based on the cutterhead speed adjustment, the earth pressure bistable oscillation characteristic value, and the peak torque frequency on the high torque side, including:

[0165] Arrange all high-pressure steady-state periods in ascending order of their starting times to generate a high-pressure steady-state period sequence.

[0166] Specifically, the high-pressure steady-state period refers to the continuous interval in which the pressure value is continuously higher than the preset first pressure threshold and the duration exceeds the preset steady-state discrimination time window. The starting time refers to the time value corresponding to the first sampling point in this continuous interval. Ascending order means reordering according to the order from early to late according to the starting time. The generated high-pressure steady-state period sequence reflects the order in which each high-pressure steady state appears during the tunneling process.

[0167] Arrange all low-pressure steady-state periods in ascending order of their starting times to generate a low-pressure steady-state period sequence.

[0168] Specifically, the low-pressure steady-state period refers to the continuous interval in which the pressure value is continuously lower than the preset second pressure threshold and the duration exceeds the preset steady-state discrimination time window. The starting time refers to the time value corresponding to the first sampling point in this continuous interval. Ascending order means reordering according to the order from early to late according to the starting time. The generated low-pressure steady-state period sequence reflects the order in which each low-pressure steady state appears during the tunneling process.

[0169] The time interval between the start times of two adjacent high-pressure steady-state periods in the high-pressure steady-state period sequence is taken as a complete alternation cycle. The difference calculation is performed on the start times of all adjacent high-pressure steady-state periods in the high-pressure steady-state period sequence to generate an alternation cycle sequence.

[0170] Specifically, two adjacent high-pressure steady-state periods refer to two high-pressure steady-state periods that are adjacent in position in the high-pressure steady-state period sequence. The start times of the two periods are recorded as the previous time and the next time, respectively. The difference operation refers to subtracting the start time of the previous high-pressure steady-state period from the start time of the next high-pressure steady-state period. The time difference obtained is a complete alternation cycle. After performing the above difference operation on all adjacent high-pressure steady-state periods, they are arranged in order to generate an alternation cycle sequence.

[0171] For each high-pressure steady-state period in the high-pressure steady-state period sequence, the first low-pressure steady-state period that is temporally adjacent to the high-pressure steady-state period and whose start time is later than the end time of the high-pressure steady-state period is extracted. The denoised earth pressure signal between the end time of the high-pressure steady-state period and the start time of the low-pressure steady-state period is used as the transition interval signal. The transition interval signal is subjected to least squares linear fitting, and the absolute value of the slope of the fitted line is used as the transition slope to generate a transition slope sequence.

[0172] Specifically, the end time of the high-pressure steady-state period refers to the time value corresponding to the last sampling point of the high-pressure steady-state period. The first low-pressure steady-state period whose start time is later than the end time of the high-pressure steady-state period refers to the low-pressure steady-state period that is located after the high-pressure steady-state period on the time axis and has the smallest time difference between the start time and the end time. The transition interval signal refers to the denoised signal segment of the earthwork pressure from the end time of the high-pressure steady-state period to the start time of the low-pressure steady-state period. Least squares linear fitting refers to performing univariate linear regression with the time value of each sampling point in the signal segment as the independent variable and the pressure value as the dependent variable to minimize the sum of squared residuals between the fitted line and each sampling point. The absolute value of the slope of the obtained fitted line is the transition slope.

[0173] The arithmetic mean of all alternating periods in the alternating period sequence is calculated to generate the average alternating period.

[0174] The average transition slope is generated by averaging all transition slopes in the transition slope sequence.

[0175] The average alternation period and the average transition slope are multiplied to generate the earth pressure bistable oscillation characteristic value.

[0176] Specifically, the average alternation period refers to the arithmetic mean of all alternation periods in the alternation period sequence, the average transition slope refers to the arithmetic mean of all transition slopes in the transition slope sequence, and the product operation refers to multiplying the average alternation period by the average transition slope, and the resulting product is the characteristic value of the earth pressure bistable oscillation.

[0177] The cutterhead speed adjustment, the earth pressure bistable oscillation characteristic value, and the peak torque frequency of the high torque side are combined into a three-dimensional vector to generate the characteristic vector of the interface tunneling condition.

[0178] Specifically, the cutterhead speed adjustment refers to the speed adjustment value calculated using the phase-speed coupling characteristic as the optimization target under the current interface tunneling condition; the earth pressure bistable oscillation characteristic value refers to the scalar value obtained by multiplying the average alternation period and the average transition slope; the high torque side torque peak frequency refers to the frequency value corresponding to the maximum amplitude extracted from the high torque side torque spectrum; and the three-dimensional vector refers to the triple formed by arranging the above three values ​​in a fixed order. This triple serves as the characteristic vector of the interface tunneling condition and is input into the dynamic optimization model of tunneling parameters.

[0179] S5. Input the feature vector into the pre-constructed dynamic optimization model of tunneling parameters, and output the correction value of the cutterhead speed and the correction value of the soil chamber pressure setting under the current interface tunneling conditions.

[0180] In this embodiment, the step of inputting the feature vector into the pre-constructed dynamic optimization model of tunneling parameters and outputting the correction values ​​of the cutterhead rotation speed and the soil chamber pressure setpoint under the current interface tunneling conditions includes:

[0181] The dynamic optimization model for tunneling parameters is obtained. The input feature dimension of the dynamic optimization model for tunneling parameters is the same as the dimension of the feature vector of the interface tunneling working condition. The output of the dynamic optimization model for tunneling parameters is the correction value of the cutterhead speed and the correction value of the soil chamber pressure setpoint.

[0182] Specifically, the dynamic optimization model for tunneling parameters refers to a pre-built and trained fully connected neural network regression model, which is stored in the non-volatile storage medium of the tunnel boring machine's industrial control computer. The number of nodes in the input layer of this model is equal to the dimension of the interface tunneling condition feature vector, both being three nodes, corresponding to the cutterhead speed adjustment, the earth pressure bistable oscillation characteristic value, and the peak torque frequency on the high torque side, respectively. The output layer of this model has two output nodes, corresponding to the cutterhead speed correction value and the earth chamber pressure setpoint correction value, respectively. The acquisition process involves loading the network structure parameters and trained weight coefficient file of this model from the non-volatile storage medium into the memory of the industrial control computer through the data reading interface of the tunnel boring machine's industrial control computer.

[0183] Input the interface tunneling condition feature vector into the tunneling parameter dynamic optimization model to generate the cutterhead speed correction value and the soil chamber pressure set value correction value.

[0184] Specifically, the interface tunneling condition feature vector is a three-dimensional vector composed of the cutterhead speed adjustment, the earth pressure bistable oscillation characteristic value, and the peak torque frequency on the high-torque side, arranged in a fixed order. Inputting this interface tunneling condition feature vector into the dynamic optimization model of tunneling parameters involves feeding the three values ​​of this three-dimensional vector as input data into a fully connected neural network already loaded into memory. The model performs forward propagation calculations based on the pre-determined connection weights and bias parameters of each layer of neurons, passing the data layer by layer from the input layer through the hidden layers to the output layer. The cutterhead speed correction value and the earth chamber pressure setpoint correction value are the actual output values ​​of the two nodes in the output layer, which are used to update the cutterhead speed and earth chamber pressure setpoints in subsequent steps.

[0185] In this embodiment, the pre-construction process of the dynamic optimization model for tunneling parameters includes:

[0186] A tunneling performance cost function is constructed with cutterhead rotation speed and soil chamber pressure as decision variables. The tunneling performance cost function is composed of a weighted sum of tunneling speed, cutterhead torque fluctuation, and soil chamber pressure deviation after normalization. The weights are determined by a combination of the analytic hierarchy process and the entropy weight method.

[0187] Specifically, the tunneling performance cost function refers to an objective function with cutterhead rotation speed and soil chamber pressure setpoint as two input variables. This function is aggregated into a single scalar value by weighted summation after normalization of three evaluation indicators: tunneling speed, cutterhead torque fluctuation amplitude, and soil chamber pressure deviation, through maximum and minimum value normalization. The combined weighting method of the analytic hierarchy process (AHP) and entropy weighting involves first constructing a judgment matrix by pairwise comparison of the importance of the three indicators and calculating eigenvectors to obtain subjective weights; then, calculating information entropy based on the degree of variation of each indicator in historical tunneling data to obtain objective weights; finally, the subjective and objective weights are weighted and averaged proportionally to obtain the combined weight. Normalization involves subtracting the minimum value of each evaluation indicator from the original value of the candidate parameter combination and dividing by the difference between the maximum and minimum values, thus transforming each indicator into a dimensionless range of zero to one.

[0188] Obtain feature vector samples of interface tunneling conditions from historical tunneling cycles.

[0189] Specifically, the historical tunneling cycle refers to the stroke segments of each propulsion cylinder that the tunnel boring machine has completed before the current construction section, and the interface tunneling condition feature vector sample refers to the feature vector collected from each historical tunneling cycle, which consists of the cutterhead speed adjustment, the earth pressure bistable oscillation characteristic value, and the peak frequency of the high torque side torque.

[0190] For each interface tunneling condition feature vector sample, with the goal of minimizing the tunneling performance cost function, a Bayesian optimization algorithm is used to search for the optimal cutterhead speed adjustment and the optimal soil chamber pressure adjustment within the range of cutterhead speed adjustment and soil chamber pressure adjustment.

[0191] Specifically, the cutterhead speed adjustment range refers to the adjustment interval determined by the lower and upper limits of the rated speed of the main drive motor of the tunnel boring machine, and the soil chamber pressure adjustment range refers to the adjustment interval determined by the upper limit of the design bearing capacity of the soil chamber partition and the lower limit of the static soil and water pressure of the stratum.

[0192] Bayesian optimization algorithm refers to constructing a Gaussian process surrogate model in a two-dimensional search space consisting of two decision variables: cutterhead speed and soil pressure, with the tunneling performance cost function as the objective function. The expected improvement acquisition function guides the selection of candidate points in each iteration. After fifty iterations, the combination of decision variables that minimizes the cost function is output. The cutterhead speed and soil pressure values ​​in this combination are used as the optimal adjustment values. The number of iterations is taken as an empirical value of fifty based on the convergence characteristics of Bayesian optimization in the two-dimensional search space to ensure that the surrogate model fully approximates the objective function.

[0193] Using the interface tunneling working condition feature vector samples as input features, and the optimal cutterhead speed adjustment and optimal soil chamber pressure adjustment obtained by the search as output labels, a training sample set is constructed.

[0194] Construct a fully connected neural network consisting of an input layer, a hidden layer, and an output layer. The hidden layer uses a linear rectified function as the activation function, and the output layer has two output nodes.

[0195] Specifically, a fully connected neural network refers to a feedforward network structure in which all neurons between two adjacent layers are connected with weights. The input layer has three neurons corresponding to the three components of the three-dimensional feature vector. The hidden layer adopts a two-layer structure with sixty-four neurons in the first layer and thirty-two neurons in the second layer. The output layer has two neurons corresponding to the predicted values ​​of the cutterhead speed adjustment and the soil chamber pressure adjustment, respectively.

[0196] The training sample set is input into a fully connected neural network for training. The mean square error between the output values ​​of two nodes in the output layer and their corresponding labels is used as the loss function. The network weights are updated through backpropagation and iterated until the loss function converges, generating a dynamic optimization model for the tunneling parameters.

[0197] Specifically, training involves sequentially feeding the input feature vectors from the training sample set into a fully connected neural network to perform forward propagation calculations to obtain the predicted values ​​of the output layer. The mean square error between the predicted values ​​of the two nodes in the output layer and the corresponding label values ​​in the training sample set is used as the loss function. An adaptive moment estimation optimizer is used to calculate the gradient value of the loss function with respect to each connection weight layer by layer from the output layer to the input layer, and the weights and bias parameters are updated along the gradient descent direction.

[0198] The convergence criterion for the loss function is that the change in the loss function value over twenty consecutive iterations is less than a preset threshold. This preset threshold is one-thousandth of the average loss function value of the first one hundred iterations during the initial training phase. The model is considered converged when the absolute value of the difference between the loss function value of each of the twenty consecutive iterations and the loss function value of the previous iteration is less than this threshold. After training stops, the current network structure and the trained weight coefficients are saved to the non-volatile storage medium of the tunnel boring machine's industrial control computer. The saved network structure and weight coefficients together constitute the dynamic optimization model for tunneling parameters.

[0199] S6. The cutter head speed adjustment is updated with the cutter head speed correction value to obtain the optimized cutter head speed, and the soil chamber pressure setting value is updated with the soil chamber pressure setting value correction value to obtain the optimized soil chamber pressure setting value.

[0200] In this embodiment, the step of updating the cutter head speed adjustment amount with the cutter head speed correction value to obtain the optimized cutter head speed, and updating the soil chamber pressure setting value with the soil chamber pressure setting value correction value to obtain the optimized soil chamber pressure setting value, includes:

[0201] Obtain the current sample value of the cutter head rotation speed timing signal.

[0202] Specifically, the cutter head speed timing signal refers to the sequence of speed values ​​collected and stored by the encoder of the cutter head drive motor at equal sampling times. The current sampling value refers to the speed value corresponding to the latest sampling time in the timing signal.

[0203] The current sampled value of the cutter head speed timing signal is added to the cutter head speed adjustment amount to generate the reference cutter head speed.

[0204] Specifically, the cutter head speed adjustment amount refers to the speed adjustment value calculated with the phase-speed coupling characteristic as the optimization target. The current sampled value is added to this adjustment amount, and the result is used as the reference cutter head speed.

[0205] The base cutter head speed and the cutter head speed correction value are added together to generate the optimized cutter head speed.

[0206] Specifically, the cutterhead speed correction value refers to the speed adjustment amount output by the dynamic optimization model of tunneling parameters. The baseline cutterhead speed is added to this correction value, and the result is used as the optimized cutterhead speed.

[0207] Obtain the current sample value of the earth chamber pressure timing signal.

[0208] Specifically, the earth pressure time series signal refers to the sequence of pressure values ​​output at equal sampling times after median filtering. The current sampling value refers to the pressure value corresponding to the latest sampling time in the time series signal.

[0209] The current sampled value of the earth chamber pressure timing signal is added to the correction value of the earth chamber pressure setpoint to generate an optimized earth chamber pressure setpoint.

[0210] Specifically, the soil chamber pressure setpoint correction value refers to the pressure setpoint adjustment amount output by the dynamic optimization model of tunneling parameters. The current sampled value is added to this correction value, and the result is used as the optimized soil chamber pressure setpoint.

[0211] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0212] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0213] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A karst stratum shield tunneling parameter dynamic optimization method based on industrial big data, characterized in that, Includes the following steps: S1. Collect industrial data during the tunneling process of the tunnel boring machine. The industrial data includes the torque timing signal of the left zone of the cutterhead, the torque timing signal of the right zone of the cutterhead, the soil chamber pressure timing signal, and the cutterhead speed timing signal. S2. Perform time-domain and frequency-domain analysis on the torque timing signals of the left and right regions of the cutterhead, extract the peak torque frequency on the high-torque side, and calculate the instantaneous phase difference sequence. Calculate the phase-speed coupling characteristic quantity based on the instantaneous phase difference sequence and the cutterhead speed timing signal. Use the phase-speed coupling characteristic quantity as the optimization target to reverse match and avoid the cutterhead structure resonance by adjusting the cutterhead speed. S3. Perform time-domain segmentation on the earth chamber pressure time sequence signal to obtain pressure segment time periods. Extract the time periods from each pressure segment time period where the pressure value is continuously higher than the preset first pressure threshold within the preset steady state discrimination time window as the high pressure steady state time period and the low pressure steady state time period where the pressure value is continuously lower than the preset second pressure threshold. S4. Calculate the alternation period and transition slope between the high-pressure steady-state period and the low-pressure steady-state period, and multiply the alternation period and transition slope to generate the earth pressure bistable oscillation characteristic value. The feature vector of the interface tunneling condition is composed of the cutterhead speed adjustment, the earth pressure bistable oscillation characteristic value, and the peak torque frequency of the high torque side. S5. Input the feature vector into the pre-constructed dynamic optimization model of tunneling parameters, and output the correction value of the cutterhead speed and the correction value of the soil chamber pressure setting under the current interface tunneling conditions. S6. The cutter head speed adjustment is updated with the cutter head speed correction value to obtain the optimized cutter head speed, and the soil chamber pressure setting value is updated with the soil chamber pressure setting value correction value to obtain the optimized soil chamber pressure setting value.

2. The karst stratum shield tunneling parameter dynamic optimization method based on industrial big data according to claim 1, characterized in that, The industrial data collected during the tunnel boring machine's excavation process includes torque timing signals for the left and right sections of the cutterhead, soil chamber pressure timing signals, and cutterhead rotation speed timing signals. The timing signal of the cutterhead rotation speed is acquired by the encoder of the cutterhead drive motor of the tunnel boring machine; The original torque voltage signal of the left zone during the preset tunneling cycle is collected by the torque sensor in the left zone of the tunnel boring machine cutterhead. The original torque voltage signal of the right zone during the preset tunneling cycle is collected by the torque sensor in the right zone of the tunnel boring machine cutterhead. Obtain the pre-calibrated calibration coefficients of the left-zone torque sensor and the right-zone torque sensor; The original torque voltage signal of the left zone is linearly converted by the calibration coefficient of the left zone torque sensor to generate the initial torque timing signal of the left zone. The original torque voltage signal of the right zone is linearly converted by the calibration coefficient of the right zone torque sensor to generate the initial torque timing signal of the right zone. Digital low-pass filtering is performed on the initial torque timing signal of the left zone and the initial torque timing signal of the right zone to obtain the torque timing signal of the left zone and the torque timing signal of the right zone of the cutterhead. The original analog current signal of the soil chamber pressure is acquired by the pressure transmitter on the upper part of the soil chamber partition. The original analog current signal of the earth chamber pressure is converted from analog to digital to generate the original digital sequence of the earth chamber pressure. The original digital sequence of earth pressure is subjected to median filtering to generate the earth pressure time sequence signal.

3. The karst stratum shield tunneling parameter dynamic optimization method based on industrial big data according to claim 1, characterized in that, The time-domain and frequency-domain analysis of the torque timing signals of the left and right regions of the cutterhead is performed to extract the peak torque frequency on the high-torque side and calculate the instantaneous phase difference sequence, including: A moving average operation is performed on the torque timing signals of the left and right regions of the cutter head to generate smoothed torque sequences for the left and right regions. The arithmetic mean of the left-zone torque smoothing sequence and the right-zone torque smoothing sequence is calculated to generate the mean torque of the left zone and the mean torque of the right zone. Compare the average torque values ​​of the left and right zones, and mark the side with the larger value as the high torque side and the side with the smaller value as the low torque side. Fast Fourier transform is performed on the torque time-series signal corresponding to the high torque side and the torque time-series signal corresponding to the low torque side to generate the torque spectrum of the high torque side and the torque spectrum of the low torque side. Extract the frequency corresponding to the maximum amplitude from the torque spectrum on the high torque side to obtain the peak torque frequency on the high torque side; Perform Hilbert transform on the torque timing signals corresponding to the high torque side and the low torque side to generate analytical signals for the high torque side and the low torque side. Instantaneous phases are extracted from the high-torque side analytical signal and the low-torque side analytical signal to generate high-torque side instantaneous phase sequences and low-torque side instantaneous phase sequences; The instantaneous phase difference sequence is generated by performing a difference operation between the instantaneous phase sequence on the high torque side and the instantaneous phase sequence on the low torque side.

4. The karst stratum shield tunneling parameter dynamic optimization method based on industrial big data according to claim 1, characterized in that, The step of calculating the phase-speed coupling characteristic quantity based on the instantaneous phase difference sequence and the cutter head speed timing signal, and using the phase-speed coupling characteristic quantity as the optimization target to reverse match and avoid cutter head structural resonance, includes: Perform a difference operation on the instantaneous phase difference sequence to generate an instantaneous phase difference rate of change sequence; Perform absolute value calculation on the instantaneous phase difference change rate sequence to generate an instantaneous phase difference change rate amplitude sequence; The unit conversion of the cutter head speed timing signal is performed to generate the cutter head speed frequency signal. The division operation between the high torque side torque peak frequency and the cutter head speed frequency signal is performed to generate the torque-speed frequency ratio sequence. The arithmetic mean of the corresponding point product operation is performed on the instantaneous phase difference change rate amplitude sequence and the torque-speed frequency ratio sequence to generate the phase-speed coupling characteristic quantity; Obtain the set of natural frequencies from the finite element modal analysis results of the tunnel boring machine cutterhead structure; The current sampled value of the cutter head speed timing signal is used as the initial speed. Within the allowable adjustment range of the cutter head speed, a candidate speed sequence is generated with a preset speed step size. Candidate speed values ​​in the candidate speed sequence whose absolute value of the difference between the converted candidate speed frequency value and the natural frequency set is less than a preset neighborhood radius are removed to generate a feasible speed sequence. For each feasible speed value in the feasible speed sequence, calculate the torque-speed frequency ratio corresponding to the feasible speed value, and calculate the arithmetic mean after performing corresponding point product operation on the torque-speed frequency ratio and the instantaneous phase difference change rate amplitude sequence, which is used as the estimated value of the phase-speed coupling characteristic quantity. The feasible speed value with the smallest estimated value of the phase-speed coupling characteristic is selected as the target speed value. The difference between the target speed and the initial speed is calculated to generate the cutter head speed adjustment amount.

5. The karst stratum shield tunneling parameter dynamic optimization method based on industrial big data according to claim 1, characterized in that, The time-domain segmentation of the earth chamber pressure time-series signal is performed to obtain pressure segment time periods. From each pressure segment time period, the time period in which the pressure value is continuously higher than a preset first pressure threshold within a preset steady-state discrimination time window is extracted as the high-pressure steady-state time period, and the low-pressure steady-state time period in which the pressure value is continuously lower than a preset second pressure threshold is extracted as the low-pressure steady-state time period, including: The soil pressure timing signal is low-pass filtered to generate a soil pressure filtered signal. Median filtering is applied to the soil pressure filter signal to generate a denoised soil pressure signal. A first-order difference operation is performed on the denoised earth pressure signal to generate an earth pressure difference sequence. Zero-crossing detection is performed on the differential pressure sequence of the earth chamber, and the indexes of sampling points in the differential pressure sequence of the earth chamber where the sign changes are extracted to generate a set of zero-crossing indexes; Using the index positions in the zero-crossing index set as the dividing boundaries, the soil pressure denoising signal is segmented to generate pressure segment time periods; Remove pressure segments whose duration is less than the preset steady-state discrimination time window from the pressure segment time periods, and generate effective pressure segment time periods; For each effective pressure segment, extract the pressure values ​​of all sampling points within the effective pressure segment, and mark the continuous intervals where the pressure value is continuously higher than the preset first pressure threshold and the duration exceeds the preset steady-state discrimination time window as high-pressure steady-state periods; For each effective pressure segment, extract the pressure values ​​of all sampling points within the effective pressure segment, and mark the continuous interval where the pressure value is continuously lower than the preset second pressure threshold and the duration exceeds the preset steady-state discrimination time window as the low-pressure steady-state period.

6. The karst stratum shield tunneling parameter dynamic optimization method based on industrial big data according to claim 5, characterized in that, The alternation period and transition slope of the high-pressure steady-state period and the low-pressure steady-state period are calculated, and the alternation period and transition slope are multiplied to generate the earth pressure bistable oscillation characteristic value. The feature vector of the interface tunneling condition is constructed based on the cutterhead speed adjustment, the earth pressure bistable oscillation characteristic value, and the peak torque frequency on the high torque side, including: Arrange all high-pressure steady-state periods in ascending order of their start times to generate a high-pressure steady-state period sequence. Arrange all low-pressure steady-state periods in ascending order of their start times to generate a low-pressure steady-state period sequence. The time interval between the start times of two adjacent high-pressure steady-state periods in the high-pressure steady-state period sequence is taken as a complete alternation cycle. The difference calculation is performed on the start times of all adjacent high-pressure steady-state periods in the high-pressure steady-state period sequence to generate an alternation cycle sequence. For each high-pressure steady-state period in the high-pressure steady-state period sequence, the first low-pressure steady-state period that is temporally adjacent to the high-pressure steady-state period and whose start time is later than the end time of the high-pressure steady-state period is extracted. The denoised earth pressure signal between the end time of the high-pressure steady-state period and the start time of the low-pressure steady-state period is used as the transition interval signal. The transition interval signal is subjected to least squares linear fitting, and the absolute value of the slope of the fitted line is used as the transition slope to generate a transition slope sequence. The arithmetic mean of all alternating periods in the alternating period sequence is calculated to generate the average alternating period. The average transition slope is generated by performing an arithmetic mean operation on all transition slopes in the transition slope sequence. The average alternation period and the average transition slope are multiplied to generate the earth pressure bistable oscillation characteristic value; The cutterhead speed adjustment, the earth pressure bistable oscillation characteristic value, and the peak torque frequency of the high torque side are combined into a three-dimensional vector to generate the characteristic vector of the interface tunneling condition.

7. The method for dynamic optimization of tunneling parameters of shield tunneling machines in karst strata based on industrial big data as described in claim 6, characterized in that, The process of inputting feature vectors into a pre-constructed dynamic optimization model for tunneling parameters and outputting correction values ​​for cutterhead rotation speed and soil chamber pressure setpoint under the current interface tunneling conditions includes: The dynamic optimization model for tunneling parameters is obtained. The input feature dimension of the dynamic optimization model for tunneling parameters is the same as the dimension of the feature vector of the interface tunneling working condition. The output of the dynamic optimization model for tunneling parameters is the correction value of the cutterhead speed and the correction value of the soil chamber pressure setpoint. Input the interface tunneling condition feature vector into the tunneling parameter dynamic optimization model to generate the cutterhead speed correction value and the soil chamber pressure set value correction value.

8. The method for dynamic optimization of tunneling parameters of shield tunneling machines in karst strata based on industrial big data as described in claim 7, characterized in that, The pre-construction process of the dynamic optimization model for tunneling parameters includes: A tunneling performance cost function is constructed with cutterhead speed and soil chamber pressure as decision variables. The tunneling performance cost function is composed of a weighted sum of tunneling speed, cutterhead torque fluctuation, and soil chamber pressure deviation after normalization. The weights are determined by a combination of the analytic hierarchy process and the entropy weight method. Obtain feature vector samples of interface tunneling conditions from historical tunneling cycles; For each interface tunneling condition feature vector sample, with the goal of minimizing the tunneling performance cost function, the Bayesian optimization algorithm is used to search for the optimal cutterhead speed adjustment and the optimal soil chamber pressure adjustment within the range of cutterhead speed adjustment and soil chamber pressure adjustment. Using the interface tunneling working condition feature vector samples as input features, and the optimal cutterhead speed adjustment and optimal soil chamber pressure adjustment obtained by the search as output labels, a training sample set is constructed. Construct a fully connected neural network consisting of an input layer, a hidden layer, and an output layer. The hidden layer uses a linear rectified function as the activation function, and the output layer has two output nodes. The training sample set is input into a fully connected neural network for training. The mean square error between the output values ​​of two nodes in the output layer and their corresponding labels is used as the loss function. The network weights are updated through backpropagation and iterated until the loss function converges, generating a dynamic optimization model for the tunneling parameters.

9. The method for dynamic optimization of tunneling parameters of shield tunneling machines in karst strata based on industrial big data as described in claim 8, characterized in that, The process of updating the cutter head speed adjustment amount with the cutter head speed correction value to obtain the optimized cutter head speed, and updating the soil chamber pressure setting value with the soil chamber pressure setting value correction value to obtain the optimized soil chamber pressure setting value, includes: Obtain the current sample value of the tool turret speed timing signal; The current sampled value of the cutter head speed timing signal is added to the cutter head speed adjustment amount to generate the reference cutter head speed; The base cutter head speed and the cutter head speed correction value are added together to generate the optimized cutter head speed. Obtain the current sample value of the earth chamber pressure timing signal; The current sampled value of the earth chamber pressure timing signal is added to the correction value of the earth chamber pressure setpoint to generate an optimized earth chamber pressure setpoint.

Citation Information

Patent Citations

  • Karst stratum shield tunneling parameter prediction method and system based on federated learning

    CN119917808A