A method for real-time regulation of rheological properties of muck in a tunnel boring machine

Abstract

This invention relates to the field of intelligent adaptive control of tunnel boring equipment, and more particularly to a real-time control method for the rheological properties of tunnel boring machine (TBM) excavation materials. The method includes the following steps: S1: Arranging the TBM operating conditions sequentially according to the time sequence of occurrence based on changes in geological conditions and equipment conditions, forming a condition sequence and determining associated condition combinations; S2: Collecting excavation material moisture content, composition, viscosity, and rheological parameters as dimensions of the state space, and determining a comprehensive stochastic index based on the state space dimension; S3: Determining the comprehensive effectiveness of the current associated condition combination for subsequent associated condition combinations, and determining the joint rate of change, dominant direction of change, and long-term evolution trend of lag and stochasticity based on the comprehensive effectiveness and the comprehensive stochastic index. This invention, by quantifying lag and stochasticity, identifies four evolutionary modes, achieving rapid response and adaptive optimization of excavation material rheological control.

Description

Technical Field

[0001] This invention relates to the field of intelligent adaptive control of tunnel boring equipment, and in particular to a method for real-time control of the rheological properties of tunnel boring machine excavation materials. Background Technology

[0002] As the core equipment for tunnel excavation, the tunnel boring machine (TBM) cuts through the ground with its cutterhead and discharges the excavated soil using a screw conveyor, achieving mechanized excavation of underground spaces. The rheological properties of excavated soil refer to its characteristics of flow and deformation under stress, directly determining the flow state of the excavated soil within the soil chamber, the ability to maintain the soil plug effect, and the stability of the soil discharge process. Existing real-time control methods rely on sensors to collect tunneling parameters and dynamically adjust the injection rates of foam and bentonite additives by establishing empirical models or rule bases. This achieves stable control of the excavated soil state under single-stratum conditions, effectively reducing the risks of abnormal cutterhead torque and surface subsidence, and accumulating important experience for intelligent TBM construction.

[0003] However, existing methods reveal significant shortcomings under complex and variable geological conditions or frequent equipment status adjustments. Traditional control strategies often rely on the acquisition of state parameters within fixed time windows and threshold-based feedback adjustment. When tunneling conditions change continuously, the response of the excavated soil rheology exhibits a strong historical dependence, meaning that the state of the preceding tunneling process directly affects the subsequent control effect. Due to the time lag between real-time data acquisition and rheological state inversion, the control system struggles to capture the rheological evolution trend in a timely manner, often resulting in excessive or insufficient additive injection. Furthermore, the impact of the same type of condition change on rheology varies significantly under different initial conditions. Existing methods treat condition changes and control effects as a fixed correspondence, lacking the ability to quantitatively assess the uncertainty of fluctuations, making it difficult to guarantee control stability in scenarios with multi-source interference coupling.

[0004] Therefore, there is an urgent need to study a real-time control method for the rheological properties of tunnel boring machine excavated soil that can dynamically adapt to continuous changes in working conditions and state-dependent characteristics. By integrating multi-source monitoring data and dynamic modeling technology, the response speed and adaptive capability of rheological control in complex tunneling environments can be improved. Summary of the Invention

[0005] To overcome the drawbacks of dynamic response lag and state-dependent instability, this invention provides a method for real-time control of the rheological properties of tunnel boring machine excavated soil.

[0006] The technical implementation scheme of the present invention is: a method for real-time control of the rheological properties of tunnel boring machine excavated soil, comprising the following steps: S1: Arrange the tunnel boring machine's operating conditions in chronological order according to changes in geological conditions and equipment conditions to form a sequence of operating conditions and determine the associated combinations of operating conditions. S2: Collect the moisture content, composition, viscosity and rheological parameters of the slag and use them as the state space dimension, and determine the comprehensive stochasticity index based on the state space dimension; S3: Determine the comprehensive effectiveness of the current associated working condition combination for subsequent associated working condition combinations, and determine the joint rate of change of lag degree and randomness, the dominant direction of change and the long-term evolution trend based on the comprehensive effectiveness and the comprehensive randomness index; S4: Obtain the current evolutionary model based on the joint rate of change, the dominant direction of change, and the long-term evolutionary trend, and determine the real-time control method based on the current evolutionary model.

[0007] Preferably, the step of arranging the tunnel boring machine's operating conditions in chronological order according to changes in geological conditions and equipment conditions to form a sequence of operating conditions and determining associated combinations of operating conditions includes: For any two consecutive working conditions in the working condition sequence, the relationship between the preceding and subsequent working conditions is divided into enhancing influence relationship, offsetting influence relationship and no significant influence relationship according to the superposition effect of the influence of the preceding and subsequent working conditions on the rheology of the slag soil. Combinations of operating conditions that have reinforcing or offsetting effects are defined as associated operating condition combinations.

[0008] Preferably, the step of classifying the relationship between preceding and subsequent operating conditions into reinforcing influence relationships, offsetting influence relationships, and no significant influence relationships includes: Enhanced influence relationship refers to the unidirectional shift in the rheological properties of slag and soil caused by preceding and subsequent working conditions. The offsetting effect relationship refers to the fact that the preceding and subsequent working conditions cause the rheological properties of the slag and soil to shift in opposite directions. No significant influence relationship means that the superposition effect of the preceding and subsequent working conditions does not cause a identifiable directional shift in the rheological properties of the slag.

[0009] Preferably, the step of collecting the moisture content, composition, viscosity, and rheological parameters of the slag and using them as dimensions of the state space, and determining the comprehensive stochastic index based on the state space dimensions, includes: Accuracy is obtained by calculating the sum of squares and the square root of the squares for each dimension of the state space. The real-time detection data of each state space dimension at the current time point are used as the initial state vector of the construction waste. If the real-time detection data is lower than the highest accuracy value among the historical observation data, the historical observation data is selected to form the initial state vector of the slag and soil. The difference between the initial state vector of the slag before the occurrence of the preceding working condition and the state vector of the slag after the occurrence of the preceding working condition is taken as the change in the rheological properties of the slag under the preceding working condition. The difference between the initial state vector of the slag before the occurrence of the subsequent working condition and the state vector of the slag after the occurrence of the subsequent working condition is taken as the change in the rheological properties of the slag under the subsequent working condition. The superposition value of the rheological changes of the slag and soil under the preceding and subsequent working conditions is taken as the rheological change of the associated working condition combination. For the same type of associated working condition combination, under different initial state vectors of slag soil, the rheological change of the associated working condition combination is repeatedly obtained to form a rheological change sample set; The variance, Jacobian matrix norm, and confidence interval width of the rheological change sample set are calculated, and then normalized and weighted to obtain the comprehensive randomness index.

[0010] Preferably, determining the overall effectiveness of subsequent associated operating condition combinations based on the current associated operating condition combination includes: The time required from the initial collection of soil moisture content, soil composition, soil viscosity and rheological parameters to the completion of accuracy index calculation and the formation of the initial state vector of the soil is taken as the first time window; The time required from the occurrence of the preceding condition in the associated working condition combination to the completion of the calculation of the rheological change of the associated working condition combination is used as the second time window. If the first time window is less than or equal to the second time window, the matching status of the associated working condition combination is determined to be a successful match; If the first time window is larger than the second time window, the matching status of the associated working condition combination is determined to be a matching failure. The calculation results of the rheological changes corresponding to the failed matching of associated load case combinations are used to predict the accuracy of the k-th associated load case combination. And calculate the overall effectiveness of this associated working condition combination for the subsequent N associated working condition combinations. .

[0011] Preferably, the accuracy of the prediction of the rheological change corresponding to the failed associated operating condition combination for the subsequent k-th associated operating condition combination includes: The formula for prediction accuracy is: ,in, This is the current calculation result. This represents the actual rheological change in subsequent combinations of related operating conditions. This represents the range of values ​​for the rheological change.

[0012] Preferably, calculating the overall effectiveness of the associated operating condition combination for the subsequent N associated operating condition combinations includes: Selecting the subsequent N related working condition combinations as the contribution range, the comprehensive effectiveness formula is as follows: , ,in, For distance decay weights, The attenuation coefficient; When the overall effectiveness is higher than the set threshold, the associated working condition combination that failed to match is determined to be effective and the calculation lag does not constitute a control obstacle. When the overall effectiveness is below the set threshold, the associated working condition combination that fails to match is deemed to be ineffective and the calculation lag constitutes a control obstacle. The overall effectiveness is used as the degree of lag; the higher the overall effectiveness, the smaller the impact of the lag on regulation.

[0013] Preferably, the determination of the joint rate of change, dominant direction of change, and long-term evolution trend of lag degree and randomness based on comprehensive effectiveness and comprehensive stochasticity index includes: Each associated working condition combination is represented as a two-dimensional feature point. The horizontal axis of the two-dimensional feature point is the comprehensive effectiveness, and the vertical axis of the two-dimensional feature point is the comprehensive randomness index. The two-dimensional feature points of N consecutive associated working condition combinations are connected in the order of occurrence to form a feature trajectory. Calculate the Euclidean distance between adjacent two-dimensional feature points in the feature trajectory as the combined rate of change of hysteresis and randomness; The angle of change of the direction of the line connecting adjacent two-dimensional feature points in the feature trajectory is calculated as the dominant direction of the change in lag degree and randomness; The cumulative offset vector from the first two-dimensional feature point to the last two-dimensional feature point in the feature trajectory is calculated as a long-term evolution trend of lag degree and randomness.

[0014] Preferably, obtaining the current evolutionary model based on the joint rate of change, the dominant direction of change, and the long-term evolutionary trend includes: Collect sample data of the dominant change direction in all historical characteristic trajectories, and plot the probability distribution histogram of the dominant change direction; When the probability distribution histogram has a bimodal feature, the angle values ​​of the directional changes corresponding to the two peaks are taken, and the valley angle value between the two peaks is used as the classification boundary to distinguish the dominant direction of change. When the probability distribution histogram does not have a bimodal feature, the median of all dominant directions of change is taken as the classification boundary; When the absolute value of the M consecutive dominant directions of change in the feature trajectory is greater than the classification boundary and the randomness component of the long-term evolution trend is greater than the lag component, the current evolution pattern is identified as a random accumulation pattern. When the absolute value of the M consecutive dominant directions of change in the feature trajectory is less than the classification boundary and the lag component of the long-term evolution trend is greater than the random component, the current evolution pattern is identified as a lag accumulation pattern. When the joint rate of change in the characteristic trajectory is consistently greater than the average historical joint rate of change, the current evolution mode is identified as an unstable oscillation mode. When the joint rate of change in the feature trajectory is consistently less than the historical average joint rate of change and the two-dimensional feature points are clustered in a small area, the current evolutionary pattern is identified as a stable convergence pattern.

[0015] Preferably, the method for determining real-time regulation based on the current evolutionary model includes: When the evolution mode is a stochastic accumulation mode, increase the sampling frequency of the state space dimension to shorten the first time window; When the evolution mode is a lag-cumulative mode, the evolution trend of the characteristic trajectory of the preceding working condition is used as a benchmark to predict the direction of rheological change after the occurrence of the subsequent working condition and adjust the additive injection parameters in advance. When the evolution mode is an unstable oscillation mode, the active control action based on the combination of related operating conditions is suspended, the basic additive injection rate is maintained, and active control is resumed after the characteristic trajectory reconverges. When the evolution mode is a stable convergence mode, the additive injection parameters are adjusted according to the real-time detected rheological changes. During the tunneling process, new associated working condition combination data are continuously collected, new two-dimensional feature points are added to the feature trajectory, and the sliding time window keeps the feature trajectory length fixed. The above steps are repeated to achieve continuous adaptive updates of the control strategy.

[0016] Beneficial effects: This invention quantifies the superposition effect of continuous working conditions by associating working conditions, collects the moisture content, composition, viscosity and rheological parameters of the slag soil to construct the state space dimension, and calculates the comprehensive randomness index to solve the problem of the difference in the influence amplitude of the same working condition under different initial conditions; it quantifies the degree of lag by matching time windows, maps lag and randomness to feature trajectories and extracts geometric parameters, identifies four evolution modes: randomness accumulation, lag accumulation, unstable oscillation and stable convergence, and executes adaptive strategies such as increasing sampling frequency, feedforward compensation, pausing active control and conventional feedback adjustment respectively; this invention decomposes dynamic response lag and state-dependent instability into sub-scenarios that can be dealt with in a targeted manner, which significantly improves the response speed and adaptive capability of slag soil rheological control under complex working conditions. Attached Figure Description

[0017] Figure 1 This is a flowchart of a method for real-time control of the rheological properties of tunnel boring machine excavated soil according to the present invention; Figure 2 This is a flowchart illustrating the process of obtaining the comprehensive randomness index in this invention. Figure 3 This is a flowchart of the method for determining the evolution mode of the present invention. Detailed Implementation

[0018] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0019] A method for real-time control of the rheological properties of tunnel boring machine excavated soil, such as Figure 1 , Figure 2 and Figure 3 As shown, it includes the following steps: S1: Based on changes in geological conditions and equipment status, the operating conditions of the tunnel boring machine are arranged sequentially according to their occurrence time to form a sequence of operating conditions and determine the associated combinations of operating conditions, including: For any two consecutive working conditions in the working condition sequence, the relationship between the preceding and subsequent working conditions is divided into enhancing influence relationship, offsetting influence relationship and no significant influence relationship according to the superposition effect of the influence of the preceding and subsequent working conditions on the rheology of the slag soil. Combinations of operating conditions that have a reinforcing or offsetting effect relationship are defined as associated operating condition combinations; Enhanced influence relationship refers to the unidirectional shift in the rheological properties of slag and soil caused by preceding and subsequent working conditions. The offsetting effect relationship refers to the fact that the preceding and subsequent working conditions cause the rheological properties of the slag and soil to shift in opposite directions. No significant influence relationship means that the superposition effect of the preceding and subsequent working conditions does not cause a identifiable directional shift in the rheological properties of the slag.

[0020] It should be noted that during tunnel boring machine (TBM) excavation, changes in geological conditions usually imply abrupt or gradual changes in stratum properties, while changes in equipment status reflect adjustments to operating parameters or the occurrence of abnormal situations. To accurately capture the historical impact of preceding conditions on subsequent responses, the control system needs to perform a time-series analysis of these two types of conditions based on their sequence of occurrence. The rheology of excavated soil exhibits a significant state memory effect; the shearing action applied by preceding conditions alters the initial state of the excavated soil, thus preventing the independent superposition of the effects of subsequent conditions. Therefore, any two consecutive conditions are used as the smallest unit of analysis. The superposition effect is essentially a vector synthesis of the rheological changes under the action of two consecutive conditions. Based on the consistency of the superposition direction, the relationships between conditions are divided into three categories: reinforcing influence, offsetting influence, and no significant influence. Reinforcing influence relationships are used to identify scenarios requiring early intervention, offsetting influence relationships are used to discover combinations that can mutually restrain each other and maintain stability, and no significant influence relationships are used to filter redundant disturbances.

[0021] S2: Collect the moisture content, composition, viscosity, and rheological parameters of the construction waste and use them as the state-space dimension. Based on the state-space dimension, determine the comprehensive stochastic index, including: Accuracy is obtained by calculating the sum of squares and the square root of the squares for each dimension of the state space. The real-time detection data of each state space dimension at the current time point are used as the initial state vector of the construction waste. If the real-time detection data is lower than the highest accuracy value among the historical observation data, the historical observation data is selected to form the initial state vector of the slag and soil. The difference between the initial state vector of the slag before the occurrence of the preceding working condition and the state vector of the slag after the occurrence of the preceding working condition is taken as the change in the rheological properties of the slag under the preceding working condition. The difference between the initial state vector of the slag before the occurrence of the subsequent working condition and the state vector of the slag after the occurrence of the subsequent working condition is taken as the change in the rheological properties of the slag under the subsequent working condition. The superposition value of the rheological changes of the slag and soil under the preceding and subsequent working conditions is taken as the rheological change of the associated working condition combination. For the same type of associated working condition combination, under different initial state vectors of slag soil, the rheological change of the associated working condition combination is repeatedly obtained to form a rheological change sample set; The variance, Jacobian matrix norm, and confidence interval width of the rheological change sample set are calculated, and then normalized and weighted to obtain the comprehensive randomness index.

[0022] It should be noted that the reference Figure 2 A capacitive humidity sensor outputs moisture content by measuring changes in the dielectric constant of the slag. A near-infrared spectrometer identifies mineral composition and particle size distribution by analyzing spectral absorption characteristics. The driving torque of the soil mixing rod is inverted based on the Metzner-Otto correlation to determine the apparent viscosity. The driving torque of the screw conveyor is obtained by fitting shear stress and shear rate to obtain yield stress and plastic viscosity. These parameters are used as dimensions of the state space to establish a mathematical description space for the rheology of the slag, providing clear physical coordinates for subsequent analysis. Accuracy is calculated using the square root of the sum of squares of the inversion errors of each dimension to screen reliable data sources. The inversion errors of each dimension have been normalized by the range of that dimension. This refers to the deviation between the measured value and the reference value of the sensor in this dimension. The reference value is the laboratory calibration value or the synchronous measurement value of the high-precision sensor. The initial state vector of the slag and soil records the state values ​​before the occurrence of the previous working conditions. When the accuracy of the real-time detection data is lower than the historical peak, the vector is constructed by using historical high-accuracy data first, so as to avoid the deviation introduced by the instantaneous disturbance of the sensor. The range of historical data is limited by a sliding time window method. The most recently collected L sets of working condition data before the current moment are selected as the historical dataset. The value of L is set according to the stability of the working condition. For example, L=100 is used in stable strata, and L=50 is used when the strata change frequently; or historical tunneling data of the same type of strata can be used as a supplementary historical data source. The rheological changes of the soil under preceding working conditions characterize the independent influence of a single working condition on the initial state, while the rheological changes under subsequent working conditions reflect the incremental changes after superposition. The difference between the two changes is used to isolate the true contribution of the subsequent working condition. The superimposed value is a vector synthesis of the two changes, representing the net rheological change under the combined action of continuous working conditions. For the same type of associated working condition combination, superimposed values ​​are repeatedly collected under different initial state vector conditions to form a sample set, and the variance of the sample set is calculated to measure the degree of dispersion. The Jacobian matrix norm is used to measure the sensitivity of the rheological changes to the initial state vector. Its calculation method is as follows: taking the rheological changes as the output and the initial state vector as the input, a locally weighted linear regression is used to fit the mapping relationship, and the Jacobian matrix is ​​obtained after calculation. Calculate the Frobenius norm; if the sample size is insufficient for fitting, use the median norm of the combination of working conditions in historical data of the same type of strata as a substitute; estimate the fitting probability distribution function through kernel density, and use the percentile method to determine the width of the confidence interval at a given confidence level to measure the range of uncertainty. After normalizing the above three indicators, assign weight coefficients to them and perform a weighted summation. The weight coefficients are determined by expert experience or analytic hierarchy process based on the contribution of each indicator to the comprehensive stochastic index. The sum of the weight coefficients is 1. The weight coefficients are dynamically updated according to the contribution ratio of each indicator to the total variance of the comprehensive stochastic index. The update cycle is set according to the frequency of working condition changes to ensure that the weights reflect the dominant sources of uncertainty under the current working conditions, thus obtaining the comprehensive stochastic index. This index quantifies the degree of certainty in the mapping relationship between operating condition combinations and rheological response, solving the deficiency in the background technology where the influence of the same operating condition under different initial conditions varies significantly but lacks quantitative evaluation methods; by transforming state-dependent uncertainty into a calculable numerical index, it provides a quantitative basis for distinguishing between stochastic cumulative modes and hysteretic cumulative modes in subsequent steps.

[0023] S3-1: Determine the overall effectiveness of subsequent associated operating condition combinations based on the current associated operating condition combinations, including: The time required from the initial collection of soil moisture content, soil composition, soil viscosity and rheological parameters to the completion of accuracy index calculation and the formation of the initial state vector of the soil is taken as the first time window; The time required from the occurrence of the preceding condition in the associated working condition combination to the completion of the calculation of the rheological change of the associated working condition combination is used as the second time window. If the first time window is less than or equal to the second time window, the matching status of the associated working condition combination is determined to be a successful match; If the first time window is larger than the second time window, the matching status of the associated working condition combination is determined to be a matching failure. The calculation results of the rheological changes corresponding to the failed matching of associated load case combinations are used to predict the accuracy of the k-th associated load case combination. And calculate the overall effectiveness of this associated working condition combination for the subsequent N associated working condition combinations. ; The formula for prediction accuracy is: ,in, This is the current calculation result. This represents the actual rheological change in subsequent combinations of related operating conditions. The range of values ​​for the rheological change; It should be noted that the range of values ​​for rheological changes is... The extreme value range of historical monitoring data on the change in rheological properties of excavated soil under the same type of tunneling stratum for the tunnel boring machine is specifically set as follows: ,in, This represents the historical maximum value of the change in soil rheology under this type of stratum and this combination of related working conditions. This represents the historical minimum value of the change in rheological properties of the slag under this type of stratum and this combination of related working conditions; if it is a new stratum / no historical data, the extreme value range of the measured change in rheological properties of the slag under simulated working conditions in the same stratum in the laboratory is taken. Selecting the subsequent N related working condition combinations as the contribution range, the comprehensive effectiveness formula is as follows: , ,in, For distance decay weights, The attenuation coefficient; When the overall effectiveness is higher than the set threshold, the associated working condition combination that failed to match is determined to be effective and the calculation lag does not constitute a control obstacle. When the overall effectiveness is below the set threshold, the associated working condition combination that fails to match is deemed to be ineffective and the calculation lag constitutes a control obstacle. The overall effectiveness is used as the degree of lag; the higher the overall effectiveness, the smaller the impact of the lag on regulation.

[0024] It should be noted that there is a time difference between data acquisition and inversion calculation. The first time window corresponds to the complete cycle required from sensor sampling to the formation of a reliable initial state vector, while the second time window records the time elapsed from the occurrence of the preceding operating condition to the completion of the calculation of the rheological change in that operating condition. By comparing the lengths of the two windows, a successful match indicates that the inversion result is ready before the subsequent operating condition occurs, while a failed match indicates that the calculation result lags behind the actual operating condition process. For cases of failed matching, it is necessary to further determine the predictive value of the lagging calculation result. The prediction accuracy is quantified by normalizing the deviation between the current calculation result and the actual subsequent result, thus determining the usability of the current information for the k-th future operating condition. Considering that a single lagging result has different degrees of impact on multiple subsequent operating conditions, the comprehensive effectiveness introduces a distance decay weight to weight the prediction accuracy of the subsequent N operating conditions, where N is the sliding time window length, used to limit the number of subsequent operating conditions participating in the comprehensive effectiveness evaluation. The value of N is determined based on the changes in the operating conditions. Frequency settings are used, with smaller values ​​(e.g., N=5) for high-frequency changes in operating conditions and larger values ​​(e.g., N=10) for low-frequency changes. For example, under conventional tunneling conditions, N=8 is set. λ is determined by fitting the rate of prediction accuracy decay with distance in historical data, used to assess the comprehensive contribution of current lag information over long time. E itself serves as a quantitative indicator of lag, with a threshold set at the lower quartile of the comprehensive effectiveness of historical associated operating condition combinations under the current geological type. This threshold is periodically updated with new sample accumulation, adapting to changes in operating conditions. When E is higher than this benchmark, the current lag information is considered to still have a positive effect on regulation; otherwise, it is considered that the lag constitutes a regulatory obstacle, and parameter adjustments based on this outdated information should be avoided. Traditional methods treat lag as a fixed time difference and ignore the value assessment of the lag information itself. This step transforms lag into a measurable effectiveness indicator, providing a quantitative basis for distinguishing between lag accumulation patterns and stochastic accumulation patterns.

[0025] S3-2: Based on the comprehensive effectiveness and comprehensive stochasticity index, determine the joint rate of change, dominant direction of change, and long-term evolution trend of lag degree and stochasticity, including: Each associated working condition combination is represented as a two-dimensional feature point. The horizontal axis of the two-dimensional feature point is the comprehensive effectiveness, and the vertical axis of the two-dimensional feature point is the comprehensive randomness index. The two-dimensional feature points of N consecutive associated working condition combinations are connected in the order of occurrence to form a feature trajectory. Calculate the Euclidean distance between adjacent two-dimensional feature points in the feature trajectory as the combined rate of change of hysteresis and randomness; The angle of change of the direction of the line connecting adjacent two-dimensional feature points in the feature trajectory is calculated as the dominant direction of the change in lag degree and randomness; The cumulative offset vector from the first two-dimensional feature point to the last two-dimensional feature point in the feature trajectory is calculated as a long-term evolution trend of lag degree and randomness.

[0026] It should be noted that the lag degree and the randomness index are placed on the horizontal and vertical axes to form a two-dimensional plane because the two indicators have different dimensions and are coupled with each other. Analyzing either indicator alone cannot reveal the dynamic relationship between the two. In the plane, each associated working condition combination is mapped as a feature point and connected in chronological order to form a feature trajectory. This transforms the joint evolution process of lag and randomness from abstract numerical values ​​into a visualized geometric path, which is convenient for subsequent extraction of morphological features. Euclidean distance measures the spatial interval between adjacent feature points on a plane, reflecting the combined change in hysteresis and stochastic index per unit time. This distance is defined as the joint rate of change and is used to determine the occurrence of abrupt changes in the system state. The angle between the line connecting adjacent feature points and the horizontal axis is determined by the ratio of the change in hysteresis to the change in stochastic index. This angle is defined as the dominant direction of change and is used to identify whether the current change is dominated by hysteresis or stochasticity. The cumulative offset vector from the first feature point to the last feature point integrates the total amount and direction of the offset of hysteresis and stochasticity during long-term evolution. This vector is defined as the long-term evolution trend and is used to determine whether the two cumulatively enhance each other or cancel each other out on the time axis.

[0027] S4-1: The current evolutionary model is obtained based on the joint rate of change, the dominant direction of change, and the long-term evolutionary trend, including: Collect sample data of the dominant change direction in all historical characteristic trajectories, and plot the probability distribution histogram of the dominant change direction; When the probability distribution histogram has a bimodal feature, the angle values ​​of the directional changes corresponding to the two peaks are taken, and the valley angle value between the two peaks is used as the classification boundary to distinguish the dominant direction of change. When the probability distribution histogram does not have a bimodal feature, the median of all dominant directions of change is taken as the classification boundary; When the absolute value of the M consecutive dominant directions of change in the feature trajectory is greater than the classification boundary and the randomness component of the long-term evolution trend is greater than the lag component, the current evolution pattern is identified as a random accumulation pattern. When the absolute value of the M consecutive dominant directions of change in the feature trajectory is less than the classification boundary and the lag component of the long-term evolution trend is greater than the random component, the current evolution pattern is identified as a lag accumulation pattern. When the joint rate of change in the characteristic trajectory is consistently greater than the average historical joint rate of change, the current evolution mode is identified as an unstable oscillation mode. When the joint rate of change in the feature trajectory is consistently less than the historical average joint rate of change and the two-dimensional feature points are clustered in a small area, the current evolutionary pattern is identified as a stable convergence pattern.

[0028] It should be noted that the reference Figure 3By plotting a histogram of the probability distribution of the dominant change direction, the concentration range and distribution pattern of the indicator in historical data can be intuitively grasped. When a bimodal feature appears, it indicates that there are two typical dominant directions in the historical data. In this case, the angle of the valley between the two peaks is taken as the classification boundary, which can effectively classify new samples into the random accumulation side or the lagged accumulation side. If the histogram does not show a bimodal feature, the median is used as the boundary to ensure the robustness of the classification. The M value is set according to the actual operating condition change frequency, usually between 3 and 5, to filter out instantaneous disturbances and avoid masking the evolution trend. The specific method for determining the M value is: set according to the sampling frequency of the state space dimension and the average time interval of the operating condition change, ensuring that the time span of M consecutive operating conditions covers at least one complete operating condition response cycle. As an example, when the sampling frequency is 1Hz and the average operating condition change interval is 5 seconds, M=3; when the sampling frequency is 0.2 ... At 20 seconds, M=5. When the dominant direction of M consecutive changes is biased towards the random side and the random component of the long-term evolution trend is dominant, it is judged as a random accumulation mode, indicating that randomness is continuously amplified in time. Conversely, when the direction is biased towards the lag side and the lag component is dominant, it is judged as a lag accumulation mode, and the lag effect continues to intensify. Among them, the long-term evolution trend vector is decomposed into horizontal axis components and vertical axis components. The horizontal axis component represents the cumulative amount of lag, and the vertical axis component represents the cumulative amount of randomness. The two are directly compared by the length of the vector's projection on the coordinate axis. If the joint rate of change continuously exceeds the historical average, it indicates frequent state changes and is classified as an unstable oscillation mode. If it is continuously lower than the historical average and the feature points are highly clustered, the system tends to be stable and is classified as a stable convergence mode. Traditional regulation lacks dynamic identification of evolution patterns. This step transforms the coupling relationship between lag and randomness into a categorizable geometric feature, providing a clear decision basis for differentiated regulation.

[0029] S4-2: Determine real-time control methods based on the current evolutionary model, including: When the evolution mode is a stochastic accumulation mode, increase the sampling frequency of the state space dimension to shorten the first time window; When the evolution mode is a lag-cumulative mode, the evolution trend of the characteristic trajectory of the preceding working condition is used as a benchmark to predict the direction of rheological change after the occurrence of the subsequent working condition and adjust the additive injection parameters in advance. It should be noted that the prediction method is as follows: take the most recent feature points in the current feature trajectory and perform linear fitting, and use the extrapolated direction as the predicted value of the rheological change direction of the subsequent working conditions. The number of feature points is set according to the degree of fluctuation of the feature trajectory. The greater the fluctuation, the fewer points are taken to ensure trend sensitivity, and the smaller the fluctuation, the more points are taken to ensure prediction stability. The additive injection parameters are adjusted one level of injection rate in advance according to the prediction direction. The level division is set according to historical control experience. When the evolution mode is an unstable oscillation mode, the active control action based on the combination of related operating conditions is suspended, the basic additive injection rate is maintained, and active control is resumed after the characteristic trajectory reconverges. When the evolution mode is a stable convergence mode, the additive injection parameters are adjusted according to the real-time detected rheological changes. During the tunneling process, new associated working condition combination data are continuously collected, new two-dimensional feature points are added to the feature trajectory, and the sliding time window keeps the feature trajectory length fixed. The above steps are repeated to achieve continuous adaptive updates of the control strategy.

[0030] It should be noted that, in the case of stochastic accumulation mode, the uncertainty of state dependence continues to amplify, and simply relying on lag feedback can no longer suppress fluctuations. At this time, increasing the sampling frequency of the state space dimension can compress the first time window, allowing the control system to capture the details of stochastic evolution with denser real-time data, thereby curbing the growth of stochasticity in its early stages. In the lag accumulation mode, the calculation results lag behind the actual operating conditions for a long time. Any adjustment based on the current value will lag behind the process. Therefore, it is necessary to predict the direction of rheological changes in subsequent operating conditions based on the evolution trend of characteristic trajectories, and adjust the additive injection parameters to the predicted values ​​in advance, using feedforward compensation to offset the time difference caused by lag. In the unstable oscillation mode, the joint rate of change far exceeds the historical average, indicating that lag and stochasticity are mutually reinforcing each other. In the case of positive feedback, if active control based on the combination of related operating conditions continues, any adjustment action may become a source of oscillation. Therefore, active control is paused, and only the basic additive injection rate is maintained. Intervention is resumed after the characteristic trajectory converges again. In the stable convergence mode, the characteristic points are clustered in a small area and the joint change rate is stable, indicating that the system has entered a stable state. At this time, it is only necessary to perform conventional feedback adjustment based on the real-time detected rheological change to maintain the balance. Traditional methods use the same set of control parameters indiscriminately for the above four states. This step achieves precise switching of control strategies through evolutionary pattern recognition, decomposes the dynamic response lag and state-dependent instability problems into sub-scenarios that can be dealt with in a targeted manner, and significantly improves the control stability and adaptability under complex operating conditions.

[0031] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for real-time adjustment of rheological properties of muck in a tunneling machine, characterized in that, Includes the following steps: S1: Arrange the tunnel boring machine's operating conditions in chronological order according to changes in geological conditions and equipment conditions to form a sequence of operating conditions and determine the associated combinations of operating conditions. S2: Collect the moisture content, composition, viscosity and rheological parameters of the slag and use them as the state space dimension, and determine the comprehensive stochasticity index based on the state space dimension; S3: Determine the comprehensive effectiveness of the current associated working condition combination for subsequent associated working condition combinations, and determine the joint rate of change of lag degree and randomness, the dominant direction of change and the long-term evolution trend based on the comprehensive effectiveness and the comprehensive randomness index; S4: Obtain the current evolutionary model based on the joint rate of change, the dominant direction of change, and the long-term evolutionary trend, and determine the real-time control method based on the current evolutionary model.

2. The method according to claim 1, wherein the method is characterized in that, The process of arranging the tunnel boring machine's operating conditions in chronological order based on changes in geological conditions and equipment conditions to form a sequence of operating conditions and determining associated combinations of operating conditions includes: For any two consecutive working conditions in the working condition sequence, the relationship between the preceding and subsequent working conditions is divided into enhancing influence relationship, offsetting influence relationship and no significant influence relationship according to the superposition effect of the influence of the preceding and subsequent working conditions on the rheology of the slag soil. Combinations of operating conditions that have reinforcing or offsetting effects are defined as associated operating condition combinations.

3. The method according to claim 2, wherein the method is characterized in that, The classification of the relationship between preceding and subsequent operating conditions into reinforcing influence relationships, offsetting influence relationships, and no significant influence relationships includes: Enhanced influence relationship refers to the unidirectional shift in the rheological properties of slag and soil caused by preceding and subsequent working conditions. The offsetting effect relationship refers to the fact that the preceding and subsequent working conditions cause the rheological properties of the slag and soil to shift in opposite directions. No significant influence relationship means that the superposition effect of the preceding and subsequent working conditions does not cause a identifiable directional shift in the rheological properties of the slag.

4. The method according to claim 1, wherein the method is characterized in that, The collected data on soil moisture content, composition, viscosity, and rheological parameters are used as dimensions of the state space. Based on these state space dimensions, a comprehensive stochastic index is determined, including: Accuracy is obtained by calculating the sum of squares and the square root of the squares for each dimension of the state space. The real-time detection data of each state space dimension at the current time point are used as the initial state vector of the construction waste. If the real-time detection data is lower than the highest accuracy value among the historical observation data, the historical observation data is selected to form the initial state vector of the slag and soil. The difference between the initial state vector of the slag before the occurrence of the preceding working condition and the state vector of the slag after the occurrence of the preceding working condition is taken as the change in the rheological properties of the slag under the preceding working condition. The difference between the initial state vector of the slag before the occurrence of the subsequent working condition and the state vector of the slag after the occurrence of the subsequent working condition is taken as the change in the rheological properties of the slag under the subsequent working condition. The superposition value of the rheological changes of the slag and soil under the preceding and subsequent working conditions is taken as the rheological change of the associated working condition combination. For the same type of associated working condition combination, under different initial state vectors of slag soil, the rheological change of the associated working condition combination is repeatedly obtained to form a rheological change sample set; The variance, Jacobian matrix norm, and confidence interval width of the rheological change sample set are calculated, and then normalized and weighted to obtain the comprehensive randomness index.

5. The method for real-time control of the rheological properties of tunnel boring machine excavated soil according to claim 1, characterized in that, The determination of the overall effectiveness of subsequent associated operating condition combinations based on the current associated operating condition combination includes: The time required from the initial collection of soil moisture content, soil composition, soil viscosity and rheological parameters to the completion of accuracy index calculation and the formation of the initial state vector of the soil is taken as the first time window; The time required from the occurrence of the preceding condition in the associated working condition combination to the completion of the calculation of the rheological change of the associated working condition combination is used as the second time window. If the first time window is less than or equal to the second time window, the matching status of the associated working condition combination is determined to be a successful match; If the first time window is larger than the second time window, the matching status of the associated working condition combination is determined to be a matching failure. The calculation results of the rheological changes corresponding to the failed matching of associated load case combinations are used to predict the accuracy of the k-th associated load case combination. And calculate the overall effectiveness of this associated working condition combination for the subsequent N associated working condition combinations. .

6. The method for real-time control of the rheological properties of tunnel boring machine excavated soil according to claim 5, characterized in that, The accuracy of predicting the k-th associated working condition combination for the calculation results of the rheological change corresponding to the failed matching combination includes: The formula for prediction accuracy is: ,in, This is the current calculation result. This represents the actual rheological change in subsequent combinations of related operating conditions. This represents the range of values ​​for the rheological change.

7. The method for real-time control of the rheological properties of tunnel boring machine excavated soil according to claim 5, characterized in that, The calculation of the overall effectiveness of the associated operating condition combination for the subsequent N associated operating condition combinations includes: Selecting the subsequent N related working condition combinations as the contribution range, the comprehensive effectiveness formula is as follows: , ,in, For distance decay weights, The attenuation coefficient; When the overall effectiveness is higher than the set threshold, the associated working condition combination that failed to match is determined to be effective and the calculation lag does not constitute a control obstacle. When the overall effectiveness is below the set threshold, the associated working condition combination that fails to match is deemed to be ineffective and the calculation lag constitutes a control obstacle. The overall effectiveness is used as the degree of lag; the higher the overall effectiveness, the smaller the impact of the lag on regulation.

8. The method for real-time control of the rheological properties of tunnel boring machine excavated soil according to claim 1, characterized in that, The determination of the joint rate of change, dominant direction of change, and long-term evolution trend of lag degree and randomness based on comprehensive effectiveness and comprehensive stochasticity index includes: Each associated working condition combination is represented as a two-dimensional feature point. The horizontal axis of the two-dimensional feature point is the comprehensive effectiveness, and the vertical axis of the two-dimensional feature point is the comprehensive randomness index. The two-dimensional feature points of N consecutive associated working condition combinations are connected in the order of occurrence to form a feature trajectory. Calculate the Euclidean distance between adjacent two-dimensional feature points in the feature trajectory as the combined rate of change of hysteresis and randomness; The angle of change of the direction of the line connecting adjacent two-dimensional feature points in the feature trajectory is calculated as the dominant direction of the change in lag degree and randomness; The cumulative offset vector from the first two-dimensional feature point to the last two-dimensional feature point in the feature trajectory is calculated as a long-term evolution trend of lag degree and randomness.

9. A method for real-time control of the rheological properties of tunnel boring machine excavated soil according to claim 1, characterized in that, The current evolutionary model is obtained based on the joint rate of change, the dominant direction of change, and the long-term evolutionary trend, including: Collect sample data of the dominant change direction in all historical characteristic trajectories, and plot the probability distribution histogram of the dominant change direction; When the probability distribution histogram has a bimodal feature, the angle values ​​of the directional changes corresponding to the two peaks are taken, and the valley angle value between the two peaks is used as the classification boundary to distinguish the dominant direction of change. When the probability distribution histogram does not have a bimodal feature, the median of all dominant directions of change is taken as the classification boundary; When the absolute value of the M consecutive dominant directions of change in the feature trajectory is greater than the classification boundary and the random component of the long-term evolution trend is greater than the lag component, the current evolution pattern is identified as a random cumulative pattern. When the absolute value of the M consecutive dominant directions of change in the feature trajectory is less than the classification boundary and the lag component of the long-term evolution trend is greater than the random component, the current evolution pattern is identified as a lag accumulation pattern. When the joint rate of change in the characteristic trajectory is consistently greater than the average historical joint rate of change, the current evolution mode is identified as an unstable oscillation mode. When the joint rate of change in the feature trajectory is consistently less than the historical average joint rate of change and the two-dimensional feature points are clustered in a small area, the current evolutionary pattern is identified as a stable convergence pattern.

10. A method for real-time control of the rheological properties of tunnel boring machine excavated soil according to claim 1, characterized in that, The method for determining real-time regulation based on the current evolutionary model includes: When the evolution mode is a stochastic accumulation mode, increase the sampling frequency of the state space dimension to shorten the first time window; When the evolution mode is a lag-cumulative mode, the evolution trend of the characteristic trajectory of the preceding working condition is used as a benchmark to predict the direction of rheological change after the occurrence of the subsequent working condition and adjust the additive injection parameters in advance. When the evolution mode is an unstable oscillation mode, the active control action based on the combination of related operating conditions is suspended, the basic additive injection rate is maintained, and active control is resumed after the characteristic trajectory reconverges. When the evolution mode is a stable convergence mode, the additive injection parameters are adjusted according to the real-time detected rheological changes. During the tunneling process, new associated working condition combination data are continuously collected, new two-dimensional feature points are added to the feature trajectory, and the sliding time window keeps the feature trajectory length fixed. The above steps are repeated to achieve continuous adaptive updates of the control strategy.

Images