Ballast distribution identification method based on shield tunneling parameters

Abstract

The invention discloses a rock ballast distribution identification method based on shield tunneling parameters, and belongs to the field of rock ballast distribution identification, and the method comprises the following steps: obtaining multi-dimensional parameter data in a shield tunneling process, and preprocessing the multi-dimensional parameter data to obtain a preprocessed tunneling parameter data set; performing dimension reduction processing on the preprocessed tunneling parameter data set to obtain a dimension-reduced feature data set; determining an optimal clustering number based on the feature data set after dimension reduction; clustering the feature data set after dimension reduction based on the optimal clustering number to obtain a preliminary clustering result; removing noise data in the preliminary clustering result to obtain an optimized clustering result; based on the corresponding relation between the optimized clustering result and actual rock ballast distribution, obtaining a rock ballast distribution identification result; and a mapping relation between the shield tunneling parameters and the rock ballast distribution is established according to the rock ballast distribution identification result and the change trend of the shield tunneling parameters, and a correlation model of the parameters and the rock ballast distribution is obtained.

Description

Technical Field

[0001] The invention belongs to the technical field of rock slag distribution identification, and in particular relates to a rock slag distribution identification method based on shield tunneling parameters. Background Art

[0002] During shield tunnel construction, accurate identification of rock ballast distribution is crucial for optimizing excavation parameters, improving construction efficiency, and ensuring safety. However, traditional methods for identifying rock ballast distribution present numerous technical challenges. For one thing, the physical properties of rock ballast are complex and diverse, with its shape, particle size, hardness, and other characteristics varying significantly depending on geological conditions. This makes it difficult for a single identification method to accurately reflect the true distribution of rock ballast. Furthermore, existing identification methods often rely on manual experience or simple image processing techniques, which are not only inefficient but also susceptible to subjective factors, making them difficult to meet the real-time and precision requirements of modern shield construction.

[0003] Furthermore, the amount of multidimensional parameter data generated during shield tunneling is enormous and complex. Extracting effective features related to rock slag distribution from this massive amount of data is a key issue. Traditional data processing methods struggle to effectively handle this high-dimensional data, easily leading to information redundancy and noise interference, which in turn affects the accuracy of rock slag distribution identification.

[0004] The continuous development of shield construction technology has led to higher requirements for intelligent and automated rock slag distribution identification. Using advanced data processing and analysis techniques to establish a precise mapping between shield tunneling parameters and rock slag distribution, thereby achieving efficient and accurate identification of rock slag distribution, has become a pressing technical challenge in the current shield construction field. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention proposes a rock slag distribution identification method based on shield tunneling parameters to solve the problems existing in the above-mentioned prior art.

[0006] To achieve the above-mentioned object, the present invention provides a method for identifying rock slag distribution based on shield tunneling parameters, comprising:

[0007] Acquiring multidimensional parameter data during shield tunneling, and preprocessing the multidimensional parameter data to obtain a preprocessed tunneling parameter data set;

[0008] Performing dimensionality reduction processing on the pre-processed tunneling parameter data set to obtain a reduced-dimensional feature data set;

[0009] Determining the optimal number of clusters based on the feature data set after dimensionality reduction;

[0010] Clustering the feature data set after dimensionality reduction based on the optimal number of clusters to obtain a preliminary clustering result;

[0011] Remove the noise data in the preliminary clustering results to obtain optimized clustering results;

[0012] Based on the corresponding relationship between the optimized clustering results and the actual rock debris distribution, obtain the rock debris distribution recognition result;

[0013] Establish a mapping relationship between the shield tunneling parameters and the rock debris distribution by establishing the relationship between the rock debris distribution recognition result and the change trend of the shield tunneling parameters, and obtain the association model between the parameters and the rock debris distribution.

[0014] Optionally, the process of preprocessing the multi-dimensional parameter data to obtain the preprocessed tunneling parameter data set includes:

[0015] Obtain multi-dimensional parameter data such as the thrust force, torque value, rotation speed value, and tunneling speed during the shield tunneling process to form an original data set;

[0016] For the missing values in the original data set, use the interpolation method to fill them to obtain a complete data set;

[0017] In the complete data set, based on the preset threshold range, judge and remove the outliers to obtain a cleaned data set;

[0018] Adopt the normalization method to map the thrust force, torque value, rotation speed value, and tunneling speed in the cleaned data set to a unified scale to obtain a normalized data set;

[0019] Based on the normalized data set, obtain the preprocessed tunneling parameter data set.

[0020] Optionally, the process of dimension reduction processing on the preprocessed tunneling parameter data set to obtain a dimension-reduced feature data set includes:

[0021] Adopt the principal component analysis method to extract features from the preprocessed tunneling parameter data set and calculate the cumulative contribution rate of the feature components;

[0022] When the cumulative contribution rate exceeds the preset threshold, retain the feature components;

[0023] According to the retained feature components, generate a dimension-reduced feature data set and determine the dimension of the data set.

[0024] Optionally, the process of determining the optimal number of clusters based on the dimension-reduced feature data set includes:

[0025] Calculate the sum of squared errors of the dimension-reduced feature data set;

[0026] Based on the sum of squared errors, obtain a curve of the sum of squared errors changing with the number of clusters;

[0027] Analyze the changing trend of the curve based on the elbow method;

[0028] When the number of clusters increases, if the decrease in the sum of squared errors in the curve trend is less than the decrease in the sum of squared errors before the increase in the number of clusters, the number before the increase in the number of clusters is determined to be the optimal number of clusters.

[0029] Optionally, clustering the feature data set after dimensionality reduction based on the optimal number of clusters to obtain a preliminary clustering result includes:

[0030] Initialize the center point coordinates of the K-Means algorithm based on the optimal number of clusters;

[0031] The Euclidean distance is used to calculate the similarity measure between each data point and the center point coordinates in the feature data set after dimensionality reduction;

[0032] The feature data set after dimensionality reduction is clustered based on the similarity measure and the optimal number of clusters to obtain a preliminary clustering result.

[0033] Optionally, the process of removing noise data from the preliminary clustering result to obtain an optimized clustering result includes:

[0034] Using a density clustering method to filter the noise data of the preliminary clustering result;

[0035] When the neighborhood density of a data point in the preliminary clustering result is lower than a fixed threshold, the current data point is determined to be a noise point and removed to obtain denoised data;

[0036] Based on the denoised data, the coordinates of the center point of each cluster are calculated, and the distance between each denoised data point and the coordinates of the center point of the cluster to which it belongs is measured using the Euclidean distance;

[0037] When the distance value is greater than the outlier threshold, the current denoised data point is marked as an outlier, and the cluster center coordinates are recalculated to obtain the optimized cluster center;

[0038] Based on the optimized cluster centers, a linear regression method is used to perform trend analysis on the denoised data points in each cluster to obtain a change trend prediction value for each cluster;

[0039] According to the predicted value, the denoised data points in each cluster are redistributed to obtain the optimized clustering results.

[0040] Optionally, the process of obtaining the rock ballast distribution identification result includes:

[0041] The confusion matrix is used to calculate the matching degree between the optimized clustering results and the actual rock ballast distribution to obtain the evaluation value;

[0042] When the evaluation value is higher than the preset evaluation value, the optimized clustering result is determined to be a valid clustering;

[0043] According to the effective clustering results, the rock chip distribution identification results are obtained.

[0044] Optionally, the process of establishing a mapping relationship between shield tunneling parameters and rock slag distribution and obtaining a correlation model between the parameters and rock slag distribution includes:

[0045] The K-nearest neighbor algorithm is used to classify the areas in the rock debris distribution identification results to obtain the classification results;

[0046] Based on the classification results, a support vector machine is used to optimize the rock slag distribution boundary to obtain an optimized rock slag distribution map;

[0047] The DBSCAN algorithm is used to perform density analysis on the optimized rock slag distribution map to obtain a density distribution map;

[0048] Analyzing the aggregation characteristics of parameter changes based on the density distribution map to generate parameter change aggregation characteristics;

[0049] Based on the parameter variation aggregation characteristics and the variation trend of the shield tunneling parameters, a linear regression method is used to establish a mapping relationship between the shield tunneling parameters and the rock ballast distribution, and a correlation model between the parameters and the rock ballast distribution is obtained;

[0050] A rock slag distribution trend prediction value is obtained based on the correlation model.

[0051] Optionally, the method further comprises calculating statistical characteristics of the tunneling parameters based on the preprocessed tunneling parameter data set, and obtaining a change trend of the shield tunneling parameters based on the statistical characteristics.

[0052] Optionally, the process of obtaining a change trend of shield tunneling parameters based on the preprocessed tunneling parameter data set includes:

[0053] Calculating the mean, variance and standard deviation of each parameter in the preprocessed tunneling parameter data set to obtain the statistical characteristics;

[0054] Drawing a variation curve of propulsion force, torque value, rotation speed value and tunneling speed based on the statistical characteristics;

[0055] Extracting the distribution characteristics of each tunneling parameter based on the change curve, and drawing a distribution characteristic diagram to obtain the distribution characteristics of the tunneling parameter;

[0056] Using a support vector machine algorithm to classify the dynamic change patterns of the distribution characteristics of the tunneling parameters to obtain pattern categories;

[0057] generating a set of rules for changes in excavation parameters based on the pattern categories, and determining dynamic change rules for the excavation parameters;

[0058] Determine the change trend of the shield tunneling parameters based on the dynamic change rules of the tunneling parameters.

[0059] Compared with the prior art, the present invention has the following advantages and technical effects:

[0060] The present invention discloses a method for identifying the muck distribution based on shield tunneling parameters. The method first obtains multi-dimensional parameter data during the shield tunneling process. After preprocessing and feature extraction, principal component analysis is used for dimensionality reduction. Then, the K-Means clustering algorithm is used to cluster the dimensionality-reduced feature data, and the density clustering method is used to filter out the noise data. Finally, the corresponding relationship between the clustering results and the actual muck distribution is analyzed, and an association model between the tunneling parameters and the muck distribution is established. The present invention realizes the automatic identification of the muck distribution during the shield tunneling process in a data-driven manner, provides an important basis for optimizing the tunneling parameters, improving the construction efficiency and safety, and has significant engineering application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] The drawings forming a part of this application are used to provide a further understanding of this application. The schematic embodiments and descriptions thereof of this application are used to explain this application and do not constitute an improper limitation to this application. In the drawings:

[0062] Figure 1 is a flow chart of the method for identifying the muck distribution based on shield tunneling parameters according to the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0063] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The following will refer to the drawings and combine the embodiments to detail this application.

[0064] It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0065] Embodiment 1

[0066] As Figure 1 shown, in this embodiment, a method for identifying the muck distribution based on shield tunneling parameters is provided, including the following steps:

[0067] Step S101, obtain multi-dimensional parameter data such as the thrust force, torque, rotational speed, and tunneling speed during the shield tunneling process, clean the missing values and outliers in the data, and map the data to a unified scale using the normalization method to obtain the preprocessed tunneling parameter data set.

[0068] Furthermore, obtain multi-dimensional parameter data such as the thrust force, torque value, rotational speed value, and tunneling speed during the shield tunneling process to form an original data set; for the missing values in the original data set, use the interpolation method to fill them to obtain a complete data set; in the complete data set, based on a preset threshold range, judge and remove the outliers to obtain a cleaned data set; use the normalization method to map the thrust force, torque value, rotational speed value, and tunneling speed in the cleaned data set to a unified scale to obtain a normalized data set; based on the normalized data set, obtain a preprocessed tunneling parameter data set.

[0069] Furthermore, as a specific implementation manner of this embodiment, in the shield tunneling project, the thrust force is manifested as the resistance generated when the tunneling cutterhead contacts the soil body, usually remaining in the range of eight thousand to twelve thousand kilonewtons in soft soil strata, and can reach more than twenty thousand kilonewtons in hard rock strata. The torque value reflects the torque required for the cutterhead to rotate, about 1.5 million newton-meters in soft soil strata, and can reach 3 million newton-meters in hard rock strata. The rotational speed value is generally between 0.5 and 3 revolutions per minute, and the tunneling speed fluctuates within the range of 20 to 40 millimeters per minute. For the missing values in the original data, the linear interpolation method is used to fill them based on the data at adjacent times. If the thrust force data at a certain time is missing, the average value of the data at its previous and subsequent times can be taken for filling. For continuous missing cases, the polynomial interpolation method is used to fit the curve using the data of more adjacent points. In the data cleaning stage, the threshold range of the thrust force is set to 5,000 to 25,000 kilonewtons, and the data points outside this range are regarded as outliers. Similarly, the torque value is limited between 500,000 and 4 million newton-meters, the rotational speed value is between 0.3 and 4 revolutions per minute, and the tunneling speed is between 10 and 60 millimeters per minute. During the normalization process, each parameter is mapped to the interval of 0 to 1. If the thrust force at a certain time is 15,000 kilonewtons, the maximum value is 25,000 kilonewtons, and the minimum value is 5,000 kilonewtons, then the normalized value is 0.5. This processing makes the parameters with different dimensions comparable.

[0070] Step S102, calculate the mean value, variance, and standard deviation of each parameter in the preprocessed tunneling parameter data set, extract the statistical characteristics of the tunneling parameters, draw the parameter change curve and distribution characteristic diagram, analyze the dynamic change law of the parameters, and obtain the change trend and distribution characteristics of the parameters.

[0071] Furthermore, calculate the statistical characteristics of the tunneling parameters based on the preprocessed tunneling parameter data set, and obtain the change trend of the shield tunneling parameters based on the statistical characteristics.

[0072] The process of obtaining the change trend of the shield tunneling parameters based on the preprocessed tunneling parameter dataset includes: calculating the mean, variance, and standard deviation of each parameter in the preprocessed tunneling parameter dataset to obtain the statistical features; drawing the change curves of the thrust force, torque value, rotation speed value, and tunneling speed based on the statistical features; extracting the distribution features of each tunneling parameter based on the change curves, and drawing the distribution feature map to obtain the distribution features of the tunneling parameters; using the support vector machine algorithm to classify the dynamic change patterns of the distribution features of the tunneling parameters to obtain the pattern categories; generating a rule set for the change of the tunneling parameters based on the pattern categories to determine the dynamic change rules of the tunneling parameters; and determining the change trend of the shield tunneling parameters based on the dynamic change rules of the tunneling parameters.

[0073] Furthermore, as a specific implementation of this embodiment, for the shield tunneling data, statistical feature analysis needs to be carried out first. In this embodiment, among the tunneling data, the average value of the thrust force is three thousand kilonewtons, and the variance is one hundred kilonewton square, reflecting the fluctuation range of the thrust force. The average value of the torque is about five hundred kilonewton meters, and the standard deviation is about fifty kilonewton meters, indicating that the torque has good stability. The rotation speed value usually remains at 1.5 revolutions per minute, and the tunneling speed remains at about 30 millimeters per minute. When drawing the change curves for these parameters, it can be found that the thrust force shows a stepwise upward trend, which is related to the formation change and the increase in tunneling distance. The change curve of the torque value shows periodic fluctuations, and each period lasts about 30 minutes, which is closely related to the cutterhead rotation and the soil cutting process. The rotation speed value curve is relatively stable, and only shows an obvious decrease when changing the soil bin. The tunneling speed curve shows similar change characteristics to the thrust force. In terms of distribution characteristics, the thrust force data shows a normal distribution characteristic, concentrated in the range of three thousand to three thousand five hundred kilonewtons. The torque value shows a right-skewed distribution, mainly distributed in the range of four hundred to six hundred kilonewton meters. The distribution of the rotation speed value is relatively concentrated, showing a typical uniform distribution characteristic. The distribution of the tunneling speed is relatively dispersed, reflecting the flexibility of speed adjustment during the construction process. When the thrust force suddenly drops below two thousand kilonewtons, or the torque value exceeds seven hundred kilonewton meters, the system will mark it as an abnormal state. This abnormality may be related to formation changes, equipment failures, or improper operations. Under normal conditions, each parameter should fluctuate within the preset range, and the change trend should be gentle. Through linear regression prediction, it is found that the thrust force shows a linear upward trend with the increase of tunneling depth, increasing by about one hundred kilonewtons for every one hundred meters forward. The prediction result of the torque value shows that it is positively correlated with the formation hardness. The support vector machine algorithm further classifies these change patterns into three categories: stable tunneling, adjusted tunneling, and bin-changing tunneling. Finally, the rules generated by the decision tree algorithm show that when the thrust force is in the range of three thousand to three thousand five hundred kilonewtons, the torque value is in the range of four hundred to six hundred kilonewton meters, the rotation speed value is in the range of 1.3 to 1.7 revolutions per minute, and the tunneling speed is in the range of 25 to 35 millimeters per minute, the shield machine is in the best tunneling state. These rules provide important operation guidance for construction personnel.

[0074] Step S103, use the principal component analysis method to reduce the dimension of the preprocessed tunneling parameter data set, extract the main feature components, and if the cumulative contribution rate of the feature components exceeds the preset threshold, retain the component to obtain the reduced-dimension feature data set.

[0075] Furthermore, the process of reducing the dimension of the preprocessed tunneling parameter data set to obtain the reduced-dimension feature data set includes: using the principal component analysis method to extract features from the preprocessed tunneling parameter data set, calculating the cumulative contribution rate of the feature components; when the cumulative contribution rate exceeds the preset threshold, retain the feature components; according to the retained feature components, generate the reduced-dimension feature data set, and determine the dimension of the data set.

[0076] Obtain the preprocessed dataset of tunneling parameters, and use the principal component analysis method to extract features from the dataset. Calculate the cumulative contribution rate of the feature components. If the cumulative contribution rate exceeds the preset threshold, retain the feature components. Generate a feature dataset with reduced dimensions based on the retained feature components, and determine the dimension of the dataset.

[0077] Furthermore, as a specific implementation manner of this embodiment, the principal component analysis method extracts features from the multi-dimensional parameter data collected during the tunneling operation of the tunneling machine. The collected original data includes parameters such as thrust force, torque, rotational speed, and tunneling speed. In the analysis, first calculate the contribution rate of the feature components. The cumulative contribution rate of the first three principal components reaches 85%, exceeding the preset threshold of 80%, so these three feature components are retained. Through this dimension reduction method, the data dimension can be reduced while retaining key information, improving the efficiency of subsequent analysis. The feature dataset with reduced dimensions can be classified by a clustering algorithm.

[0078] Step S104, based on the feature dataset with reduced dimensions, use the elbow method to determine the optimal number of clusters of the K-Means clustering algorithm. If the decrease in the sum of squared errors is significantly reduced when the number of clusters increases, select this number as the optimal number of clusters to obtain the optimal number of clusters.

[0079] Further, the process of determining the optimal number of clusters based on the feature dataset with reduced dimensions includes: calculating the sum of squared errors of the feature dataset with reduced dimensions; obtaining a curve of the sum of squared errors changing with the number of clusters based on the sum of squared errors; analyzing the change trend of the curve based on the elbow method; when the number of clusters increases, if the decrease in the sum of squared errors in the change trend of the curve is less than the decrease in the sum of squared errors before the increase in the number of clusters, determine the number before the increase in the number of clusters as the optimal number of clusters.

[0080] For the feature dataset with reduced dimensions, calculate a curve of the sum of squared errors changing with the number of clusters. Use the elbow method to analyze the change trend of the curve. If the decrease in the sum of squared errors is significantly reduced, determine the current number of clusters as the optimal number of clusters.

[0081] Furthermore, as a specific implementation manner of this embodiment, the sum of squared errors is an important indicator for measuring the difference between data points and cluster centers in clustering analysis. Considering the parameter change situations during the construction process of the tunneling machine, such as tunneling speed, torque, and thrust force, these parameters will show different characteristics under different geological conditions. The elbow method determines the optimal number of clusters by observing the inflection point of the sum of squared errors curve. For example, when analyzing the data of a certain tunneling working condition, when the number of clusters increases from two to three, the sum of squared errors decreases significantly by 50%, while when increasing from three to four, it only decreases by 10%. At this time, three can be determined as the optimal number of clusters.

[0082] Step S105: Initialize the central points of the K-Means clustering algorithm according to the optimal number of clusters, use the Euclidean distance as the similarity measure, and cluster the dimensionality-reduced feature dataset to obtain a preliminary clustering result.

[0083] Furthermore, the process of clustering the dimensionality-reduced feature dataset based on the optimal number of clusters to obtain a preliminary clustering result includes: initializing the central point coordinates of the K-Means algorithm based on the optimal number of clusters; calculating the similarity measure values between each data point in the dimensionality-reduced feature dataset and the central point coordinates using the Euclidean distance; and performing clustering division on the dimensionality-reduced feature dataset based on the similarity measure values and the optimal number of clusters to obtain a preliminary clustering result.

[0084] More specifically, as a specific implementation manner of this embodiment, when using the K-Means algorithm for clustering, taking the tunneling speed and thrust force as examples, it is assumed that one hundred data points are collected in a certain tunneling section. Through iterative calculation, three clustering centers are finally obtained, which respectively represent three typical working conditions of low speed and low thrust, medium speed and medium thrust, and high speed and high thrust. The coordinates of the clustering centers reflect the average characteristics of this type of working condition. For example, the tunneling speed of the first type of central point is two meters per hour, and the thrust force is fifty kilonewtons. In terms of distance calculation, the Euclidean distance can reflect the similarity degree between the data point and the clustering center. Taking a certain data point as an example, its tunneling speed is 2.5 meters per hour, and the thrust force is 60 kilonewtons. By calculating its Euclidean distances from the three clustering centers, it is found that the distance from the second type of center is the smallest, indicating that this point belongs to the medium speed and medium thrust working condition. In terms of prediction, a linear regression model is established for each category respectively. Taking the first type of working condition as an example, it is analyzed that as the tunneling depth increases, the tunneling speed shows a gradually decreasing trend. The prediction shows that within the subsequent 50-meter section, the tunneling speed will drop from two meters per hour to 1.5 meters per hour. This trend prediction helps to adjust the tunneling parameters in a timely manner to avoid efficiency reduction or equipment damage caused by parameter mismatch. By analyzing different types of working conditions, the influence law of geological condition changes on tunneling parameters can be found. For example, in soft rock sections, the tunneling speed and thrust force usually belong to the high speed and high thrust category, while in hard rock sections, they mostly show the characteristics of low speed and low thrust. This classification result provides an important basis for subsequent parameter optimization and control strategy formulation. Based on the clustering result and the prediction trend, a corresponding parameter adjustment plan can be formulated. For example, when it is predicted that the hard rock section is about to be entered, the tunneling speed and thrust force are reduced in advance to achieve a smooth transition of the tunneling parameters.

[0085] Step S106: For the noise data in the preliminary clustering result, use the density clustering method for filtering. If the neighborhood density of the data point is lower than the preset threshold, it is determined as a noise point and eliminated to obtain a denoised clustering result.

[0086] Further, the process of removing noise data from the preliminary clustering result to obtain an optimized clustering result includes: filtering the noise data of the preliminary clustering result by using a density clustering method; when the neighborhood density of a data point in the preliminary clustering result is lower than a fixed threshold, determining that the current data point is a noise point and removing it to obtain denoised data; based on the denoised data, calculating the center point coordinates of each cluster, and using the Euclidean distance to measure the distance value between each denoised data point and the center point coordinates of the cluster to which it belongs; when the distance value is greater than an outlier threshold, marking the current denoised data point as an outlier, and recalculating the cluster center coordinates to obtain an optimized cluster center; based on the optimized cluster center, using a linear regression method to perform trend analysis on the denoised data points in each cluster to obtain the change trend prediction value of each cluster; according to the prediction value, reassigning the denoised data points in each cluster to obtain an optimized clustering result.

[0087] Step S107: Analyze the correspondence between the optimized clustering result and the actual rock debris distribution, and use a confusion matrix to evaluate the accuracy of the clustering result. If the matching degree between the clustering result and the actual distribution is higher than a preset threshold, it is determined as an effective clustering to obtain the rock debris distribution recognition result.

[0088] Further, the process of obtaining the rock debris distribution recognition result includes: calculating the matching degree between the optimized clustering result and the actual rock debris distribution by using a confusion matrix to obtain an evaluation value; when the evaluation value is higher than a preset evaluation value, determining that the optimized clustering result is an effective clustering; according to the effective clustering result, obtaining the rock debris distribution recognition result.

[0089] Furthermore, as a specific implementation of this embodiment, the confusion matrix is an important tool for evaluating the accuracy of clustering results. The matching degree is calculated by statistically analyzing the correspondence between the actual rock debris distribution and the clustering results. In the field application of a quarry, the clustering results show that the area of the rock debris distribution region is 100 square meters. Compared with the actual rock debris distribution area of 98 square meters, the matching degree reaches 95%, exceeding the preset threshold of 90%. Therefore, it is determined as a valid clustering. After confirming the clustering validity, the K-nearest neighbor algorithm is used to classify the rock debris distribution region. Taking a mountain quarry as an example, the rock debris is divided into three categories: coarse particle area, medium particle area, and fine particle area according to the particle size. By selecting appropriate neighborhood parameters, the system automatically divides the rock debris with similar characteristics into the same category, forming clear distribution region boundaries. The support vector machine optimization technology can accurately define the rock debris distribution boundary. In the application of an open-pit mine, by establishing the optimal classification hyperplane, the originally fuzzy rock debris accumulation boundary is optimized into a clear demarcation line. The optimized rock debris distribution map shows that the boundary positioning accuracy is improved to the centimeter level, providing an accurate reference for subsequent construction. When the principal component analysis method is used to extract the rock debris distribution characteristics, key indicators such as particle size distribution, stacking height, and density are focused on. At a project site, more than a dozen characteristic indicators are simplified into three principal components through dimensionality reduction processing, retaining 95% of the original information and greatly improving the data processing efficiency. The density clustering algorithm DBSCAN can effectively analyze the rock debris distribution density. In an application case of a quarry, by setting appropriate neighborhood radii and minimum sample numbers, the system identifies three high-density regions and two low-density regions, providing a scientific basis for stockyard management. The rock debris accumulation volume in the high-density region reaches 1.8 tons per cubic meter, while that in the low-density region is only 1.2 tons. When linear regression is used to predict the rock debris change trend, variables such as mining intensity and climate factors are comprehensively considered. The trend analysis of an open-pit mining area shows that under normal mining conditions, the rock debris accumulation volume shows periodic changes, with a monthly growth rate of about 5%. Based on this, it is predicted that the total accumulation volume next month will reach 80,000 cubic meters. These predicted data provide important references for production plan adjustment and equipment configuration optimization. Through the comprehensive application of these technical means, a complete rock debris distribution analysis and prediction system is established. The system can not only accurately identify the current rock debris distribution status but also predict future change trends, providing strong support for the safe production and efficiency improvement of the quarry.

[0090] Step S108: Compare the rock debris distribution recognition result with the change trend of the shield tunneling parameters, and use the regression analysis method to establish the mapping relationship between the parameters and the rock debris distribution, obtaining the correlation model between the parameters and the rock debris distribution.

[0091] Further, the process of establishing the mapping relationship between the shield tunneling parameters and the rock debris distribution and obtaining the correlation model between the parameters and the rock debris distribution includes: classifying the regions in the rock debris distribution recognition result by using the K-nearest neighbor algorithm to obtain the classification result; based on the classification result, optimizing the rock debris distribution boundary by using the support vector machine to obtain the optimized rock debris distribution map; performing density analysis on the optimized rock debris distribution map by using the DBSCAN algorithm to obtain the density distribution map; based on the density distribution map, analyzing the aggregation characteristics of the parameter changes to generate the aggregation characteristics of the parameter changes; based on the aggregation characteristics of the parameter changes and the change trend of the shield tunneling parameters, establishing the mapping relationship between the shield tunneling parameters and the rock debris distribution by using the linear regression method to obtain the correlation model between the parameters and the rock debris distribution; and obtaining the predicted value of the rock debris distribution trend based on the correlation model.

[0092] Furthermore, as a specific implementation of this embodiment, shield tunneling parameter trend analysis extracts real-time monitoring data of key parameters such as cutterhead torque, thrust, and silo pressure. Taking cutterhead torque as an example, when encountering rock debris accumulation during tunneling, torque will fluctuate significantly. By establishing a mapping relationship between torque changes and rock debris distribution, rock debris accumulation can be predicted. In a certain project in this embodiment, a sudden increase in torque exceeding 30% of the baseline value often indicates that the cutterhead is passing through an area of rock debris accumulation. When extracting key parameters using principal component analysis, multiple parameters such as silo pressure, thrust speed, and cutterhead speed can be subjected to dimensionality reduction. By calculating the contribution of each parameter to rock debris distribution, the parameter combinations with the most significant impact on rock debris distribution are selected. It was found that the combined changes in silo pressure and thrust speed can explain 85% of the changes in rock debris distribution. When using the nearest neighbor algorithm for parameter classification, the earth pressure equilibrium state can be used as the classification basis. When the silo pressure is within a preset range and the cutterhead torque fluctuations are minimal, the tunneling state is determined to be stable. Otherwise, there may be a risk of rock slag accumulation. When optimizing parameter distribution boundaries using a support vector machine, focus is placed on unusual parameter combinations, such as an abnormal increase in silo pressure and a sudden decrease in advance speed. During the density analysis phase, a density-based clustering method is used to identify clusters of parameter changes. For example, in a certain project section, high-density clustering of silo pressure and cutterhead torque within a specific range of values was observed, which was confirmed by analysis to be a typical characteristic of rock slag accumulation. This clustering characteristic can be used to construct an early warning system. When analyzing parameter change clustering characteristics, focus on the changing patterns of parameter combinations. As a specific implementation of this embodiment, when silo pressure continues to rise while cutterhead torque exhibits periodic fluctuations, it often indicates the formation of rock slag accumulation. Using linear regression, a predictive model is established that combines these characteristics with rock slag distribution. The development of the predictive model requires considering the synergistic effects of multiple parameters. When multiple parameters, such as silo pressure and cutterhead torque, exhibit abnormalities simultaneously, the system will issue a rock slag accumulation risk warning. The accuracy of the predicted values can be verified using historical data. In a specific project, this method has demonstrated a prediction accuracy of over 85%. This prediction model can effectively guide construction personnel to adjust excavation parameters in a timely manner and prevent the risk of rock debris accumulation.

[0093] The above are merely preferred embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A method for identifying the distribution of rock debris based on shield tunneling parameters, characterized in that, The following steps are involved: Acquiring multidimensional parameter data during shield tunneling, and preprocessing the multidimensional parameter data to obtain a preprocessed tunneling parameter data set; Performing dimensionality reduction processing on the pre-processed tunneling parameter data set to obtain a reduced-dimensional feature data set; Determining the optimal number of clusters based on the feature data set after dimensionality reduction; Clustering the feature data set after dimensionality reduction based on the optimal number of clusters to obtain a preliminary clustering result; Removing noise data from the preliminary clustering results to obtain optimized clustering results; Based on the correspondence between the optimized clustering result and the actual rock slag distribution, a rock slag distribution identification result is obtained; The rock slag distribution identification result and the change trend of the shield tunneling parameters are used to establish a mapping relationship between the shield tunneling parameters and the rock slag distribution, and a correlation model between the parameters and the rock slag distribution is obtained.

2. The method for identifying the rock debris distribution based on the shield tunneling parameters according to claim 1, wherein The process of preprocessing the multidimensional parameter data to obtain a preprocessed tunneling parameter data set includes: Obtain multi-dimensional parameter data of propulsion force, torque value, rotation speed value and tunneling speed during shield tunneling to form an original data set; For the missing values in the original data set, interpolation method is used to fill them and obtain a complete data set; In the complete data set, based on the preset threshold range, outliers are judged and removed to obtain the cleaned data set; The normalization method is used to map the propulsion force, torque value, rotation speed value and tunneling speed in the cleaned data set to a unified scale to obtain a normalized data set. A preprocessed tunneling parameter data set is obtained based on the normalized data set.

3. The method for identifying the rock debris distribution based on the shield tunneling parameters according to claim 2, wherein The process of performing dimensionality reduction processing on the pre-processed tunneling parameter data set to obtain a reduced-dimensional feature data set includes: The principal component analysis method is used to extract features from the pre-processed tunneling parameter data set, and the cumulative contribution rate of the characteristic components is calculated; When the cumulative contribution rate exceeds the preset threshold, the characteristic component is retained; Based on the retained feature components, a feature dataset with dimensionality reduction is generated, and the dimension of the dataset is determined.

4. The method for identifying the distribution of rock debris based on shield tunneling parameters according to claim 3, wherein The process of determining the optimal number of clusters based on the feature data set after dimensionality reduction includes: Calculating the sum of squared errors of the feature data set after dimensionality reduction; Obtaining a curve showing how the sum of squared errors varies with the number of clusters based on the sum of squared errors; Analyze the changing trend of the curve based on the elbow method; When the number of clusters increases, if the decrease in the sum of squared errors in the curve trend is less than the decrease in the sum of squared errors before the increase in the number of clusters, the number before the increase in the number of clusters is determined to be the optimal number of clusters.

5. The method for identifying the distribution of rock debris based on shield tunneling parameters according to claim 4, wherein The process of clustering the feature data set after dimensionality reduction based on the optimal number of clusters to obtain a preliminary clustering result includes: Initialize the center point coordinates of the K-Means algorithm based on the optimal number of clusters; The Euclidean distance is used to calculate the similarity measure between each data point and the center point coordinates in the feature data set after dimensionality reduction; The feature data set after dimensionality reduction is clustered based on the similarity measure and the optimal number of clusters to obtain a preliminary clustering result.

6. The method for identifying muck distribution based on shield tunneling parameters according to claim 5, wherein, The process of removing noise data from the preliminary clustering result to obtain an optimized clustering result includes: Using a density clustering method to filter the noise data of the preliminary clustering result; When the neighborhood density of a data point in the preliminary clustering result is lower than a fixed threshold, it is determined that the current data point is a noise point and is removed to obtain denoised data; Based on the denoised data, calculate the center point coordinates of each cluster, and use the Euclidean distance to measure the distance value between each denoised data point and the center point coordinates of its affiliated cluster; When the distance value is greater than the outlier threshold, mark the current denoised data point as an outlier, and recalculate the cluster center coordinates to obtain an optimized cluster center; Based on the optimized cluster center, use the linear regression method to perform trend analysis on the denoised data points in each cluster to obtain the change trend prediction values of each cluster; According to the prediction values, reassign the denoised data points in each cluster to obtain an optimized clustering result.

7. The method for identifying the distribution of rock debris based on shield tunneling parameters according to claim 6, characterized in that, The process of obtaining the rock debris distribution recognition result includes: Use a confusion matrix to calculate the matching degree between the optimized clustering result and the actual rock debris distribution to obtain an evaluation value; When the evaluation value is higher than the preset evaluation value, it is determined that the optimized clustering result is a valid cluster; According to the valid clustering result, obtain the rock debris distribution recognition result.

8. The method for identifying the distribution of rock debris based on shield tunneling parameters according to claim 7, wherein The process of establishing the mapping relationship between the shield tunneling parameters and the rock debris distribution to obtain the association model between the parameters and the rock debris distribution includes: Use the K-nearest neighbor algorithm to classify the regions in the rock debris distribution recognition result to obtain a classification result; Based on the classification result, use the support vector machine to optimize the rock debris distribution boundary to obtain an optimized rock debris distribution map; Use the DBSCAN algorithm to perform density analysis on the optimized rock debris distribution map to obtain a density distribution map; Based on the density distribution map, analyze the aggregation characteristics of parameter changes to generate parameter change aggregation characteristics; Based on the parameter change aggregation characteristics and the change trend of the shield tunneling parameters, use the linear regression method to establish the mapping relationship between the shield tunneling parameters and the rock debris distribution to obtain the association model between the parameters and the rock debris distribution; Based on the association model, obtain the rock debris distribution trend prediction value.

9. The method for identifying the distribution of rock debris based on shield tunneling parameters according to claim 8, wherein The method further includes calculating the statistical characteristics of the tunneling parameter based on the preprocessed tunneling parameter dataset, and obtaining the change trend of the shield tunneling parameter based on the statistical characteristics.

10. The method for identifying the distribution of rock debris based on shield tunneling parameters according to claim 9, characterized in that, The process of obtaining the change trend of the shield tunneling parameter based on the preprocessed tunneling parameter dataset includes: Calculate the mean, variance, and standard deviation of each parameter in the preprocessed tunneling parameter dataset to obtain the statistical characteristics; Based on the statistical characteristics, draw the change curves of the thrust force, torque value, rotation speed value, and tunneling speed; Based on the change curves, extract the distribution characteristics of each tunneling parameter, and draw the distribution characteristic map to obtain the distribution characteristics of the tunneling parameter; Use the support vector machine algorithm to perform dynamic change pattern classification on the distribution characteristics of the tunneling parameter to obtain a pattern category; Based on the pattern category, generate a rule set for the change of the tunneling parameter, and determine the dynamic change rule of the tunneling parameter; Based on the dynamic change rule of the tunneling parameter, determine the change trend of the shield tunneling parameter.

Images