TBM tunneling parameter intelligent analysis and recommendation method and system

Abstract

The invention provides a TBM tunneling parameter intelligent analysis and recommendation method and system, and the method comprises the steps: sorting key tunneling parameters, calculating the slope of adjacent parameters in the key tunneling parameters, and determining the reasonable distribution interval of the key tunneling parameters; using a kmeans algorithm to perform segmented clustering on cutterhead torque and propulsion speed data, and performing classified calculation according to a clustering result to judge a stratum surrounding rock grade probability; based on the working condition mode, piecewise linear fitting is carried out on the current tunneling parameters, an envelope line of piecewise linear fitting is calculated, and ideal torque, thrust and corresponding propelling speed are recommended; and executing a multi-dimensional fusion algorithm to score the recommended or actual tunneling parameters, and dynamically adjusting the mode weight to optimize the recommendation model. By means of the technical scheme, the problems that in the prior art, TBM tunneling parameter setting depends on experience, and intelligent analysis and multi-parameter comprehensive consideration are lacked can be effectively solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of geological exploration and tunnel construction, and particularly relates to a TBM tunneling parameter intelligent analysis and recommendation method and system. BACKGROUND

[0002] With the continuous progress of tunnel construction technology, TBM is increasingly widely used in hard rock tunnel construction. However, the parameter setting and optimization in the TBM tunneling process is still a complex and critical problem. The traditional method mainly relies on the experience of engineers and field tests, which not only consumes time and effort, but also is difficult to ensure the accuracy of parameter setting and the maximization of tunneling efficiency. In recent years, although there have been some data-driven tunneling parameter analysis methods, these methods are often limited to simple statistics of a single parameter, lacking a comprehensive understanding and effective use of the complex relationship between multiple parameters in the tunneling process.

[0003] The existing technology often only focuses on the statistical distribution of a single tunneling parameter, ignoring the interaction and mutual influence between parameters, resulting in limited parameter optimization effect. Lack of intelligent analysis methods based on big data and machine learning, unable to realize accurate prediction and optimization recommendation of tunneling parameters. The evaluation of tunneling parameters depends on artificial judgment and experience, lacking scientific and objective evaluation standards and system. SUMMARY

[0004] The present application aims to at least solve one of the technical problems in the related art to some extent.

[0005] The present application proposes a TBM tunneling parameter intelligent analysis and recommendation method based on a multi-parameter comprehensive analysis framework of tunneling parameter package algorithm, which can automatically calculate the reasonable distribution interval of a single parameter, and judge the stratum surrounding rock grade probability through cutterhead torque-propulsion rate segmented clustering+classification calculation, realizing comprehensive understanding and effective use of tunneling parameters.

[0006] Another object of the present application is to propose a TBM tunneling parameter intelligent analysis and recommendation system.

[0007] To achieve the above purpose, the present application proposes a TBM tunneling parameter intelligent analysis and recommendation method, which comprises:

[0008] Obtain TBM tunneling data and preprocess; wherein the tunneling data includes total thrust, cutterhead speed, cutterhead torque, cutterhead power, and propulsion speed;

[0009] Sort the key tunneling parameters after preprocessing, and calculate the slope of adjacent parameters in the key tunneling parameters to determine the reasonable distribution interval of the key tunneling parameters;

[0010] The kmeans algorithm is used for segmented clustering of the cutter head torque and the advancing speed, classification calculation is performed according to the clustering result, and the stratum surrounding rock grade probability is judged;

[0011] Based on the preset working condition mode, segmented linear fitting is performed on the current tunneling parameters, the envelope line of the segmented linear fitting is calculated, and ideal torque, thrust and the corresponding advancing speed are recommended;

[0012] A multi-dimensional fusion algorithm is executed to score the recommended or actual tunneling parameters, and the mode weight is dynamically adjusted to optimize the recommendation model.

[0013] The TBM tunneling parameter intelligent analysis and recommendation method of the embodiment of the application can also have the following additional technical features:

[0014] In an embodiment of the application, for the current batch of data, according to the speed section where it is located, the following calculations are performed:

[0015] First, segmented linear fitting is performed on each cluster of torque and speed;

[0016] The initial torque value is determined to be less than torque Vmax*0.9; the initial torque value T1;

[0017] The envelope line of the thrust-torque is calculated, and the envelope line density threshold is set;

[0018] The threshold value indicates the probability of being boxed into the envelope line;

[0019] The torque at the top of the envelope line is calculated; the initial torque value T2;

[0020] The minimum value of T1 and T2 is calculated as the starting torque value T0;

[0021] The T0 starting point is taken along the fitting line of the torque and speed corresponding class, if the torque decreases, the speed remains unchanged or increases, then the final torque value peak torque is found, and the corresponding advancing speed is peak rate;

[0022] If the peak torque is less than T2, the density threshold is increased, the envelope line and the top torque value T2' are recalculated, but the new top torque value is approximately equal to the peak torque, and the iteration is stopped;

[0023] Based on the peak torque value, the peak thrust is obtained, the torque in the range of 0.99-1 is taken, and the range of the thrust is calculated;

[0024] If the envelope line calculation has a higher left side, 0.8 times the highest point of the convex, then the left side subpeak torque, sub peak thrust is calculated at the same time, based on the sub peak torque in the piecewise linear fitting to calculate the subpeak rate;

[0025] Get the recommended parameter table, that is, the recommended tunneling parameter table of each cluster, which includes aggressive and conservative data;

[0026] According to a set of data in the current tunneling process, the classification condition is calculated, the parameter values in all classes are obtained by table lookup, and the weighted average is used to calculate the recommended peak parameter.

[0027] In an embodiment of the application, a multi-dimensional fusion algorithm is executed to score the recommended or actual tunneling parameters, and the mode weight is dynamically adjusted to optimize the recommended model, including:

[0028] Classify a new batch of data, classify based on the cluster of stratum analysis of the ring / batch data, and obtain the probability in each class;

[0029] Look up the recommended parameter table, score each set of data, and score if the condition is met, otherwise deduct the score in proportion;

[0030] Torque < torque Vmax*0.9;

[0031] Torque < peak torque;

[0032] Thrust < thrust Vmax;

[0033] Torque + thrust point is within the envelope line range;

[0034] The advance speed is close to the peak rate, with a deviation of ±5% or less;

[0035] Sum and average the scores of all points to get the final score.

[0036] In an embodiment of the application, the advance speed is clustered to obtain a clustering calculation result, including:

[0037] The cutter torque on the x-axis and the advance speed on the y-axis on the coordinate axis are segmented to obtain a plurality of groups, segmented by the x-axis, clustered by the y-axis, and the number of clusters is 2N*3+2, wherein N is the number of classes of surrounding rock;

[0038] The advance speed of each group is kmeans clustered, and the number of clusters of each group is consistent;

[0039] The data participating in the calculation is subjected to a limitation condition: the advancing speed is less than 1.5 times of the advancing speed Vmax, the cutter head torque is less than 1.5 times of the cutter head torque Vmax, and the advancing force is greater than 0.

[0040] In an embodiment of the present application, the advancing speed of each group is subjected to kmeans clustering, the number of clusters of each group is consistent, and the method comprises the following steps:

[0041] Randomly selecting k samples as initial centroids {μ1, μ2,..., μ k};

[0042] For each sample x∈X, the distance of x to all centroids is calculated: d(x, μ i ) = √Σ(x j - μ ij ) 2 x is assigned to the nearest cluster: c(x) = argmin i ||x-μ i || 2 ;

[0043] For each cluster C i : the centroid is recalculated: μ i = (1 / |C i |) Σx∈C i x;

[0044] If the centroid change is less than a threshold value ε or the maximum number of iterations is reached, the process is stopped.

[0045] The data of each segment is classified, and the label size of each segment is required to be consistent in order.

[0046] The clustering centroids of each group are sorted in ascending order, the smallest cluster number is 0, and the largest number is 2N*3+1, to obtain the ordered clustering of all data.

[0047] In an embodiment of the present application, the single batch of data is classified, and Vmax is calculated, comprising the following steps:

[0048] All data are sorted in a column to obtain a sorting result.

[0049] The sorted data is divided into 1000 packages, and the data amount in each package is a package length step.

[0050] The minimum value P_min and the maximum value P_max in each package are taken, and the slope P_k of each package is calculated as (P_max-P_min) / step.

[0051] Take the median value of all data in the package, and calculate the weighted average of the selected package slope, denoted as K500;

[0052] Multiply K500 by a predetermined multiple to obtain the threshold value K_threshold for calculating Vmax;

[0053] From the maximum point of the sorting result to the median point of the sorting, the window size is step, and the sliding step is 1. The slope between the maximum and minimum values in the window is calculated. If the slope is greater than the threshold value K_threshold, it is recorded as Vmax'.

[0054] K_threshold, until the smallest value, denoted as Vmax.

[0055] According to the clustering calculation result, the category of each batch of data in the segmented clustering is calculated, and the distribution of the batch of data in each category is calculated to obtain the clustering proportion distribution data, including:

[0056] Get a batch of new data, and search for the first belonging segment of each group of cutter speed;

[0057] Calculate the second belonging segment of the cutter torque of each group of data in the first belonging segment;

[0058] Kmeans clustering is performed on the second belonging segment, and the cluster number where the propelling speed is located is calculated;

[0059] Calculate the cluster number to which all data belongs, and calculate the number of data in each cluster number based on the cluster number to which all data belongs, to calculate the clustering proportion distribution data.

[0060] In an embodiment of the present application, the normalized Gaussian distribution is used to calculate the surrounding rock grade probability according to the clustering proportion distribution data, including:

[0061] Based on the clustering proportion distribution data, the weighted average method is used to calculate the center of data distribution to represent the position of the concentration trend of data distribution on the index;

[0062] Map each surrounding rock grade to an ideal position in the data space to obtain an ideal position for each surrounding rock grade;

[0063] Find the two grade intervals where the center is located, and perform linear interpolation to obtain a continuous magnitude value;

[0064] Assuming that the ideal position of each surrounding rock grade is a Gaussian distribution center, the Gaussian kernel function is used to calculate the similarity between the center and the ideal position of each surrounding rock grade;

[0065] All similarities are normalized to obtain the probability of each surrounding rock grade.

[0066] To achieve the above purpose, another aspect of the present application provides a TBM tunneling parameter intelligent analysis and recommendation system, characterized in that it comprises:

[0067] A tunneling data acquisition module is configured to acquire TBM tunneling data and perform preprocessing; wherein the tunneling data includes total thrust, cutterhead speed, cutterhead torque, cutterhead power, and propulsion speed.

[0068] A distribution interval calculation module is configured to sort the preprocessed key tunneling parameters and calculate the slope of adjacent parameters in the key tunneling parameters to determine the reasonable distribution interval of the key tunneling parameters.

[0069] A clustering calculation module is configured to use a kmeans algorithm to perform segmented clustering on the cutterhead torque and propulsion speed, and perform classification calculation based on the clustering results to determine the formation surrounding rock grade probability.

[0070] A segmented linear fitting module is configured to perform segmented linear fitting on the current tunneling parameters based on a preset working condition mode, calculate the envelope line of the segmented linear fitting, and recommend ideal torque, thrust, and corresponding propulsion speed.

[0071] A model optimization module is configured to perform a multi-dimensional fusion algorithm to score the recommended or actual tunneling parameters, and dynamically adjust the mode weight to optimize the recommendation model.

[0072] The TBM tunneling parameter intelligent analysis and recommendation method and system of the embodiments of the present application propose a multi-parameter comprehensive analysis framework based on tunneling parameter package algorithm, can automatically calculate the reasonable distribution interval of a single parameter, and determine the formation surrounding rock grade probability through cutterhead torque-propulsion rate segmented clustering+classification calculation, realize comprehensive understanding and effective utilization of tunneling parameters. In combination with working condition mode design, the present application develops a tunneling parameter recommendation calculation module, which can calculate ideal torque, thrust, and corresponding propulsion rate according to the current tunneling parameters, and provide scientific guidance for tunneling operation. A tunneling parameter evaluation system based on multi-dimensional fusion algorithm is proposed, which can comprehensively consider multiple tunneling parameters and geological conditions to objectively and scientifically evaluate the recommended parameters or actual tunneling parameters. An adaptive dynamic learning model of TBM tunneling parameters is constructed. The model can automatically adjust the parameter analysis and recommendation strategy based on real-time tunneling data and construction feedback. With continuous accumulation of construction data, the system can continuously optimize the parameters and logic of the multi-parameter comprehensive analysis algorithm, intelligent recommendation algorithm, and multi-dimensional fusion evaluation algorithm, realize the leap from "data-driven" to "autonomous evolution".

[0073] Additional aspects and advantages of the present application will be set forth in part in the description which follows, and in part will become apparent to those having ordinary skill in the art upon examination of the following or can be learned from practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0074] The above and / or additional aspects and advantages of the present application will become apparent and be more readily understood through reference to the following description, taken in conjunction with the accompanying drawings, wherein:

[0075] Figure 1 is a flow chart of a TBM tunneling parameter intelligent analysis and recommendation method according to an embodiment of the present application;

[0076] Figure 2 is a structural diagram of a TBM tunneling parameter intelligent analysis and recommendation system according to an embodiment of the present application. DETAILED DESCRIPTION

[0077] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict. The present application will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

[0078] In order to enable persons skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all. Based on the embodiments in the present application, all other embodiments obtained by persons skilled in the art without creative labor should fall within the protection scope of the present application.

[0079] The TBM tunneling parameter intelligent analysis and recommendation method and system according to the embodiments of the present application will be described below with reference to the accompanying drawings.

[0080] Figure 1 is a flow chart of a TBM tunneling parameter intelligent analysis and recommendation method according to an embodiment of the present application, as shown in Figure 1 , including:

[0081] S1, obtaining TBM tunneling data and performing preprocessing; wherein the tunneling data includes total thrust, cutter head speed, cutter head torque, cutter head power, and advancing speed;

[0082] Specifically, the TBM system first loads the tunneling data and removes the data during non-tunneling period (such as data with speed close to 0) to ensure the validity and accuracy of the data.

[0083] Specifically, the total thrust, cutter head speed, cutter head torque, and cutter head power are all greater than 0, and the advancing speed is >0.5m / h.

[0084] Optionally, preprocessing the operational data before conducting TBM tunneling data analysis is a key step to ensure the accuracy and effectiveness of subsequent analysis.

[0085] S2, sort the preprocessed key tunneling parameters, calculate the slope of adjacent parameters among the key tunneling parameters, and determine the reasonable distribution range of the key tunneling parameters.

[0086] S3 uses the kmeans algorithm to perform segmented clustering of cutterhead torque and feed speed, and performs classification calculations based on the clustering results to determine the probability of the surrounding rock grade.

[0087] Specifically, key tunneling parameters such as cutterhead torque, feed rate, and thrust are sorted, the slope (k value) of adjacent parameters is calculated, and a reasonable distribution range of parameters is determined based on the set threshold.

[0088] For example, the cutter head speed can be segmented in increments of 0.2. Each segment performs the following steps:

[0089] S21. Perform segmented clustering on the data of cutterhead torque (x-axis) - feed speed (y-axis). Segment by x-axis, cluster by y-axis. The number of clusters is 2N*3+2, where N is the number of rock categories. In the segmented clustering, the closer the category is to the x-axis, the higher the rock level is, and the farther the category is from the x-axis, the lower the rock level is.

[0090] S22, use the kmeans algorithm to classify the data within each x segment, requiring that the label size and sorting of each segment be consistent;

[0091] S23, imposes restrictions on the data involved in the calculation: the propulsion speed is less than 1.5 times the propulsion speed Vmax, the cutterhead torque is less than 1.5 times the cutterhead torque Vmax, and the propulsion force is greater than 0.

[0092] Specifically, in the analysis of TBM (Tunnel Boring Machine) operating data, to more precisely analyze the tunneling characteristics under different working conditions, the cutterhead speed is first segmented. Specifically, the cutterhead speed range is divided into multiple speed intervals, each segment being 0.2 times the speed of the machine, and each interval is assigned a label (such as A, B, C, etc.) for subsequent analysis. Within each speed interval, the data is further processed by performing the following steps:

[0093] The data is segmented along the x-axis according to the range of the cutter head torque. For example, if the cutter head torque range is 0 to 1000 kN·m, it can be segmented in increments of 100 kN·m.

[0094] Within each cutterhead torque segment, cluster analysis is performed on the feed speed (y-axis). The number of clusters is set to 2N×3+2, where N is the number of surrounding rock categories. For example, if the surrounding rock is divided into 3 categories, the number of clusters is 20.

[0095] In segmented clustering, the closer a category is to the x-axis (lower cutterhead torque), the higher the surrounding rock grade (i.e., the harder the surrounding rock); the farther away a category is from the x-axis (higher cutterhead torque), the lower the surrounding rock grade (i.e., the softer the surrounding rock).

[0096] The K-means algorithm is used to classify the data within each x segment: within each cutterhead torque segment, the K-means algorithm is used to perform cluster analysis on the propulsion speed data. It is required that the cluster results (labels) of each segment are ordered in the same order to facilitate subsequent analysis and comparison.

[0097] To ensure the validity and reliability of the data, the following restrictions are imposed on the data used in the calculations:

[0098] The propulsion speed must be less than 1.5 times the propulsion speed Vmax.

[0099] The cutter head torque must be less than 1.5 times the cutter head torque Vmax.

[0100] The thrust must be greater than 0 to exclude invalid data caused by equipment downtime or sensor failure.

[0101] In this embodiment of the invention, all data are segmented according to the cutterhead torque in a certain step size, resulting in several groups. Then, the propulsion speed of each group is clustered using k-means clustering, with each group having the same number of clusters. The clustering steps are as follows:

[0102] 1. Initialization: Randomly select k samples as initial centroids {μ1, μ2, ..., μ...} k}

[0103] 2. Iterate until convergence:

[0104] a. Allocation steps:

[0105] For each sample x∈X:

[0106] Calculate its distance to all centroids: d(x,μ) i )=√Σ(x j -μ ij ) 2

[0107] Assign x to the nearest cluster: c(x) = argmin i ||x-μ i || 2

[0108] b. Update steps:

[0109] For each cluster C i :

[0110] Recalculate the centroid: μ i =(1 / |C i |)Σx∈C i x

[0111] c. Check termination conditions:

[0112] Stop if the centroid change is less than the threshold ε or the maximum number of iterations is reached.

[0113] The k-means algorithm is used to classify the data in each segment, requiring that the labels of each segment be ordered in the same order.

[0114] Sort the centroids of each cluster in ascending order, with the smallest cluster numbered 0 and the largest numbered 2N*3+1, to obtain the ordered clusters of all data.

[0115] Furthermore, the new operational data from each single loop / batch will be further classified and calculated, and the data will be processed according to the Vmax requirements mentioned above. The detailed calculation steps for Vmax are as follows:

[0116] Sort all the data in one column to get the sorted result, such as the speed of advancement.

[0117] The sorted data is divided into 1000 packages, and the amount of data in each package is the package length step;

[0118] Take the minimum value P_min and the maximum value P_max in each package, and calculate the slope P_k = (P_max - P_min) / step for each package.

[0119] Take the 50 packages before and after the package containing the median value from all data, calculate the slope of the selected package using a weighted average, and denot it as K500;

[0120] Multiply K500 by a factor (e.g., 50) to obtain the threshold K_threshold for calculating Vmax;

[0121] Iterative calculations begin from the maximum value in the sorted results to the median value. The window size is step, and the sliding step size is 1. The slope between the maximum and minimum values ​​within the window is calculated. If the slope is greater than the threshold K_threshold, it is recorded as Vmax'. This process continues until the minimum value is reached, which is then recorded as Vmax.

[0122] The above steps transform complex continuous data into segmented discrete data, and cluster analysis identifies tunneling characteristics under different surrounding rock conditions. This method not only improves the accuracy of data analysis but also provides a reliable foundation for subsequent geological classification and optimization of construction parameters.

[0123] Specifically, according to the set dimensions, each ring or batch of new data can be segmented by rotational speed (e.g., A, B, C, etc.), and classified and calculated according to the above-mentioned cutterhead torque segmentation and feed speed clustering (for example, if it is segment A, then the clustering results of cutterhead torque and feed speed within segment A are selected for classification and calculation). The category of each data point in the segmented cluster is calculated. After calculating all the data, the distribution of the data in this ring / batch within each category can be obtained, that is, the proportion in each cluster. The specific steps are as follows:

[0124] S31, acquire a new batch of data, first find the segment to which the cutter head speed belongs for each group;

[0125] S32, calculate the segment to which the cutter head torque belongs within this speed segment for each set of data;

[0126] S33, find the k-means cluster of the segment, calculate the cluster number of the advancing speed, and repeat S31 and S32 to calculate the cluster number to which all data belong.

[0127] S34, based on the results of S33, can calculate the number of data in each cluster number, and thus calculate its proportion probability (normalized probability).

[0128] Specifically, first, a new set of operating data is obtained from the control system or sensors of the TBM tunneling machine. This data includes parameters such as cutterhead torque, feed speed, cutterhead rotation speed, and feed force.

[0129] Based on a preset rotational speed interval (e.g., each step is 0.2 r / min), the cutter head rotational speed range is divided into multiple speed intervals, and each interval is assigned a label (e.g., A, B, C, etc.). For each set of data, the speed interval to which the cutter head rotational speed belongs is determined.

[0130] Within each speed range, the data is segmented according to the range of cutter head torque. For each set of data, the torque segment in which the cutter head torque falls is determined.

[0131] Within each cutterhead torque segment, the K-means algorithm is used to perform cluster analysis on the feed rate data.

[0132] For each set of data, find its corresponding cluster number.

[0133] For all data, repeat the above steps to calculate the cluster number for each data group. Count the number of data points in each cluster. Calculate the percentage probability for each cluster, which is the number of data points in that cluster divided by the total number of data points. Normalize all percentage probabilities so that their sum equals 1.

[0134] Through the above steps, we can obtain the cluster number of each data set within different speed and torque segments, as well as the probability of each cluster number. These results can be used for further judgment of surrounding rock grade and stratigraphic classification, providing a scientific basis for the selection of TBM tunneling machine construction parameters and stratigraphic classification.

[0135] Furthermore, this invention constructs a graded probability classification algorithm, taking the probability values ​​of each cluster as input and outputting the probability of each grade of surrounding rock. The calculation approach is as follows: by locating the distribution trend through weighted centering, and combining it with the ideal grade location for linear interpolation and Gaussian modeling, the probability of belonging to each discrete grade is output. The steps are as follows:

[0136] S41, use the weighted average method to calculate the "centroid" of the data distribution: center is the position of the central tendency of the entire distribution on the index (which may be a floating-point number).

[0137] First, based on each cluster number and its corresponding probability value, the "center" of the data distribution is calculated using a weighted average method, with the following formula:

[0138]

[0139] Among them, c i It is the i-th cluster number, p i It is the probability value of the i-th cluster.

[0140] "center" represents the position of the central tendency of the entire distribution on the index, and may be a floating-point number.

[0141] S42 maps each level to an "ideal location" in the data space.

[0142] Assume that each surrounding rock grade corresponds to an ideal location, which can be determined based on historical data or prior knowledge.

[0143] For example, if the surrounding rock is divided into 3 grades, the ideal locations could be μ1, μ2, and μ3.

[0144] S43, find the two level intervals where the center is located, perform linear interpolation, and obtain a continuous "magnitude" value.

[0145] Determine the two level intervals where "center" is located, for example, μi and μi+1.

[0146] Calculate a continuous "magnitude" value using a linear interpolation formula:

[0147]

[0148] S44, assuming the ideal position for each level is the center of a Gaussian distribution. Calculate the similarity from the center to each level_position (using a Gaussian kernel).

[0149] Assuming the ideal location for each rock grade is a Gaussian distribution center, the similarity from "center" to the ideal location for each grade is calculated using a Gaussian kernel function:

[0150]

[0151] Where σ is the bandwidth parameter of the Gaussian kernel.

[0152] S45 normalizes the probabilities of all levels so that the sum is 1.

[0153] Specifically, for all similarity values ​​S i Normalize the sum to make it equal to 1:

[0154]

[0155] Where N is the total number of surrounding rock grades.

[0156] S4, based on the preset working condition mode, performs piecewise linear fitting on the current tunneling parameters, calculates the envelope of the piecewise linear fitting, and recommends ideal torque, thrust and its corresponding propulsion speed.

[0157] Specifically, for the current batch of data, the following calculations are performed based on the speed range in which it falls:

[0158] First, perform piecewise linear fitting on each cluster of torque and speed, for example, divide it into 6 segments;

[0159] The initial torque value should be less than torque Vmax * 0.9 (initial torque value T1);

[0160] Calculate the envelope of thrust-torque, set a threshold for envelope density; the larger the threshold, the more conservative the approach.

[0161] The threshold indicates the probability of the data points being within the envelope. For example, a threshold of 0.1 indicates that 90% of the data points are within the envelope, and a threshold of 0.02 indicates that 98% of the data points are within the envelope.

[0162] Calculate the torque at the top of the envelope (initial torque value T2);

[0163] Calculate the minimum value of T1 and T2 as the initial torque value T0;

[0164] Starting from T0, along the fitted line of the corresponding class of torque and rate, if the torque decreases while the rate remains unchanged or increases, the final torque value (peak torque) is found, and the corresponding propulsion rate is the peak rate.

[0165] If the peak torque is less than T2, increase the density threshold, recalculate the envelope and vertex torque value T2', but the new vertex torque value is approximately equal to the peak torque, and stop the calculation iteration;

[0166] The peak thrust can be obtained from the peak torque value. For example, by taking a torque in the range of 0.99 to 1, the range of thrust can be calculated.

[0167] If the envelope calculation shows a bulge on the left side that is higher, for example, 0.8 times the highest point, then the sub-peak torque and sub-peak thrust on the left side are calculated simultaneously. Based on the sub-peak torque, the sub-peak rate can be calculated in the piecewise linear fitting.

[0168] The left sub-vertex illustrates the parameters of the strata under collapse conditions;

[0169] The recommended parameter table is obtained, which is the suggested mining parameter table for each cluster, including aggressive (low density threshold) and conservative (high density threshold) data.

[0170] Based on a set of data from the current tunneling process, the classification is calculated, the parameter values ​​for each class are obtained from the table above, and the recommended peak parameter is calculated using a weighted average (sumproduct).

[0171] S5 executes a multi-dimensional fusion algorithm to score the recommended or actual tunneling parameters and dynamically adjusts the pattern weights to optimize the recommendation model.

[0172] Specifically, a multi-dimensional fusion algorithm is used to score the recommended parameters or actual tunneling parameters, comprehensively considering multiple factors such as torque, thrust, and advance rate, as well as constraints within the envelope range. Based on the scores, the mode weights are dynamically adjusted to optimize the tunneling parameter recommendation model.

[0173] (1) Classify the newly acquired batch of data, and classify them based on the clustering of ring / batch data stratigraphic analysis to obtain the probability (or proportion) in each category;

[0174] (2) Check the recommended parameter table, score each set of data, and give points to those that meet the conditions; otherwise, deduct points proportionally.

[0175] (3) Torque should be less than torque Vmax * 0.9;

[0176] (4) Torque should <peak torque;

[0177] (5) Thrust < Thrust Vmax;

[0178] (6) The torque + thrust point should be within the envelope range;

[0179] (7) The propulsion rate should be close to the peak rate, for example, within ±5% of the deviation;

[0180] (8) Sum and average the scores of all points to get the final score. The higher the score, the better.

[0181] This invention employs a multi-dimensional fusion algorithm to comprehensively score recommended or actual operating parameters during shield or TBM tunneling, aiming to achieve intelligent and adaptive control of the tunneling process. The scoring system comprehensively considers multiple key parameter dimensions, including dynamic and efficiency indicators such as cutterhead torque, thrust, and advance rate, while also incorporating stability factors such as vibration and tilt angle, as well as safety indicators such as ground pressure and cutter wear, constructing a comprehensive evaluation model reflecting the tunneling status. During the scoring process, the system introduces an envelope mechanism, setting safe threshold ranges for each parameter and implementing dynamic penalties when parameters exceed these limits to ensure equipment operation within a safe and controllable range. Based on real-time scoring results, the system can dynamically adjust the weight allocation of different parameter dimensions, thereby continuously optimizing the tunneling parameter recommendation model. This allows the model to automatically adapt and evolve according to external conditions such as geological changes and equipment status, improving tunneling efficiency and construction safety. This mechanism not only enhances the intelligent decision-making capability of the tunneling system but also provides strong support for efficient and stable tunneling under complex geological conditions.

[0182] The TBM tunneling parameter intelligent analysis and recommendation method according to embodiments of the present invention can accurately predict and optimize tunneling parameters through intelligent analysis and recommendation algorithms, thereby improving tunneling efficiency and quality. It reduces tunneling failures and downtime caused by improper parameter settings, lowering operating and maintenance costs. Furthermore, by monitoring and evaluating tunneling parameters in real time, it can promptly identify potential safety hazards, ensuring construction safety.

[0183] To achieve the above embodiments, such as Figure 2 As shown, this embodiment also provides a TBM tunneling parameter intelligent analysis and recommendation system 10, including:

[0184] The tunneling data acquisition module 100 is used to acquire TBM tunneling data and perform preprocessing; wherein, the tunneling data includes total thrust, cutterhead rotation speed, cutterhead torque, cutterhead power, and propulsion speed;

[0185] The distribution interval calculation module 200 is used to sort the preprocessed key tunneling parameters, calculate the slope of adjacent parameters among the key tunneling parameters, and determine the reasonable distribution interval of the key tunneling parameters.

[0186] The clustering calculation module 300 is used to perform segmented clustering of cutterhead torque and feed speed using the kmeans algorithm, and to perform classification calculations based on the clustering results to determine the probability of the surrounding rock grade of the strata.

[0187] The piecewise linear fitting module 400 is used to perform piecewise linear fitting on the current tunneling parameters based on a preset working condition mode, calculate the envelope of the piecewise linear fitting, and recommend ideal torque, thrust and corresponding propulsion speed.

[0188] Model optimization module 500 is used to execute multi-dimensional fusion algorithms to score recommendation or actual mining parameters and dynamically adjust pattern weights to optimize the recommendation model.

[0189] The TBM tunneling parameter intelligent analysis and recommendation system according to embodiments of the present invention can accurately predict and optimize tunneling parameters through intelligent analysis and recommendation algorithms, thereby improving tunneling efficiency and quality. It reduces tunneling failures and downtime caused by improper parameter settings, lowering operating and maintenance costs. Furthermore, by monitoring and evaluating tunneling parameters in real time, it can promptly identify potential safety hazards, ensuring construction safety.

[0190] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0191] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A method for intelligent analysis and recommendation of TBM tunneling parameters, characterized in that, include: Acquire TBM tunneling data and perform preprocessing; wherein, the tunneling data includes total thrust, cutterhead rotation speed, cutterhead torque, cutterhead power, and propulsion speed; The preprocessed key tunneling parameters are sorted, and the slopes of adjacent parameters are calculated to determine the reasonable distribution range of the key tunneling parameters. The kmeans algorithm is used to perform segmented clustering of cutterhead torque and feed speed, and classification calculations are performed based on the clustering results to determine the probability of the surrounding rock grade of the strata. Based on the preset working condition mode, piecewise linear fitting is performed on the current tunneling parameters, the envelope of the piecewise linear fitting is calculated, and ideal torque, thrust and corresponding propulsion speed are recommended. A multi-dimensional fusion algorithm is used to score the recommended or actual tunneling parameters, and the pattern weights are dynamically adjusted to optimize the recommendation model.

2. The method according to claim 1, characterized in that, For the current batch of data, based on the corresponding speed range, the following calculations are performed: First, piecewise linear fitting is performed on each cluster of torque and speed; The initial torque value should be less than torque Vmax * 0.9; initial torque value T1; Calculate the thrust-torque envelope and set an envelope density threshold; The threshold indicates the probability of the bounding box being included; Calculate the torque at the top of the envelope; initial torque value T2; Calculate the minimum value of T1 and T2 as the initial torque value T0; Starting from T0, along the fitted line of the corresponding class of torque and rate, if the torque decreases while the rate remains unchanged or increases, the final torque value, peak torque, is found, and the corresponding propulsion speed is peak rate. If the peak torque is less than T2, increase the density threshold, recalculate the envelope and vertex torque value T2', but the new vertex torque value is approximately equal to the peak torque, and stop the calculation iteration; Peak thrust is obtained based on peak torque value. The range of thrust is calculated by taking torque in the range of 0.99 to 1. If the envelope calculation shows a higher bulge on the left side, 0.8 times the highest point, then the sub-peak torque and sub-peak thrust on the left side are calculated simultaneously, and the sub-peak rate is calculated based on the sub-peak torque in the piecewise linear fitting. The recommended parameter table is obtained, which is the suggested mining parameter table for each cluster, including both aggressive and conservative data; Based on a set of data from the current tunneling process, the classification is calculated, the parameter values ​​for all classes are obtained by looking up a table, and a weighted average is used to calculate the recommended peak parameter.

3. The method according to claim 2, characterized in that, A multi-dimensional fusion algorithm is executed to score the recommended or actual tunneling parameters, and the pattern weights are dynamically adjusted to optimize the recommendation model, including: A new batch of data is collected and classified, and the classification is performed based on the clustering of ring / batch data stratigraphic analysis to obtain the probability of each category; Check the recommended parameter table, score each set of data, award points if the conditions are met, and deduct points proportionally otherwise; Torque < Torque Vmax * 0.9; Torque <peak torque; Thrust < Thrust Vmax; The torque and thrust points are within the envelope range; The propulsion speed is close to the peak rate, with a deviation within ±5%. Sum and average the scores of all points to obtain the final score.

4. The method according to claim 1, characterized in that, Clustering the propulsion speeds yields the following clustering results: The cutterhead torque on the x-axis and the feed speed on the y-axis are segmented to obtain several groups. The segments are divided according to the x-axis and clustered according to the y-axis. The number of clusters is 2N*3+2, where N is the number of surrounding rock categories. Perform k-means clustering on the advancement speed of each group, with the same number of clusters in each group; Constraints are imposed on the data used in the calculation: the propulsion speed is less than 1.5 times the propulsion speed Vmax, the cutterhead torque is less than 1.5 times the cutterhead torque Vmax, and the propulsion force is greater than 0.

5. The method according to claim 1, characterized in that, For each group's propulsion speed, perform k-means clustering, ensuring the same number of clusters per group, including: Randomly select k samples as the initial centroids {μ1,μ2,...,μ k }; For each sample x∈X, calculate its distance to all centroids: d(x,μ) i )=√Σ(x j -μ ij ) 2 Assign x to the nearest cluster: c(x) = argmin i ||x-μ i || 2 ; For each cluster C i Recalculate the centroid: μ i =(1 / |C i |)Σx∈C i x; If the change in centroid is less than the threshold ε or the maximum number of iterations is reached, then stop. Classify the data in each segment, ensuring that the label size and order are consistent for each segment; Sort the centroids of each cluster in ascending order, with the smallest cluster numbered 0 and the largest numbered 2N*3+1, to obtain the ordered clusters of all data.

6. The method according to claim 1, characterized in that, Classify the data in a single batch and calculate Vmax, including: Sort all the data in one column to get the sorted result; The sorted data is divided into 1000 packages, and the amount of data in each package is the package length step; Take the minimum value P_min and the maximum value P_max in each package, and calculate the slope P_k = (P_max - P_min) / step for each package; Take the 50 packages before and after the package containing the median value from all data, calculate the slope of the selected package using a weighted average, and denot it as K500; Multiply K500 by a preset factor to obtain the threshold K_threshold for calculating Vmax; Iterative calculations begin from the maximum value in the sorted results to the median value. The window size is step, and the sliding step size is 1. The slope between the maximum and minimum values ​​within the window is calculated. If the slope is greater than the threshold K_threshold, it is recorded as Vmax'. This process continues until the minimum value is reached, which is then recorded as Vmax. Based on the clustering calculation results, the category of each batch of data in the segmented clustering is calculated, and the distribution of all data in each category is calculated to obtain the clustering proportion distribution data, including: Acquire a new batch of data and search for the first segment of each group of cutterhead rotation speeds; Calculate the second segment where the cutter head torque is located within the first segment for each set of data; Perform k-means clustering on the second segment and calculate the cluster number where the advancement speed belongs; Calculate the cluster number to which all data belong, and calculate the number of data in each cluster number based on the cluster number to which all data belong, so as to calculate the cluster proportion distribution data.

7. The method according to claim 6, characterized in that, The probability of surrounding rock grade is calculated based on the cluster proportion distribution data using a normalized Gaussian distribution, including: The center of gravity of the data distribution is calculated using a weighted average method based on the cluster distribution data, in order to represent the position of the central tendency of the data distribution on the index. The ideal position for each rock mass grade is obtained by mapping each rock mass grade to an ideal position in the data space; Find the two level intervals where the center is located, and perform linear interpolation to obtain a continuous magnitude value; Assuming the ideal location for each rock grade is a Gaussian distribution center, the similarity between the center and the ideal location for each rock grade is calculated using a Gaussian kernel function; Normalize all similarities to obtain the probability of each surrounding rock grade.

8. A TBM tunneling parameter intelligent analysis and recommendation system, characterized in that, include: The tunneling data acquisition module is used to acquire TBM tunneling data and perform preprocessing; wherein, the tunneling data includes total thrust, cutterhead rotation speed, cutterhead torque, cutterhead power, and propulsion speed; The distribution interval calculation module is used to sort the preprocessed key tunneling parameters, calculate the slope of adjacent parameters among the key tunneling parameters, and determine the reasonable distribution interval of the key tunneling parameters. The clustering calculation module is used to perform segmented clustering of cutterhead torque and feed speed using the kmeans algorithm, and to perform classification calculations based on the clustering results to determine the probability of the surrounding rock grade of the strata. The piecewise linear fitting module is used to perform piecewise linear fitting on the current tunneling parameters based on a preset working condition mode, calculate the envelope of the piecewise linear fitting, and recommend ideal torque, thrust and corresponding propulsion speed. The model optimization module is used to execute multi-dimensional fusion algorithms to score the recommended or actual tunneling parameters and dynamically adjust the pattern weights to optimize the recommendation model.

Images