Wear analysis method and system for shield machine or TBM cutter with correlation characteristics

Abstract

The present invention discloses a method and system for analyzing the wear of a shield machine or TBM cutter using associated features, belonging to the technical field of associated feature analysis. The method comprises the following steps: collecting parameter data related to cutter wear in real time, and constructing a parameter time series sequence; calculating statistical features, derivative features and change curves of rock and soil layers, and updating the parameter time series sequence; analyzing the dynamic changes of the parameter time series sequence, calculating feature importance scores using a random forest algorithm to avoid interference from unimportant information during the analysis process, and constructing an associated feature model; constructing three-dimensional point cloud data of the rock and soil layer, generating fused point cloud data containing cutter wear labels, providing rich dimensional data, and constructing a point cloud model for determining the cutter status, reducing the risk of misjudgment and missed judgment, and determining the cutter status; once it is determined that the cutter is damaged, generating a cutter damage signal, thereby providing a strong guarantee for the safe and efficient operation of the shield machine or TBM.

Description

Technical Field

[0001] The present invention belongs to the technical field of correlation feature analysis, and relates to a method and system for correlation feature analysis of wear of a shield machine or a TBM cutter. Background Art

[0002] The shield machine, or TBM, is the core equipment in underground tunnel construction. It integrates excavation, support, excavation, and guidance functions, greatly improving the efficiency and safety of tunnel construction. Among them, the cutter, as a key component of the shield machine or TBM, bears the heavy responsibility of crushing rock and soil. Its performance directly affects the shield machine or TBM's excavation speed and the quality of the tunnel.

[0003] The existing Chinese patent with authorization announcement number CN110119551B discloses a shield machine tool wear degradation correlation feature analysis method based on machine learning, which is used to solve the technical problem of narrow application scope in the existing technology, and at the same time improve the accuracy of tool health status monitoring and life prediction. The implementation steps include: obtaining the original feature data set data3 under the slightly weathered stratum; obtaining the data resource set Data; constructing a LightGBM feature sorting model; obtaining the important feature subset Fea; discretizing the important feature subset Fea; performing association rule mining on the discretized important feature subset data_lsh; and obtaining the shield machine tool wear degradation correlation features.

[0004] Although the existing technology fully considers the impact of all collected data on shield machine tool wear, has a wide range of applications and high accuracy, it does not consider the real-time and dynamic nature of the characteristic data. In particular, during the actual operation of the shield machine, due to the continuous changes in working conditions and external environment, such as changes in formation conditions, adjustments to construction processes, and changes in the equipment's own state, such as the impact of the tool on other parameters at different wear stages, these factors may change over time and interfere with the judgment of tool wear. Therefore, this application provides a shield machine or TBM cutter correlation feature analysis wear method and system to address the problem of the lack of real-time and dynamic nature of characteristic data and improve the accuracy and real-time nature of wear analysis. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the purpose of the present invention is to provide a method and system for analyzing the wear of shield machines or TBM cutters based on the correlation characteristics. By constructing a parameter time series sequence and updating the sequence content in combination with the rock and soil layer change curve, the real-time performance of the data is improved; and the parameter time series sequence is dynamically analyzed to capture the fluctuations between sequences, thereby enhancing the ability to respond to complex environments.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] Shield machine or TBM cutter wear analysis method based on correlation characteristics, including:

[0008] Step S1: collecting parameter data related to tool wear in real time, performing preprocessing, and constructing a parameter time series;

[0009] Step S2: Calculate the statistical characteristics, derived characteristics and the change curve of the rock and soil layer, and update the parameter time series;

[0010] Step S3: Analyze the dynamic changes of the parameter time series, calculate the feature importance score using the random forest algorithm, and build a correlation feature model;

[0011] Step S4: constructing three-dimensional point cloud data of the rock and soil layer, generating fused point cloud data containing tool wear labels, and constructing a point cloud model for determining the tool status;

[0012] Step S5: Calculate the predicted probability of the tool state and determine the tool state; once it is determined that the tool is damaged, generate a tool damage signal.

[0013] S2.1: Based on the real-time collected cutter head torque and propulsion speed , calculate the ratio of cutter head torque to propulsion speed ;

[0014] S2.2: Based on the real-time collected thrust , calculate the product of the thrust and the cutter head torque ;

[0015] S2.3: Continuously record the actual values of geotechnical parameters at different time points and draw the geotechnical parameter curves changing with time by data fitting;

[0016] S2.4: Save the calculated statistical characteristics, derived characteristics and change curves of the rock and soil layers to the parameter time series In which, the parameter time series sequence includes the historical time series sequence and real-time timing sequences .

[0017] Specifically, the specific steps of data fitting in S2.3 include:

[0018] During the excavation process, a fitting curve is generated based on the nonlinear relationship between a certain geotechnical parameter and the excavation distance;

[0019] Constructing a measurement sample group , calculate the sum of squares of the errors ;

[0020] respectively 、 and Find the partial derivatives and set them to zero to obtain the partial derivative equations;

[0021] Solve the partial derivative equations and calculate the fitting coefficient group The final value of .

[0022] Specifically, the specific steps of solving the partial derivative equations include:

[0023] Assume that the functions in the partial derivative equations are partial derivative functions 、 、 , and use the curved trapezoid area approximation to construct a variable upper limit integral function for each partial derivative function;

[0024] Calculate the gradient of each variable upper limit integral function with respect to each fitting coefficient respectively;

[0025] Assume the weight coefficient is , reconstruct and assign the variable upper limit integral function to generate a newly constructed assignment function;

[0026] Combine multiple assignment functions into a function vector and calculate the Jacobian matrix of the function vector ;

[0027] set up 、 、 The initial value is , and use the Newton iteration method to calculate the current fitting coefficient value, function vector and Jacobian matrix. 、 、 Make updates;

[0028] Calculate the iterative error of the fitted coefficient set , set the iteration termination threshold to , to determine whether the iterative operation is completed; when When , it is determined to be non-convergent and the iterative operation continues; when When , it is determined that the iteration has converged and the fitting coefficient value of the latest iteration is output.

[0029] Specifically, the specific steps of step S3 include:

[0030] S3.1: For the parameter timing sequence , using STL decomposition Decompose into trend term, seasonal term and residual term;

[0031] S3.2: Use the trend items in the historical time series to train the trend model and output the predicted value of the trend item;

[0032] S3.3: Using the residual items in the historical time series to train the residual model, fitting the dynamic characteristics of the residual items, and generating predicted values of the residual items;

[0033] S3.4: Combine the trend item prediction value, the seasonal item in the historical time series, and the residual item prediction value to obtain a historical prediction value.

[0034] Specifically, the specific steps of step S3 also include:

[0035] S3.5: Calculate the linear correlation between any two variables in the historical forecast values and construct a correlation matrix;

[0036] S3.6: Set the actual data feature vector to contain characteristic parameters, generating dimensional random vector

[0037] S3.7: Set the actual data feature vector to , the weight vector is , fuse the random vector with the actual data feature vector to generate a fused feature vector .

[0038] Specifically, the specific steps of step S3 also include:

[0039] S3.8: Use the random forest algorithm to construct multiple decision trees, calculate the feature importance scores, and generate the fusion feature vector Importance sequence ;

[0040] S3.9: Taking tool wear as the target variable, the correlation function between various influencing factors and tool wear is established through feature importance score and fusion feature vector to generate the correlation feature model.

[0041] Specifically, the specific steps of step S4 include:

[0042] S4.1: Use geological radar and laser scanning equipment to scan the rock and soil layers in front of and around the shield machine or TBM, perform data verification, and generate three-dimensional point cloud data of the rock and soil layers;

[0043] S4.2: Fusing the three-dimensional point cloud data with features in the associated feature model to generate fused point cloud data with a tool wear label;

[0044] S4.3: Preprocessing the fused point cloud data, including gridding and normalization.

[0045] S4.4: Build a point cloud model, including an input layer, a multi-layer perceptron layer, a max pooling layer, and a fully connected layer, where the output is the predicted probability of the tool state;

[0046] S4.5: Randomly initialize the parameters in the point cloud model, and train the point cloud model based on the processed fused point cloud data.

[0047] Specifically, the specific steps of data verification in S4.1 include:

[0048] Set the initial search area with the data point to be verified as the center, obtain the number of surrounding points in the search area, and calculate the initial point distribution density ;

[0049] Set the search density range to , and judging whether the point distribution density is within the search density range;

[0050] like , the point distribution density is within the search density range;

[0051] like , expand the search range until the point distribution density is within the search density range;

[0052] like , reduce the search range until the point distribution density is within the search density range;

[0053] Get the location coordinates of multiple sampling points and generate a sample data set , and calculate the mean of the sample data ;

[0054] Calculate sample data With the average The difference between , and calculate the average of the differences ;

[0055] Set the discrete threshold of the search area to , and determine whether there is an error in the data point; if , there is no error in the data point, and the data point is defined as a calibrated point; , the data point has an error, defining the data point as a preliminary error point;

[0056] In the search area of the preliminary error point, the data of the surrounding points is verified to obtain the number of errors in the surrounding points. and the average , and calculate the average value of the surrounding error ;in, For the The average value of the errors;

[0057] Set the geological change difference threshold to , determine whether the initial error point is a boundary point; if , the preliminary error point is the boundary point; if , generate alternative data points without errors;

[0058] Generate 3D point cloud data of geotechnical layers, including calibrated points and boundary points.

[0059] Specifically, the specific steps of step S5 include:

[0060] S5.1: Obtaining real-time timing sequences , calculate the predicted probability of tool status ;

[0061] S5.2: Set the damage threshold of the tool to , judge the damage of the current tool; if , the tool is in an undamaged state; if , the tool is damaged, generating a tool damaged signal;

[0062] S5.3: After the tool change is completed, various parameter data of the new tool are collected, and the parameter timing sequence and timestamp information are updated;

[0063] S5.4: Utilize the updated parameter timing sequence to optimize and adjust the associated feature model and point cloud model, and perform damage analysis on the new tool.

[0064] The shield machine or TBM cutter wear correlation feature analysis system includes: data acquisition module, feature analysis module, prediction module and update module;

[0065] The data acquisition module uses various sensors to collect parameter data related to tool wear in real time;

[0066] The characteristic analysis module calculates statistical parameters and derived parameters based on the initial acquisition parameters, and constructs a change curve of the rock and soil layer in combination with the tunneling conditions of the shield machine or TBM to generate a parameter time series;

[0067] The prediction module includes an association unit and a judgment unit;

[0068] The association unit is used to dynamically analyze the parameter time series, calculate the importance score of each feature, and build an association feature model;

[0069] The judgment unit is used to construct three-dimensional point cloud data of the rock and soil layer, combine it with the associated feature model, generate fused point cloud data with tool wear labels, and build a point cloud model for judging the tool status; output the predicted probability of the tool status, and judge the tool status;

[0070] The updating module is used to collect parameter data of the new tool after the tool change is completed, update the parameter timing sequence, and optimize and adjust the associated feature model and the point cloud model using the new data.

[0071] Beneficial effects of the present invention:

[0072] By calculating statistical and derivative features, information closely related to tool wear is mined from different dimensions, and a geotechnical change curve is constructed. This further refines the understanding of the dynamic external environment, continuously updates the parameter time series, ensures data consistency and timeliness, and avoids analytical deviations caused by data loss. Analyzing the dynamic changes in the parameter time series helps capture subtle fluctuations and trend changes during the operation of the shield machine or TBM. The fusion of the three-dimensional point cloud data of the geotechnical layer with the constructed associated feature model generates fused point cloud data containing tool wear labels, which improves the intuitiveness and accuracy of tool status judgment and enhances the ability to cope with complex working conditions. When the model determines that the tool is damaged, not only can the tool changing robot be activated in a timely manner, but the parameter data after the tool change will also be used to optimize the model, allowing the model to continuously adapt to the ever-changing operating status and environment of the shield machine or TBM, always maintaining a high level of judgment ability, and providing strong guarantees for the safe and efficient operation of the shield machine or TBM. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 Schematic diagram of the wear analysis method for shield machine or TBM cutter correlation characteristics;

[0074] Figure 2 A flow chart of the time sequence for updating parameters in the wear analysis method for shield machine or TBM cutter associated features;

[0075] Figure 3 Construct a flow chart for the associated feature model in the wear analysis method of shield machine or TBM cutter;

[0076] Figure 4 A flow chart for determining tool damage in the wear analysis method for shield machine or TBM cutter associated characteristics;

[0077] Figure 5 Analyze the wear system structure diagram for the shield machine or TBM cutter with related characteristics. DETAILED DESCRIPTION

[0078] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations on the technical solution of the present invention. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0079] Example 1

[0080] refer to Figures 1 to 4 As shown, this embodiment introduces a method for analyzing wear of a shield machine or TBM cutter by correlation characteristics, including the following steps:

[0081] Step S1: During the operation of the shield machine or TBM, various sensors are used to collect parameter data related to tool wear in real time, including equipment operation parameters such as propulsion speed, cutter head torque, cutter head speed, penetration, and thrust force, external environmental parameters such as stratum soil type, strength parameters, hardness, water content, mineral type and content, and construction process parameters such as excavation mode and grouting volume, and the collected parameter data are preprocessed, including: using the mean to fill missing values in the data, using the Z-Score method to detect and eliminate outliers, using the Gaussian filter method to reduce data noise, and constructing a parameter time series sequence to ensure the real-time data collection and timely capture the dynamic changes of the shield machine or TBM operation status and external environment; wherein, the parameter sequence includes parameters related to tool wear and the corresponding actual parameter values, and the actual parameter values are saved in chronological order, and the operation sampling frequency is , the environmental sampling frequency is , the construction sampling frequency is ;

[0082] Step S2: Based on the working principle of the shield machine or TBM and the tool wear theory, the statistical characteristics and derived characteristics related to tool wear are calculated, the change curve of the rock and soil layer is constructed, and the parameter time series is updated;

[0083] Step S3: Analyze the dynamic changes of the parameter time series, use the random forest algorithm to calculate the feature importance score, clarify the influence of various features on tool wear, build an associated feature model, and comprehensively consider the complex relationship between various features and tool wear to achieve an effective description of the tool wear state;

[0084] Step S4: constructing three-dimensional point cloud data of the rock and soil layer, combining it with the associated feature model to generate fused point cloud data containing tool wear labels, and constructing a point cloud model for determining the tool status;

[0085] Step S5: Based on the real-time time series and point cloud model, the predicted probability of the tool state is output and the tool state is judged; once the tool is judged to be damaged, a tool damage signal is generated, and the control center starts the tool changing robot; after the tool change is completed, the parameter data of the new tool is collected, the parameter time series is updated, and the associated feature model and point cloud model are optimized and adjusted using the new data to adapt to the new operating status and environmental changes of the shield machine or TBM, thereby improving the accuracy and reliability of model judgment.

[0086] Specifically, the specific steps of step S2 include:

[0087] S2.1: The tool cutting process is essentially a mechanical process. According to the relationship between force and motion, torque reflects the resistance that the tool needs to overcome when rotating, and the propulsion speed reflects the tool's forward speed in the axial direction. In order to more directly understand the tool's cutting efficiency and wear status, the ratio of the cutter head torque to the propulsion speed is introduced as an evaluation indicator. and propulsion speed , calculate the ratio of cutter head torque to propulsion speed , during the cutting process, the ratio Indicates the rotational resistance of the tool under unit propulsion speed. From the perspective of mechanics, the ratio The larger it is, the greater the resistance the tool needs to overcome during cutting, which will inevitably lead to increased friction and force between the tool and the cutting medium, thereby accelerating tool wear. The expression is as follows:

[0088] ;

[0089] In the formula, by comparing the value Continuous monitoring and analysis can timely detect changes in tool cutting status; when the ratio When it is larger, it means that at the same propulsion speed, the torque that the cutter head needs to overcome is larger, the resistance encountered by the tool during cutting is larger, the cutting process is more difficult, and the cutting efficiency is lower. At this time, the friction and force between the tool and the medium increase, and the tool wear is accelerated; on the contrary, the wear is relatively slow. For example, in hard rock formations, the ratio larger;

[0090] S2.2: In actual construction, when the shield machine or TBM is in an unstable cutting condition, such as encountering a sudden change in local geology, the thrust and cutter torque will fluctuate greatly. The thrust makes the tool closely contact with the cutting surface and provides the axial force for cutting, while the cutter torque drives the tool to rotate and cut. and cutter head torque , calculate the product of the thrust and the cutter head torque , To a certain extent, it reflects the load borne by the tool. Based on the principle of force synthesis and effect, the combined action of the two forces affects the wear of the tool. By comparing the load of the tool in different time periods, the tool The value can be used to determine whether the cutting condition of the shield machine or TBM is stable and whether the wear trend of the tool is normal; the expression is as follows:

[0091] ;

[0092] Where, the thrust The larger the value, the greater the pressure on the tool in the axial direction, and the cutter torque The larger the value, the greater the resistance torque the tool encounters during the rotational cutting process; When it is larger, it means that the shield machine or TBM consumes more energy during the cutting process, the tool bears a greater load during operation, and the wear will be correspondingly aggravated; on the contrary, the wear is relatively light;

[0093] S2.3: As the shield machine or TBM advances, continuously record the actual values of geotechnical parameters at different time points. These parameters include, but are not limited to, particle size distribution, void ratio, rock strength, and rock hardness. To better track the dynamic changes of geotechnical parameters during the tunneling process, plot a time-varying curve of the geotechnical layer, with the tunneling distance as the horizontal axis and the geotechnical parameters as the vertical axis. This clearly demonstrates the changing trends of the geotechnical parameters as the tunneling progresses. Record the relevant parameter values at regular intervals (e.g., 5 meters) and plot the curve using data fitting.

[0094] S2.4: Save the calculated statistical characteristics, derived characteristics and change curves of the rock and soil layers to the parameter time series In which, the parameter time series sequence includes the historical time series sequence and real-time timing sequences , For the parameter timing sequence The value of the moment, including The parameter data, statistical characteristics, derived characteristics, and change curve of the rock and soil layer at each moment. The derived characteristics include the ratio of cutterhead torque to propulsion speed and the product of jacking force and cutterhead torque. The statistical characteristics include the mean, variance, maximum, and minimum values of each parameter in the original parameter time series. The mean is used to measure the smoothness of the shield machine or TBM operation and reflects the average level of various parameters of the shield machine or TBM in a specific time period. The variance is used to measure the degree of dispersion of the data. The extreme value is used to measure the extreme operating state of the shield machine or TBM in a specific time period. The expression is as follows:

[0095] ;

[0096] ;

[0097] ;

[0098] ;

[0099] In the formula, the parameter time series Parameters in For example, 、 、 、 The parameters are exist The mean, variance, maximum and minimum values at each moment, and , For parameters exist The value of the moment, ;

[0100] Specifically, the steps of data fitting in S2.3 include:

[0101] Taking the excavation distance and a certain geotechnical parameter as an example, under actual geological conditions, the stratum situation is often very complex. Therefore, the nonlinear relationship between the geotechnical parameter and the excavation distance is considered during the excavation process to represent the changing trend of the geotechnical parameter during the excavation process and generate a fitting curve. The expression of the fitting curve is as follows:

[0102] ;

[0103] Where, is the fitting coefficient group, Excavation distance The initial values of geotechnical parameters are is the primary variation coefficient of geotechnical parameters with excavation distance, is the quadratic variation coefficient of geotechnical parameters with excavation distance, for The values of geotechnical parameters at time for The value of the excavation distance at the moment;

[0104] Construct measurement sample groups based on real-time geotechnical parameters and excavation distances ,in, 、 Respectively The geotechnical parameter values and excavation distances of the sample groups; Due to the differences between the measured values and the theoretical values calculated by the expression, the sum of the squares of the errors between the measured values and the theoretical values is calculated. , comprehensively measures the degree of deviation between the theoretical value and the actual value of all measurement points, and comprehensively reflects the fitting effect based on the given expression. Among them, the smaller the sum of squared errors, the closer the fitting curve is to the actual measured data, and the higher the degree of fitting to the actual data. The expression is as follows:

[0105] ;

[0106] Where, To measure the number of samples, yes Moment The actual geotechnical parameter values of the measurement points, is the corresponding excavation distance, ;

[0107] In order to minimize the sum of squares of the errors, 、 and Calculate partial derivatives to explore how changes in fitting coefficients affect the sum of squared errors, and set the partial derivatives to zero. At this time, the point where the partial derivative is zero corresponds to the extreme point on the fitting curve, which is also the minimum point of the sum of squared errors. The following partial derivative equations are obtained:

[0108] ;

[0109] When using the Newton iteration method to solve the optimal solution, it has a second-order convergence speed. Compared with some first-order convergence methods, such as the gradient descent method, it can converge to the solution of the equation group faster and improve the solution efficiency. At the same time, since the partial derivative equation group is a nonlinear equation group, fast convergence can save a lot of computing time and resources. Using the Newton iteration method to solve the partial derivative equation group, the fitting coefficient group is calculated. The final value of .

[0110] Specifically, the specific steps of using Newton's iteration method to solve the partial derivative equations include:

[0111] Set the partial derivative function to 、 、 Since the partial derivative function is a complex nonlinear function, it is difficult to process it directly. We use the curved trapezoid area approximation to approximate the area under the curve as the sum of the areas of multiple small trapezoids, and transform the complex nonlinear function into an easy-to-process form. By constructing a variable upper limit integral function, we consider the cumulative effect of the partial derivative function in a certain interval as a whole. For each partial derivative function , using the method of approximating the area of a curved trapezoid to construct four variable upper limit integral functions 、 、 、 ,in, ;

[0112] Different variable upper bound integral functions have different effects on the final result. By introducing weight coefficients, we can perform weighted combination of various variable upper bound integral functions, highlight the important parts, balance the contributions of different functions, and calculate the value of each variable upper bound integral function. about 、 、 Gradient ;in, ;

[0113] Assume the weight coefficient is , reconstruct and assign the variable upper limit integral function to construct a new function, the expression is as follows:

[0114] ;

[0115] in, is the newly constructed assignment function, weight coefficient Determined by those skilled in the art to ensure the convergence and speed of iteration;

[0116] By combining multiple assignment functions into a function vector, it is helpful to uniformly process and analyze these functions. The function vector is defined as ,calculate The Jacobian matrix ;

[0117] set up 、 、 The initial value of , and use the Newton iteration method to calculate the current fitting coefficient value, function vector and Jacobian matrix. 、 、 Update to minimize the sum of squared errors and gradually converge. The expression is as follows:

[0118] ;

[0119] Where, is the number of iterations, For the The fitting coefficient value of the iteration, is the Jacobian matrix exist The inverse matrix at ;

[0120] Based on the fitting coefficient value of the latest iteration, calculate the iteration error compared with the previous iteration , set the iteration termination threshold to , to determine whether the iterative operation is completed; the expression is as follows:

[0121] ;

[0122] when When , it is determined to be non-convergent and the iterative operation continues; when When , it is determined that the iteration has converged, and the fitting coefficient value of the latest iteration is output ,at this time is the fitting coefficient group The final value of .

[0123] Specifically, the specific steps of step S3 include:

[0124] S3.1: For parameter timing sequences , using STL decomposition Decomposed into trend term, seasonal term and residual term, the trend term reflects the overall trend of data changes over time, the seasonal term reflects the periodic law of data changes over time, and the residual term reflects the random fluctuations and abnormal changes in the data. By decomposing, complex time series data can be simplified into three parts that are easier to analyze and predict; the expression is as follows:

[0125] ;

[0126] in, For trend items The value of the moment, For seasonal items The value of the moment, For the residual term The value of the moment;

[0127] S3.2: Define the basic Informer model as a trend model, using historical time series Trend items in Train the trend model, process the historical time series through the self-attention mechanism, continuously adjust the model parameters, explore the long-term dependencies in the trend items, and output the trend item prediction value , the expression of the self-attention mechanism is as follows:

[0128] ;

[0129] ;

[0130] ;

[0131] in, for Trend items at the moment The corresponding predicted value, It is a multi-head attention function that uses multiple attention heads in parallel to allow the trend model to capture sequence features in different subspaces. is a concatenation function used to connect the output results of multiple attention heads by dimension. For the An attention head, , is the number of attention heads, forming the concatenated feature matrix, is the attention calculation function, is the activation function, which is used to convert the input value into a probability distribution. 、 、 They are query matrix, key matrix, and value matrix respectively. 、 、 and is the weight matrix, is the dimension of the key vector, is the number of attention heads in the multi-head attention mechanism; the self-attention mechanism can automatically focus on the importance of information at different positions in the sequence, discover the correlation between wear trends and working conditions (such as continuous cutting force and speed) over a long period of time in the past, and thus predict future wear trends;

[0132] S3.3: Define ARIMA The model is a residual model, the loss function is the mean square error, and the residual term in the historical time series is used Train the residual model, fit the dynamic characteristics of the residual term, and generate the predicted value of the residual term , which helps capture the impact of random fluctuations and anomalies in the data on the overall data, as well as the impact of unforeseen factors on tool wear, such as the instantaneous wear increase caused by material unevenness, and improves the accuracy of the overall prediction; the expression is as follows:

[0133] ;

[0134] Where, for The residual term at time The corresponding predicted value, is the backward shift operator, is the autoregressive part, is the difference order, is the moving average part, is a white noise sequence;

[0135] S3.4: Combine the trend term forecast value obtained by the trend model, the seasonal term and the residual term forecast value obtained by the residual model, comprehensively consider the influence of trend, season and residual, and fully consider the dynamic characteristics of time series data at different time scales and change patterns to generate more accurate forecast values and obtain historical time series. Historical forecast value , which enhances the ability to cope with complex and changing construction environments and parameter changes. For example, for the prediction of formation pressure, it can understand not only long-term trend changes but also periodic fluctuations and possible random anomalies; the expression is as follows:

[0136] ;

[0137] Where, for Historical forecast value at the moment;

[0138] S3.5: Use the Pearson correlation coefficient to calculate historical forecasts The linear correlation between any two variables in the matrix is calculated to construct a correlation matrix to further understand the dynamic relationship between different variables;

[0139] S3.6: Set the actual data feature vector to contain characteristic parameters, generating dimensional random vector , where the actual data feature vector includes feature time series, historical prediction value and correlation matrix, is the first random vector elements, ;

[0140] S3.7: Set the actual data feature vector to , the weight vector is , fuse the random vector with the actual data feature vector to generate a fused feature vector , so that the fused vector contains the key features of the actual data and has some new characteristics or dimensions, avoiding the limitation of relying on a single type of data for analysis, and more comprehensively depicting the various states and relationships of the shield machine or TBM during the construction process. The expression is as follows:

[0141] ;

[0142] Where, is the first fusion feature vector elements, is the first eigenvector of the actual data elements, for The corresponding weight;

[0143] S3.8: Use the random forest algorithm to construct multiple decision trees. Each decision tree selects the optimal feature to reduce impurity when the node is split. Calculate the contribution of each feature to the reduction of impurity when all decision tree nodes are split. Summarize the contribution of each feature in all decision trees and calculate the average to obtain the feature importance score, and generate a fused feature vector. Importance sequence , so as to determine which features have a greater impact on tool wear; among them, is the first fusion feature vector Elements the importance of correspondence;

[0144] S3.9: Using the multivariate linear regression method, with tool wear as the target variable, a correlation feature model is constructed. By using feature importance scores and fusion feature vectors, a correlation function is established between various influencing factors and tool wear. The expression is as follows:

[0145] ;

[0146] Where, is the tool wear, is a constant term.

[0147] Specifically, the specific steps of step S4 include:

[0148] S4.1: During the construction of a shield machine or TBM, in order to accurately grasp the detailed conditions of the rock and soil strata, the rock and soil strata are scanned with the help of external detection equipment. Geological radar and laser scanning equipment are used to scan the rock and soil strata in front of and around the shield machine or TBM. Geological radar is used to process the reflected waves from the interfaces of different strata to generate two-dimensional image data of the strata. At the same time, the laser scanning equipment is used to measure the time difference between the emission and reception of the laser beam. In combination with the principle of the speed of light, the distance between the scanning point and the equipment is calculated to obtain the three-dimensional coordinate information of the spatial point. The two-dimensional image data of the geological radar is combined with the coordinate information of the laser scanning to construct initial three-dimensional point cloud data, which intuitively displays the spatial structure, fluctuation changes, and distribution of different strata of the rock and soil strata. The initial three-dimensional point cloud data is then verified to generate three-dimensional point cloud data of the rock and soil strata to ensure the accuracy of the collected three-dimensional point cloud data.

[0149] S4.2: Fusing the 3D point cloud data with the features in the associated feature model, mapping the feature values to corresponding points in the 3D point cloud model, and generating fused point cloud data with tool wear labels, including spatial position information and environmental feature information related to tool wear;

[0150] S4.3: Preprocess the fused point cloud data and convert it into a format suitable for CNN input, including gridding and normalization.

[0151] S4.4: Use a convolutional neural network to construct a point cloud model, which includes an input layer, a multi-layer perceptron (MLP) layer, a maximum pooling layer, and a fully connected layer. The input layer receives preprocessed point cloud data, the MLP layer extracts and transforms the features of each point through a multi-layer neural network, the maximum pooling layer extracts global features, and the fully connected layer maps the extracted features to the classification results and outputs the predicted probability of the tool state.

[0152] S4.5: Randomly initialize the parameters in the point cloud model and train the point cloud model based on the processed fused point cloud data; define the loss function of the point cloud model as the cross-entropy loss function, calculate the gradient of the loss function with respect to the network parameters through the back-propagation algorithm, and use stochastic gradient descent to update the network parameters, and continuously adjust the weights and biases of the model to minimize the loss function.

[0153] Specifically, the steps for data verification in S4.1 include:

[0154] With a single data point to be verified as the center, set an initial circular search area with a search radius of , quickly locate all surrounding data points in the search area and obtain the number of surrounding points , and calculate the initial point distribution density , the expression is as follows:

[0155] ;

[0156] Set the search density range to , and judge whether the initial point distribution density is within the search density range; among them, is the high density threshold of the search density range, is the low density threshold of the search density range;

[0157] like , the point distribution density of the surrounding points is within the search density range, and the search radius is kept ;

[0158] like , then the surrounding points are sparsely distributed and the covered information is insufficient. Increase the search radius to , and recalculate the point distribution density within the new radius until the point distribution density is within the search density range; is the search coefficient;

[0159] like , then the surrounding points are too densely distributed, there is data redundancy or abnormal aggregation, and the search radius is reduced to , and recalculate the point distribution density within the new radius until the distribution density is within the search density range to ensure that the verification range can cover sufficient valid information without containing too many interference points;

[0160] Use a pseudo-random number generator to randomly generate the position coordinates of multiple sampling points in the search area, collect data for sampling points at different locations, and generate a sample data set , and calculate the mean of the sample data ;in, is the number of samples in the sample data set, The first Sample data, ;

[0161] Calculate sample data and the average The difference between , and calculate the average of the differences , as a discrete index of the search area; the expression is as follows:

[0162] ;

[0163] ;

[0164] Among them, the discrete index is used to comprehensively measure the degree of data dispersion in the entire search area, which is helpful for judging the fluctuation of data in the entire calibration range and providing a basis for subsequent judgment of whether there is an error in the data point or whether the geological conditions in the corresponding area are abnormal;

[0165] Set the discrete threshold of the search area to , and determine whether there is an error in the data point to be verified; if , the data point to be verified does not have any error, the verification is correct, and the data point is defined as a calibrated point; , the data point to be verified has an error, and the data point is defined as a preliminary error point;

[0166] In the search area of the initial error point, perform data verification on the surrounding points to obtain the number of errors in the surrounding points. And the specific average , and calculate the average value of the surrounding error , the expression is as follows:

[0167] ;

[0168] Where, For the The average value of the errors, ;

[0169] Set the geological change difference threshold to , determine whether the initial error point is a boundary point; if , the initial error point is the true error, and the initial error point is defined as the boundary point; if , the preliminary error point is a false error and is not a boundary point. Based on the surrounding points, a data interpolation algorithm is used to generate a substitute data point so that there is no error in the substitute data point;

[0170] Generate 3D point cloud data of geotechnical layers, including calibrated points and boundary points.

[0171] Specifically, the specific steps of step S5 include:

[0172] S5.1: Obtaining real-time timing sequences , and after a series of processing, it is converted into a format suitable for point cloud model input to obtain the predicted probability of the current tool state ;

[0173] S5.2: Set the damage threshold of the tool to , judge the damage of the current tool; if , the tool is judged to be in an undamaged state and the shield machine or TBM continues to operate normally; if , then the tool is judged to be damaged and a tool damage signal is generated, which contains relevant information about the tool damage, such as the approximate location of the tool damage, the degree of damage (predicted probability ), and transmit the tool damage signal to the tool changing robot;

[0174] S5.3: After the tool change is completed, various parameter data of the new tool are collected and added to the original parameter time sequence in chronological order. At the same time, the timestamp information is updated to ensure the time sequence integrity of the data;

[0175] S5.4: Utilize the updated parameter time series to optimize and adjust the associated feature model. Using incremental learning, add new point cloud data (including point cloud data of the new tool under different working conditions) to the training set, fine-tune the point cloud model, update the model parameters, and analyze the damage condition of the new tool.

[0176] Example 2

[0177] See also Figure 5 , another embodiment provided by the present invention: a shield machine or TBM cutter associated feature analysis wear system, comprising: an acquisition module, a feature analysis module, a prediction module and an update module;

[0178] The acquisition module uses various sensors to collect parameter data related to tool wear in real time, including equipment operating parameters such as propulsion speed, cutterhead torque, and thrust force; external environmental parameters such as soil type, hardness, and moisture content; and construction process parameters such as excavation mode and grouting volume.

[0179] The feature analysis module extracts various parameters in the time period based on the initial acquisition parameters. A series of statistical characteristics are analyzed and derived parameters are introduced, such as the ratio of cutterhead torque to propulsion speed and the product of jacking force and cutterhead torque. Combined with the tunneling conditions of the shield machine or TBM, the change curve of the rock and soil layer is constructed through data fitting, and the parameter time series is generated and updated in real time.

[0180] The prediction module uses the STL decomposition method to dynamically analyze parameter time series, decomposing them into trend terms, seasonal terms, and residual terms, clarifying the complex time series data structure. It also uses the random forest algorithm to calculate the importance score of each feature and construct an associated feature model. It constructs 3D point cloud data of the rock and soil layer and, combined with the associated feature model, generates fused point cloud data containing tool wear labels. It then constructs a point cloud model for determining tool status, outputs the predicted probability of tool status, and determines the tool status.

[0181] The update module is used to collect the parameter data of the new tool after the tool change is completed, update the parameter timing sequence, and use the new data to optimize and adjust the associated feature model and point cloud model to adapt to the new operating status and environmental changes of the shield machine or TBM, thereby improving the accuracy and reliability of model discrimination.

[0182] Specifically, the prediction module includes an association unit and a judgment unit;

[0183] The correlation unit is used to dynamically analyze the parameter time series, calculate the importance score of each feature, build a correlation feature model, and comprehensively consider the complex relationship between various features and tool wear to achieve an effective description of the tool wear status;

[0184] The judgment unit uses geological radar and laser scanning equipment to scan the rock and soil layers in front of and around the shield machine or TBM, constructs three-dimensional point cloud data of the rock and soil layers, combines it with the associated feature model, generates fused point cloud data with tool wear labels, and constructs a point cloud model specifically for judging the tool status; based on the real-time time series and point cloud model, it outputs the predicted probability of the tool status and judges the tool status; once it is determined that the tool is damaged, a tool damage signal is generated, and the control center starts the tool changing robot.

[0185] In summary, the present invention utilizes various sensors to collect parameter data related to tool wear in real time, constructs a parameter time series sequence that can reflect dynamic changes, collects actual parameter values at different sampling frequencies and saves them in chronological order; calculates relevant statistics and derivative features, constructs a rock and soil layer change curve, and timely updates the parameter time series sequence. By analyzing the dynamic changes of the parameter time series sequence, the random forest algorithm is used to clarify the influence of various features on tool wear, and an associated feature model is constructed to effectively describe the tool wear state; at the same time, the three-dimensional point cloud data of the rock and soil layer is constructed and fused with the associated feature model to generate fused point cloud data containing tool wear labels and establish a point cloud model; the tool state is judged based on the real-time sequence and point cloud model. Once the tool is damaged, the tool changing robot is started. After the tool change, new data is collected to optimize the associated features and point cloud model, thereby improving the model's adaptability to new working conditions of the shield machine or TBM and the accuracy of judgment.

[0186] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiment. All technical solutions based on the concept of the present invention are within the scope of protection of the present invention. It should be noted that for those skilled in the art, various improvements and modifications that do not depart from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for analyzing wear of shield machine or TBM cutter by correlation characteristics, characterized in that: include: Step S1: collecting parameter data related to tool wear in real time, performing preprocessing, and constructing a parameter time series; Step S2: Calculate the statistical characteristics, derived characteristics and the change curve of the rock and soil layer, and update the parameter time series; Step S3: Analyze the dynamic changes of the parameter time series, calculate the feature importance score using the random forest algorithm, and build a correlation feature model; Step S4: constructing three-dimensional point cloud data of the rock and soil layer, generating fused point cloud data containing tool wear labels, and constructing a point cloud model for determining the tool status; Step S5: Calculate the predicted probability of the tool state and determine the tool state; once it is determined that the tool is damaged, generate a tool damage signal.

2. The shield machine or TBM cutter wear analysis method according to claim 1, characterized in that: The specific steps of step S2 include: S2.1: Based on the real-time collected cutter head torque and propulsion speed , calculate the ratio of cutter head torque to propulsion speed ; S2.2: Based on the real-time collected thrust , calculate the product of the thrust and the cutter head torque ; S2.3: Continuously record the actual values of geotechnical parameters at different time points and draw the geotechnical parameter curves changing with time by data fitting; S2.4: Save the calculated statistical characteristics, derived characteristics and change curves of the rock and soil layers to the parameter time series In which, the parameter time series sequence includes the historical time series sequence and real-time timing sequences .

3. The shield machine or TBM cutter wear analysis method according to claim 2, characterized in that: The specific steps of data fitting in S2.3 include: During the excavation process, a fitting curve is generated based on the nonlinear relationship between a certain geotechnical parameter and the excavation distance; Constructing a measurement sample group , calculate the sum of squares of the errors ; respectively 、 and Find the partial derivatives and set them to zero to obtain the partial derivative equations; Solve the partial derivative equations and calculate the fitting coefficient group The final value of .

4. The shield machine or TBM cutter wear analysis method based on correlation characteristics according to claim 3, characterized in that: The specific steps of solving the partial derivative equations include: Assume that the functions in the partial derivative equations are partial derivative functions 、 、 , and use the curved trapezoid area approximation to construct a variable upper limit integral function for each partial derivative function; Calculate the gradient of each variable upper limit integral function with respect to each fitting coefficient respectively; Assume the weight coefficient is , reconstruct and assign the variable upper limit integral function to generate a newly constructed assignment function; Combine multiple assignment functions into a function vector and calculate the Jacobian matrix of the function vector ; set up 、 、 The initial value is , and use the Newton iteration method to calculate the current fitting coefficient value, function vector and Jacobian matrix. 、 、 To update: Calculate the iterative error of the fitted coefficient set , set the iteration termination threshold to , to determine whether the iterative operation is completed; when When , it is determined to be non-convergent and the iterative operation continues; when When , it is determined that the iteration has converged and the fitting coefficient value of the latest iteration is output.

5. The shield machine or TBM cutter wear analysis method based on correlation characteristics according to claim 4, characterized in that: The specific steps of step S3 include: S3.1: For the parameter timing sequence , using STL decomposition Decompose into trend term, seasonal term and residual term; S3.2: Use the trend items in the historical time series to train the trend model and output the predicted value of the trend item; S3.3: Using the residual items in the historical time series to train the residual model, fitting the dynamic characteristics of the residual items, and generating predicted values of the residual items; S3.4: Combine the trend item prediction value, the seasonal item in the historical time series, and the residual item prediction value to obtain a historical prediction value.

6. The shield machine or TBM cutter wear analysis method based on correlation characteristics according to claim 5, characterized in that: The specific steps of step S3 also include: S3.5: Calculate the linear correlation between any two variables in the historical forecast values and construct a correlation matrix; S3.6: Set the actual data feature vector to contain characteristic parameters, generating dimensional random vector ; S3.7: Set the actual data feature vector to , the weight vector is , fuse the random vector with the actual data feature vector to generate a fused feature vector .

7. The shield machine or TBM cutter wear analysis method according to claim 6, characterized in that: The specific steps of step S3 also include: S3.8: Use the random forest algorithm to construct multiple decision trees, calculate the feature importance score, and generate the fusion feature vector Importance sequence ; S3.9: Taking tool wear as the target variable, the correlation function between various influencing factors and tool wear is established through feature importance score and fusion feature vector to generate the correlation feature model.

8. The shield machine or TBM cutter wear analysis method based on correlation characteristics according to claim 7, characterized in that: The specific steps of step S4 include: S4.1: Use geological radar and laser scanning equipment to scan the rock and soil layers in front of and around the shield machine or TBM, perform data verification, and generate three-dimensional point cloud data of the rock and soil layers; S4.2: Fusing the three-dimensional point cloud data with features in the associated feature model to generate fused point cloud data with a tool wear label; S4.3: Preprocessing the fused point cloud data, including gridding and normalization. S4.4: Build a point cloud model, including an input layer, a multi-layer perceptron layer, a max pooling layer, and a fully connected layer, where the output is the predicted probability of the tool state; S4.5: Randomly initialize the parameters in the point cloud model, and train the point cloud model based on the processed fused point cloud data.

9. The shield machine or TBM cutter wear analysis method based on correlation characteristics according to claim 8, characterized in that: The specific steps of data verification in S4.1 include: Set the initial search area with the data point to be verified as the center, obtain the number of surrounding points in the search area, and calculate the initial point distribution density ; Set the search density range to , and judging whether the point distribution density is within the search density range; like , the point distribution density is within the search density range; like , expand the search range until the point distribution density is within the search density range; like , reduce the search range until the point distribution density is within the search density range; Get the location coordinates of multiple sampling points and generate a sample data set , and calculate the mean of the sample data ; Calculate sample data With the average The difference between , and calculate the average of the differences ; Set the discrete threshold of the search area to , and determine whether there is an error in the data point; if , there is no error in the data point, and the data point is defined as a calibrated point; , the data point has an error, defining the data point as a preliminary error point; In the search area of the preliminary error point, the data of the surrounding points is verified to obtain the number of errors in the surrounding points. and the average , and calculate the average value of the surrounding error ;in, For the The average value of the errors; Set the geological change difference threshold to , determine whether the initial error point is a boundary point; if , the preliminary error point is the boundary point; if , generate alternative data points without errors; Generate 3D point cloud data of geotechnical layers, including calibrated points and boundary points.

10. The shield machine or TBM cutter wear analysis method based on correlation characteristics according to claim 9, characterized in that: The specific steps of step S5 include: S5.1: Obtaining real-time timing sequences , calculate the predicted probability of tool status ; S5.2: Set the damage threshold of the tool to , judge the damage of the current tool; if , the tool is in an undamaged state; if , the tool is damaged, generating a tool damaged signal; S5.3: After the tool change is completed, various parameter data of the new tool are collected, and the parameter timing sequence and timestamp information are updated; S5.4: Utilize the updated parameter timing sequence to optimize and adjust the associated feature model and point cloud model, and perform damage analysis on the new tool.

11. A shield machine or TBM cutter wear analysis system based on correlation characteristics, which is used to implement the shield machine or TBM cutter wear analysis method based on correlation characteristics as claimed in any one of claims 1 to 10, characterized in that: include: Data acquisition module, feature analysis module, prediction module and update module; The data acquisition module uses various sensors to collect parameter data related to tool wear in real time; The characteristic analysis module calculates statistical parameters and derived parameters based on the initial acquisition parameters, and constructs a change curve of the rock and soil layer in combination with the tunneling conditions of the shield machine or TBM to generate a parameter time series; The prediction module includes an association unit and a judgment unit; The association unit is used to dynamically analyze the parameter time series, calculate the importance score of each feature, and build an association feature model; The judgment unit is used to construct three-dimensional point cloud data of the rock and soil layer, combine it with the associated feature model, generate fused point cloud data with tool wear labels, and build a point cloud model for judging the tool status; output the predicted probability of the tool status, and judge the tool status; The updating module is used to collect parameter data of the new tool after the tool change is completed, update the parameter timing sequence, and optimize and adjust the associated feature model and the point cloud model using the new data.

Images