A Shield Machine Tool Wear Prediction Method and System

Abstract

The present invention provides a method and system for predicting the wear of shield machine cutters, belonging to the field of shield technology. By integrating shield tunneling parameters and vibration signals, the present invention realizes the intelligent prediction from the data of the shield tunneling process to the wear degree of the cutters, provides a new solution for the accurate evaluation of the cutter state and efficient cutter replacement during the shield tunneling process, and also provides important technical support for shield construction under complex stratum conditions, and can maximize the benefits of cutter replacement during the shield tunneling process.

Description

Technical Field

[0001] The present invention belongs to the technical field of shield tunneling, and particularly relates to a method and system for predicting the wear of shield machine cutters. Background Art

[0002] Shield tunneling is a process in which the cutters of a shield machine interact with the formation. The degree of cutter wear directly affects the construction progress and efficiency of the project. In hard rock formations, shield tunneling is usually accompanied by frequent operations of opening the chamber to replace cutters. The technology of opening the chamber to replace cutters is highly difficult and risky, seriously affecting the safe and efficient construction of the whole process of shield tunneling. Therefore, accurately determining the cutter replacement time is crucial for reducing project losses and improving construction efficiency.

[0003] At present, the cutter replacement time mainly depends on the experience judgment of on-site construction personnel. Some studies have tried to monitor the whole process of cutter wear, establish the mapping relationship between shield tunneling parameters and the degree of cutter wear, and use the regression fitting model to predict the wear process of the cutter. However, these methods still face two major bottlenecks: firstly, the limited scale of data samples leads to insufficient generalization ability of the model; secondly, the fitting ability of empirical formulas is weak, and the prediction accuracy is low when applied across projects.

[0004] Therefore, it is necessary to provide a method and system for predicting the wear of shield machine cutters to solve the above problems. Summary of the Invention

[0005] The present invention provides a method and system for predicting the wear of shield machine cutters. By fusing tunneling parameters and vibration signals, the degree of cutter wear is predicted, so as to accurately determine the cutter replacement time, which can effectively solve at least one technical problem involved in the background art.

[0006] In order to solve the above technical problems, the present invention is implemented as follows:

[0007] A method for predicting the wear of shield machine cutters includes the following steps:

[0008] Step S1, collecting tunneling parameters and vibration signals during the tunneling process of the shield machine;

[0009] Step S2, performing collaborative filtering on the tunneling parameters and vibration signals based on the singular value decomposition method to capture the principal component signals related to the cutter wear condition in the tunneling parameters and vibration signals;

[0010] Step S3, correcting the traditional cutter wear empirical formula based on the work log of the shield machine in the actual engineering scenario, and generating cutter wear degree data by using the corrected cutter wear formula;

[0011] Step S4: Feature extraction is performed on the vibration signal and the principal component signal of the tunneling parameters respectively. The extracted vibration signal features and tunneling parameter features are fused through matrix splicing to construct a unified feature representation. Then, a mapping model from the fused features to the tool wear degree is established using a fully connected neural network. A large number of tool wear degree data under different tunneling parameters and vibration signals are generated using the tool wear formula to train the mapping model.

[0012] Step S5: For any shield machine tool wear prediction process, the fused features of the vibration signal and the principal component signal of the tunneling parameters are input into the mapping model, and the prediction result of the tool wear degree is output.

[0013] As a preferred improvement, the tunneling parameters include the average propulsion speed, total propulsion force, cutterhead torque, cutterhead inclination angle, cutterhead speed, cutterhead penetration, propulsion cylinder stroke, and articulated cylinder stroke; the vibration signals include the cutting vibration signal between the tool and the formation and the tool rotation vibration signal.

[0014] As a preferred improvement, step S2 specifically includes the following steps:

[0015] Step S21: Obtain the power spectrum data of the tunneling parameters and vibration signals, and use the Gaussian function superposition model to fit the power spectra of the tunneling parameters and vibration signals respectively.

[0016] Step S22: Calculate the half-bandwidth of each Gaussian component in the tunneling parameters and vibration signals based on the fitted Gaussian function model, and determine the principal component frequency band range of the tunneling parameters and vibration signals based on the half-bandwidth.

[0017] Step S23: Stack the one-dimensional tunneling parameters and vibration signals into a three-dimensional tensor using the bitwise cycling method.

[0018] Step S24: Decompose the obtained three-dimensional tensor into three tensor matrices U, S, and V using the three-dimensional tensor singular value decomposition method.

[0019] Step S25: Search for the columns in the spectrum of the tensor matrix U that are located between the principal component frequency band ranges, extract them separately, and regenerate the signals using the reverse cycling method to obtain the principal component signals of the tunneling parameters and vibration signals.

[0020] As a preferred improvement, in step S21, the Gaussian function superposition model is expressed as:

[0021] ;

[0022] In the formula, represents the fitting function of the power spectrum; represents the serial number index of the significant peak in the power spectrum; represents the total number of significant peaks; The power value representing the significant peak; The average power of the significant peak; The standard deviation of the power value of the significant peak;

[0023] The fitting function of the power spectrum is solved by the Levenberg–Marquardt algorithm.

[0024] As a preferred improvement, in step S22, the calculation process of the half bandwidth is expressed as:

[0025] ;

[0026] In the formula, represents the half bandwidth;

[0027] The main component frequency band range is expressed as: , in the formula, represents the larger value between the tunneling parameters and the frequency corresponding to the highest point in the frequency spectrum of the vibration signal.

[0028] As a preferred improvement, in step S23, the bit - by - bit circular method specifically includes the following steps: Define the window length, step size, and number of cycles, perform circular shifting on the tunneling parameters and vibration signals according to the defined window length, step size, and cycle coefficient to generate multiple groups of window segments, and stack the window matrices of the tunneling parameters and vibration signals along the third dimension to form a three - dimensional tensor.

[0029] As a preferred improvement, in step S24, the three - dimensional tensor singular value decomposition method decomposes the three - dimensional tensor into the product of third - order tensors based on the tensor tube product, and the decomposition form is expressed as:

[0030] ;

[0031] In the formula, represents the three - dimensional tensor; represents an orthogonal matrix of order represents an orthogonal matrix of order represents a rectangular diagonal matrix composed of non - negative diagonal elements arranged in descending order, and the elements on the diagonal are singular values.

[0032] As a preferred improvement, step S3 specifically includes the following steps:

[0033] Step S31, collect the time series of the propulsion speed of each ring of the shield machine, perform wavelet transform on the time series of the propulsion speed of each ring, take the low - frequency approximation of the last layer of the wavelet transform to form a low - frequency approximation sequence, and calculate the standard deviation of the low - frequency approximation sequence; the calculation process of the standard deviation is expressed as:

[0034] ;

[0035] Wherein, represents the standard deviation of the low-frequency approximation sequence of the th ring of the shield machine; represents the th element in the low-frequency approximation sequence of the th ring of the shield machine; represents the average value of the low-frequency approximation sequence of the th ring of the shield machine; N represents the total number of elements in the low-frequency approximation sequence of the th ring of the shield machine;

[0036] Step S32: Combine the standard deviations of all rings to form a standard deviation sequence, and perform Fourier noise reduction on the standard deviation sequence. The process of Fourier noise reduction is expressed as:

[0037] ;

[0038] ;

[0039] ;

[0040] Wherein, represents the Fourier transform; represents the sequence after the Fourier transform; represents truncating the sequence after the Fourier transform, and all elements from the 1 / 3 length position of the sequence to the end of the sequence are assigned a value of 0; represents the inverse Fourier transform; represents the sequence after the inverse Fourier transform;

[0041] Step S33: The noise-reduced standard deviation passes through to obtain the final tool wear index .

[0042] As a preferred improvement, a one-dimensional convolutional neural network is used to perform feature extraction on the vibration signal; a fully connected neural network is used to perform feature extraction on the tunneling parameters.

[0043] A system for implementing the above-mentioned shield machine tool wear prediction method, comprising:

[0044] A signal acquisition module, configured to acquire tunneling parameters and vibration signals during the tunneling process of the shield machine;

[0045] A principal component extraction module, configured to perform collaborative filtering on the tunneling parameters and vibration signals based on the singular value decomposition method, and capture the principal component signals related to the tool wear condition in the tunneling parameters and vibration signals;

[0046] A tool wear degree data generation module, which is used to correct the traditional tool wear empirical formula based on the work log of the shield machine in the actual engineering scenario, and generate tool wear degree data by using the corrected tool wear formula;

[0047] A mapping model construction module, which is used to extract features from the vibration signal and the principal component signal of the tunneling parameters respectively, fuse the extracted vibration signal features and tunneling parameter features through matrix splicing to construct a unified feature representation, then use a fully connected neural network to establish a mapping model from the fused features to the tool wear degree, and generate a large amount of tool wear degree data under different tunneling parameters and vibration signals by using the tool wear formula to train the mapping model;

[0048] A prediction module, which is used for any shield machine tool wear prediction process, inputs the fused features of the vibration signal and the principal component signal of the tunneling parameters into the mapping model, and outputs the prediction result of the tool wear degree.

[0049] The beneficial effects of the present invention are as follows:

[0050] By fusing the shield tunneling parameters and vibration signals, the present invention realizes the intelligent prediction from the shield tunneling process data to the tool wear degree, provides a new solution for the accurate assessment of the tool state and efficient tool changing during the shield tunneling process, and also provides important technical support for the shield construction under complex stratum conditions, and can maximize the tool changing benefit during the shield tunneling process. Specific implementation manners

[0051] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0052] This implementation manner provides a method for predicting the tool wear of a shield machine, including the following steps:

[0053] Step S1, collect the tunneling parameters and vibration signals during the tunneling process of the shield machine.

[0054] The tunneling parameters include the average propulsion speed, total propulsion force, cutter head torque, cutter head inclination angle, cutter head speed, cutter head penetration, propulsion cylinder stroke and articulated cylinder stroke. The tunneling parameters are collected from the operation data of the shield machine.

[0055] The vibration signal includes the cutting vibration signal between the cutter and the formation and the rotational vibration signal of the cutter. The cutting vibration signal between the cutter and the formation is collected by a uniaxial sensor arranged at the bulkhead of the soil chamber behind the cutter head; the rotational vibration signal of the cutter is collected by a triaxial sensor arranged at the motor, gearbox, speed reducer, bearing and screw conveyor in the tunneling system. The acquisition frequency of all sensors is 1000 Hz, which can collect 1000 sets of high-resolution vibration data per second, so as to fully record the dynamic information during tunneling.

[0056] During the tunneling process of the shield machine, both the vibration signal of the cutter cutting the formation and the rotational vibration signal of the cutter itself can reflect the vibration state of the cutter. Therefore, by collecting the vibration signals of these two parts simultaneously, the vibration condition of the cutter during tunneling can be obtained more comprehensively.

[0057] Step S2: Based on the singular value decomposition method, perform collaborative filtering on the tunneling parameters and vibration signals to capture the principal component signals related to the tool wear condition in the tunneling parameters and vibration signals.

[0058] In the traditional signal processing field, usually only a single signal is considered. For example, only the influence of tunneling parameters on tool wear is considered, and after filtering out the noise, the subsequent wear prediction process is carried out. The disadvantage is that the principal components of the signal cannot be effectively determined, and it is difficult to determine the selection of the cut-off frequency. The applicant found in the research that the rotation of the cutter is driven by the motor of the motor tunneling system and the conduction between the components in the tunneling system (including but not limited to the gearbox, speed reducer, bearing), and it can effectively propagate in all directions to the screw conveyor that transports the muck in the front to the rear. The cutting action between the cutter and the formation can be accompanied by strong vibrations, which can axially propagate to the rear bulkhead of the soil chamber, etc. Therefore, the tool wear condition is related to both the tunneling parameters and the vibration signals at the same time, and there is a certain coupling correlation between the tunneling parameters and the vibration signals. The present invention uses the idea that there are common principal components in the tunneling parameters and vibration signals, and extracts the common principal components of the two signals to be used to identify the working state of the cutter.

[0059] Step S2 specifically includes the following steps:

[0060] Step S21: Obtain the power spectrum data of the tunneling parameters and vibration signals, and use the Gaussian function superposition model to fit the power spectra of the tunneling parameters and vibration signals respectively.

[0061] The number of Gaussian functions can be selected according to the number of significant peaks in the power spectrum. For example, when the number of significant peaks is three, three Gaussian functions can be selected for combination to fit the power spectrum. The superposition model of the Gaussian function is expressed as:

[0062] ;

[0063] In the formula, represents the fitting function of the power spectrum; represents the serial number index of the significant peak; represents the total number of significant peaks; represents the power value of the significant peak; represents the average power of the significant peak; represents the standard deviation of the power values of the significant peaks.

[0064] The fitting function of the power spectrum can be solved by the Levenberg–Marquardt algorithm, and the solving process belongs to the conventional technology in this field, so it will not be elaborated in this embodiment.

[0065] Step S22: Calculate the half-bandwidth of each Gaussian component of the tunneling parameters and the vibration signal based on the fitted Gaussian function model, and determine the main component frequency band range of the tunneling parameters and the vibration signal based on the half-bandwidth.

[0066] The calculation process of the half-bandwidth is expressed as:

[0067] ;

[0068] In the formula, represents the half-bandwidth;

[0069] Since the periodic content in the signal occupies most of the energy, the main components are generally distributed between where, in the formula, represents the larger one of the frequencies corresponding to the highest points in the frequency spectra of the tunneling parameters and the vibration signal;

[0070] Step S23: Stack the one-dimensional tunneling parameters and vibration signals into a three-dimensional tensor by the bitwise cyclic method;

[0071] The bitwise cyclic method specifically includes the following steps: Define the window length, step size, and number of cycles, perform cyclic shifts on the tunneling parameters and vibration signals according to the defined window length, step size, and cyclic coefficient to generate multiple groups of window segments, and stack the window matrices of the tunneling parameters and vibration signals along the third dimension (channel) to form a three-dimensional tensor.

[0072] Step S24: Decompose the obtained three-dimensional tensor into three tensor matrices of U, S, and V by the three-dimensional tensor singular value decomposition method.

[0073] The three-dimensional tensor singular value decomposition method decomposes the three-dimensional tensor into the product of three-order tensors based on the tensor tube product, and the decomposition form is expressed as:

[0074] ;

[0075] In the formula, represents the three-dimensional tensor; denote an n - order orthogonal matrix; denote an m - order orthogonal matrix; denote a rectangular diagonal matrix composed of non - negative diagonal elements arranged in descending order, and the elements on the diagonal are singular values;

[0076] Step S23: Search for the columns in the spectrum of the tensor matrix U that are within the range of the principal component frequency bands, extract them separately, and then use the anti - cycle method to regenerate the signal to obtain the principal component signals of the tunneling parameters and the vibration signal.

[0077] The principal component signals of the tunneling parameters and the vibration signal filter out the unnecessary noise and can intuitively display the working state of the cutter.

[0078] Step S3: Based on the work log of the shield machine in the actual engineering scenario, correct the traditional cutter wear empirical formula, and use the corrected cutter wear formula to generate cutter wear degree data.

[0079] Step S3 specifically includes the following steps:

[0080] Step S31: Collect the propulsion speed time series of each ring of the shield machine, perform wavelet transform on the propulsion speed time series of each ring, take the low - frequency approximation of the last layer of the wavelet transform to form a low - frequency approximation sequence, and calculate the standard deviation of the low - frequency approximation sequence.

[0081] The essence of wavelet transform is a multi - level decomposition process. Each layer represents a multi - scale analysis of the signal, gradually peeling off different frequency components of the signal. Specifically: The first layer decomposes the original signal into high - frequency details and low - frequency approximation; the second layer decomposes the low - frequency approximation of the first layer again to obtain higher - frequency high - frequency details and lower - frequency low - frequency approximation; repeat this process until the preset decomposition layer number is reached.

[0082] In this embodiment, wavelet transform is used for signal denoising and includes a total of 4 layers. Noise generally appears in the high - frequency details. Therefore, this application retains the low - frequency approximation of the fourth layer and discards the high - frequency details. The low - frequency approximation of the fourth layer has a strong correlation with the wear state of the cutter and can directly reflect the wear condition of the cutter.

[0083] The calculation process of the standard deviation is expressed as:

[0084] ;

[0085] In the formula, denotes the standard deviation of the low - frequency approximation sequence of the th ring of the shield machine; denotes the th element in the low - frequency approximation sequence of the​ elements; represent the average value of the low-frequency approximation sequence of the th ring of the shield machine; N represents the total number of elements in the low-frequency approximation sequence of the th ring of the shield machine;

[0086] Step S32: Combine the standard deviations of all rings to form a standard deviation sequence, and perform Fourier noise reduction on the standard deviation sequence.

[0087] Fourier noise reduction transforms the original signal from the time domain to the frequency domain through Fourier transform to obtain the spectrum of the signal, then removes the high-frequency noise part by direct truncation, and finally restores the original signal through inverse Fourier transform.

[0088] The process of Fourier noise reduction is expressed as

[0089] ;

[0090] ;

[0091] ;

[0092] In the formula, represents Fourier transform; represents the sequence after Fourier transform; represents truncating the sequence after Fourier transform, and all elements from the 1 / 3 length position of the sequence to the end of the sequence are assigned 0; represents inverse Fourier transform; represents the sequence after inverse Fourier transform.

[0093] Step S33: The noise-reduced standard deviation passes through to obtain the final tool wear index .

[0094] It can be understood that the traditional tool wear empirical formula belongs to a general formula fitted with a large amount of data. In actual engineering scenarios, there is often a problem of insufficient generalization ability, manifested as large oscillations in the tool wear curve, which cannot accurately reflect the actual wear degree. Therefore, the present invention collects the work logs in actual engineering scenarios to correct it so that it can better perform tool wear prediction in actual engineering scenarios.

[0095] It should be noted that the work log generally only records the relevant data when the tool is completely worn out (reaching the tool change standard). Therefore, if only the relevant data in the work log is used as the input for the subsequent fully connected neural network, there will be problems such as too few and single data samples. The corrected tool wear formula can generate tool wear data throughout the tool's life cycle, which can provide a large number of data samples for subsequent training and is conducive to the rapid fitting of the model.

[0096] Step S4: Feature extraction is performed on the vibration signal and the principal component signal of the tunneling parameters respectively. The extracted vibration signal features and tunneling parameter features are fused through matrix splicing to construct a unified feature representation. Then, a mapping model from the fused features to the tool wear degree is established using a fully connected neural network. A large number of tool wear degree data under different tunneling parameters and vibration signals are generated using the tool wear formula, and the mapping model is trained.

[0097] Since the time scales of the tunneling parameters and the vibration signal do not match, mainly manifested in inconsistent frequencies, before feature extraction, it is also necessary to preprocess the vibration signal and the tunneling parameters. The present invention uses the root mean square to calculate the effective value of the signal to align the time scales. The processing process is expressed as:

[0098] ;

[0099] In the formula, represents the output signal after alignment; represents the input signal, represents the index in the signal sequence, represents the number of samples in the signal sequence.

[0100] Since the vibration signal has strong periodicity, the present invention uses a one-dimensional convolutional neural network (1D CNN) to perform the feature extraction process of the vibration signal. The sliding of the convolutional kernel during the convolution process can effectively detect short-term trends or fluctuations; its parameter sharing mechanism enables the model to effectively capture position-independent patterns in the input sequence; it does not rely on time step calculation and is more efficient.

[0101] The one-dimensional convolutional neural network includes two convolutional layers. A batch normalization layer and a ReLU activation function are connected behind each convolutional layer. Then, a max-pooling layer is used to reduce the spatial dimension of the feature map, gradually extracting more abstract and invariant features from the vibration signal. Finally, the feature map is flattened and passed through two fully connected layers to further optimize the learned feature representation, and a 32-dimensional feature vector is finally output.

[0102] The fully connected neural network is a basic deep learning model, which consists of a series of fully connected layers. Each neuron between layers is interconnected. It performs well in processing high-dimensional structured data (such as tabular data, etc.) and can effectively model the complex non-linear relationship between high-dimensional structured tunneling parameters and tool wear. Therefore, the present invention uses a fully connected neural network to extract features of tunneling parameters.

[0103] The fully connected neural network includes a fully connected layer, a batch normalization layer, a ReLU activation function, and a regularization layer connected in sequence. The input tunneling parameters are first mapped to a 64-dimensional feature space and then further reduced to 32. Batch normalization is applied after each linear transformation to improve the stability of feature extraction, and the regularization layer is used to randomly deactivate a small number of neurons during training to reduce overfitting.

[0104] Step S5, for any shield machine tool wear prediction process, input the fusion features of the vibration signal and the principal component signal of the tunneling parameters into the mapping model, and output the prediction result of the tool wear degree.

[0105] To verify the effectiveness of the method provided by the present invention, 14 input parameters of the shield tunneling parameters (average propulsion speed, total propulsion force, cutterhead torque, inclination angle, cutterhead speed, penetration rate, stroke of No. 2 propulsion cylinder, stroke of No. 5 propulsion cylinder, stroke of No. 7 propulsion cylinder, stroke of No. 10 propulsion cylinder, stroke of No. 1 articulation cylinder, stroke of No. 2 articulation cylinder, stroke of No. 3 articulation cylinder, stroke of No. 4 articulation cylinder) in a certain construction project and the minute-level effective values of the vibration signals of all measuring points (motor, gearbox, reducer, bearing, screw conveyor - XYZ three axes; upper and lower soil chamber partitions - X single axis), a total of 17 input parameters are used to train the model. The R-square (evaluation index) of the model on the test set reaches an accuracy of 95%, showing excellent performance. In addition, two machine learning models of other methods are set as comparative tests, using the random forest algorithm based only on tunneling parameters and the FNN algorithm based only on tunneling parameters respectively, and the accuracies are both lower than 90%, further verifying the superiority of the present invention.

[0106] This embodiment also provides a system for executing the above-mentioned shield machine tool wear prediction method, including:

[0107] A signal acquisition module for collecting tunneling parameters and vibration signals during the tunneling process of the shield machine;

[0108] A principal component extraction module for performing collaborative filtering on the tunneling parameters and vibration signals based on the singular value decomposition method to capture the principal component signals related to the tool wear condition in the tunneling parameters and vibration signals;

[0109] The tool wear degree data generation module is used to correct the traditional tool wear empirical formula based on the work log of the shield machine in the actual engineering scenario, and generate tool wear degree data by using the corrected tool wear formula;

[0110] The mapping model construction module is used to extract features from the vibration signal and the principal component signal of the tunneling parameters respectively, fuse the extracted vibration signal features and tunneling parameter features through matrix splicing to construct a unified feature representation, then use a fully connected neural network to establish a mapping model from the fused features to the tool wear degree, and generate a large amount of tool wear degree data under different tunneling parameters and vibration signals by using the tool wear formula to train the mapping model;

[0111] The prediction module is used for any shield machine tool wear prediction process, input the fused features of the vibration signal and the principal component signal of the tunneling parameters into the mapping model, and output the prediction result of the tool wear degree.

[0112] The above describes the embodiments of the present invention, but the present invention is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many forms without departing from the purpose of the present invention and the scope protected by the claims, and all belong to the protection scope of the present invention.

Claims

1. A method for predicting the tool wear of a shield machine, characterized in that, It includes the following steps: Step S1, collect the tunneling parameters and vibration signals during the tunneling process of the shield machine; Step S2, perform collaborative filtering on the tunneling parameters and vibration signals based on the singular value decomposition method to capture the principal component signals related to the tool wear condition in the tunneling parameters and vibration signals; Step S3, correct the traditional tool wear empirical formula based on the work log of the shield machine in the actual engineering scenario, and generate tool wear degree data using the corrected tool wear formula; Step S4, extract features from the principal component signals of the vibration signals and tunneling parameters respectively, fuse the extracted vibration signal features and tunneling parameter features through matrix splicing to construct a unified feature representation, then use a fully connected neural network to establish a mapping model from the fused features to the tool wear degree, generate a large amount of tool wear degree data under different tunneling parameters and vibration signals using the tool wear formula, and train the mapping model; Step S5, for any shield machine tool wear prediction process, input the fused features of the principal component signals of the vibration signals and tunneling parameters into the mapping model, and output the prediction result of the tool wear degree; Step S2 specifically includes the following steps: Step S21, obtain the power spectrum data of the tunneling parameters and vibration signals, and use the Gaussian function superposition model to fit the power spectra of the tunneling parameters and vibration signals respectively; Step S22, calculate the half bandwidth of each Gaussian component in the tunneling parameters and vibration signals based on the fitted Gaussian function model, and determine the principal component frequency band range of the tunneling parameters and vibration signals based on the half bandwidth; Step S23, use the bitwise circular method to stack the one-dimensional tunneling parameters and vibration signals into a three-dimensional tensor; Step S24, use the three-dimensional tensor singular value decomposition method to decompose the obtained three-dimensional tensor into three tensor matrices U, S, and V; Step S25, find the columns in the spectrum of the tensor matrix U that are located between the principal component frequency band ranges, extract them separately and use the reverse circular method to regenerate the signals to obtain the principal component signals of the tunneling parameters and vibration signals.

2. The shield machine tool wear prediction method according to claim 1, wherein The tunneling parameters include the average propulsion speed, total propulsion force, cutterhead torque, cutterhead inclination angle, cutterhead speed, cutterhead penetration, propulsion cylinder stroke, and articulated cylinder stroke; the vibration signals include the cutting vibration signal between the tool and the formation and the tool rotation vibration signal.

3. The shield machine tool wear prediction method according to claim 1, characterized in that In Step S21, the Gaussian function superposition model is expressed as: ; wherein, represents the fitting function of the power spectrum; represents the serial number index of the significant peak in the power spectrum; represents the total number of significant peaks; represents the power value of the significant peak; represents the average power of the significant peak; represents the standard deviation of the power values of the significant peaks; The fitting function of the power spectrum is solved by the Levenberg–Marquardt algorithm.

4. The shield machine tool wear prediction method according to claim 3, wherein In Step S22, the calculation process of the half bandwidth is expressed as: ； In the formula, represents the half bandwidth; The main component frequency band range is expressed as: , where represents the larger value of the frequencies corresponding to the highest points in the power spectra of the tunneling parameters and the vibration signals.

5. The shield machine tool wear prediction method according to claim 1, wherein In Step S23, the bitwise circular method specifically includes the following steps: define the window length, step size, and number of cycles, perform circular shifting on the tunneling parameters and vibration signals according to the defined window length, step size, and number of cycles to generate multiple groups of window segments, and stack the window matrices of the tunneling parameters and vibration signals along the third dimension to form a three-dimensional tensor.

6. The shield machine tool wear prediction method according to claim 1, characterized in that In Step S24, the three-dimensional tensor singular value decomposition method decomposes the three-dimensional tensor into the product of third-order tensors based on the tensor contraction, and the decomposition form is expressed as: ； In the formula, represents a three-dimensional tensor; represents an orthogonal matrix of order represents an orthogonal matrix of order represents a rectangular diagonal matrix composed of non-negative diagonal elements arranged in descending order, and the elements on the diagonal are singular values.

7. The shield machine tool wear prediction method according to claim 1, wherein Step S3 specifically includes the following steps: Step S31: Collect the propulsion speed time series of each ring of the shield machine, perform wavelet transform on the propulsion speed time series of each ring, take the low-frequency approximation of the last layer of the wavelet transform to form a low-frequency approximation sequence, and calculate the standard deviation of the low-frequency approximation sequence. The calculation process of the standard deviation is expressed as: ； In the formula, represents the standard deviation of the low-frequency approximation sequence of the th ring of the shield machine; represents the th element in the low-frequency approximation sequence of the th ring of the shield machine; represents the average value of the low-frequency approximation sequence of the th ring of the shield machine; N represents the total number of elements in the low-frequency approximation sequence of the th ring of the shield machine; Step S32: Combine the standard deviations of all rings to form a standard deviation sequence, and perform Fourier noise reduction on the standard deviation sequence. The process of Fourier noise reduction is expressed as: ； ； ； In the formula, represents the Fourier transform; represents the sequence after Fourier transform; represents truncating the sequence after Fourier transform, and all elements from the 1 / 3 length position of the sequence to the end of the sequence are assigned 0; represents the inverse Fourier transform; represents the sequence after inverse Fourier transform; Step S33, the standard deviation after noise reduction is passed through to obtain the final tool wear index .

8. The shield machine tool wear prediction method according to claim 1, characterized in that, The vibration signal uses a one-dimensional convolutional neural network to perform feature extraction; the tunneling parameters use a fully connected neural network to perform feature extraction.

9. A system for implementing the shield machine tool wear prediction method according to any one of claims 1-8, characterized in that, It includes: A signal acquisition module for collecting the tunneling parameters and vibration signals during the tunneling process of the shield machine; A principal component extraction module for performing collaborative filtering on the tunneling parameters and vibration signals based on the singular value decomposition method to capture the principal component signals related to the tool wear condition in the tunneling parameters and vibration signals; A tool wear degree data generation module for correcting the traditional tool wear empirical formula based on the work log of the shield machine in the actual engineering scenario, and generating tool wear degree data using the corrected tool wear formula; A mapping model construction module for respectively performing feature extraction on the principal component signals of the vibration signal and the tunneling parameters, fusing the extracted vibration signal features and tunneling parameter features through matrix splicing to construct a unified feature representation, then using a fully connected neural network to establish a mapping model from the fused features to the tool wear degree, generating a large amount of tool wear degree data under different tunneling parameters and vibration signals using the tool wear formula, and training the mapping model; A prediction module for, for any shield machine tool wear prediction process, inputting the fused features of the principal component signals of the vibration signal and the tunneling parameters into the mapping model and outputting the prediction result of the tool wear degree.

Images