TBM Cutter Eccentric Wear Identification Method, System and Device Based on Fault Simulation

Abstract

The present application provides a method, system and device for identifying cutter head eccentric wear of TBM based on fault simulation, which relates to the technical field of fault identification. The method includes: simulating the cutter head eccentric wear fault of the target TBM under the constraint of preset test conditions to obtain a target simulation record; extracting multi-domain features from the eccentric wear time series signal to obtain a cutter head eccentric wear feature set; introducing a feature screening mechanism to obtain a target eccentric wear feature set; obtaining output information through a cutter head eccentric wear fault identification model; judging whether the target eccentric wear failure rate in the output information reaches a predetermined failure rate threshold; if it reaches, issuing an eccentric wear fault warning instruction to perform cutter replacement treatment on the target TBM. Through the present application, the problem that the existing method cannot comprehensively detect and identify the cutter head eccentric wear fault under actual working conditions due to the influence of the coupling effect of multiple factors can be solved, the accuracy of identifying the cutter head eccentric wear fault is improved, and the working efficiency and construction safety of the roadheader are improved.

Description

Technical Field

[0001] This application relates to the technical field of fault identification, and particularly to a method, system, and device for identifying TBM hob eccentric wear based on fault simulation. Background Art

[0002] In tunnel boring engineering, a tunnel boring machine (TBM) is the main construction equipment. As a key component of the TBM, the hob directly contacts the rock mass, is responsible for breaking the rock, and pushes the tunnel forward. Its performance and condition directly affect the safety, efficiency, and quality of construction. Currently, the abnormal damage of TBM hobs is one of the key issues affecting construction safety and efficiency. Existing research mainly focuses on the normal wear of hobs, including mechanism analysis, experience summary, and model prediction based on data-driven methods. However, in engineering practice, the problem of abnormal damage of hobs, especially eccentric wear faults, has been insufficiently studied, but its harmfulness cannot be ignored. Different from the regular characteristics of the gradual consumption of the cutter ring surface material in normal wear, the abnormal damage of hobs is complex and diverse. The eccentric wear fault is one of the most common and significantly harmful fault types. Its occurrence is usually related to the damage of the hob bearing or the obstruction of foreign objects such as rock debris to the rotation of the cutter shaft, resulting in only partial contact between the cutter ring and the rock mass. As a result, the relative motion mode between the hob and the rock mass changes from rolling to sliding, and the geometric shape of the contact area between the cutter ring and the rock mass also changes. If the eccentrically worn hob is not identified and processed in time, it will not only exacerbate the wear of adjacent hobs, but may also lead to the premature failure of the hob, and even endanger the overall structure of the cutter head, causing major construction safety hazards.

[0003] In summary, in existing research, the interaction between a single eccentrically worn hob and the rock mass is mostly taken as the object, making it difficult to comprehensively consider the comprehensive influence of the number, distribution, and degree of eccentric wear of hobs on the overall load of the TBM. Moreover, the results of indoor cutting experiments are difficult to accurately reflect the characteristics of the multi-factor coupling effect in actual construction conditions. In addition, since the monitoring and recording of eccentric wear faults at the construction site have not yet formed a systematic standard, relevant data samples under actual working conditions are extremely scarce, greatly restricting the development of hob fault detection and identification technologies based on on-site data. Summary of the Invention

[0004] The purpose of this application is to provide a method, system, and device for identifying TBM hob eccentric wear based on fault simulation, so as to solve the problems in existing research that mostly take the interaction between a single eccentrically worn hob and the rock mass as the object, making it difficult to comprehensively consider the comprehensive influence of the number, distribution, and degree of eccentric wear of hobs on the overall load of the TBM, and the results of indoor cutting experiments are difficult to accurately reflect the characteristics of the multi-factor coupling effect in actual construction conditions. In addition, since the monitoring and recording of eccentric wear faults at the construction site have not yet formed a systematic standard, relevant data samples under actual working conditions are extremely scarce, greatly restricting the development of hob fault detection and identification technologies based on on-site data.

[0005] In view of the above problems, the present application provides a method, system and device for identifying TBM cutter hob eccentric wear based on fault simulation.

[0006] In a first aspect, the present application provides a method for identifying TBM cutter hob eccentric wear based on fault simulation. The method for identifying TBM cutter hob eccentric wear based on fault simulation is implemented by a system for identifying TBM cutter hob eccentric wear based on fault simulation. Among them, the method for identifying TBM cutter hob eccentric wear based on fault simulation includes: obtaining preset test conditions, and performing a cutter hob eccentric wear fault simulation on a target TBM under the constraint of the preset test conditions to obtain a target simulation record; analyzing the target simulation record to obtain an eccentric wear time series signal, and performing multi-domain feature extraction on the eccentric wear time series signal to obtain a cutter hob eccentric wear feature set; introducing a feature screening mechanism to perform screening analysis on the cutter hob eccentric wear feature set to obtain a target eccentric wear feature set; using the target eccentric wear feature set as input information of a cutter hob eccentric wear fault identification model, and obtaining output information through the cutter hob eccentric wear fault identification model; judging whether the target eccentric wear failure rate in the output information reaches a predetermined failure rate threshold; if the target eccentric wear failure rate reaches the predetermined failure rate threshold, issuing an eccentric wear fault warning instruction; performing a tool change process on the target TBM based on the eccentric wear fault warning instruction.

[0007] In a second aspect, the present application further provides a system for identifying TBM cutter hob eccentric wear based on fault simulation, which is used to execute the method for identifying TBM cutter hob eccentric wear based on fault simulation as described in the first aspect. Among them, the system for identifying TBM cutter hob eccentric wear based on fault simulation includes: an eccentric wear simulation module, which is used to obtain preset test conditions and perform a cutter hob eccentric wear fault simulation on a target TBM under the constraint of the preset test conditions to obtain a target simulation record; a feature extraction module, which is used to analyze the target simulation record to obtain an eccentric wear time series signal and perform multi-domain feature extraction on the eccentric wear time series signal to obtain a cutter hob eccentric wear feature set; a feature screening module, which is used to introduce a feature screening mechanism to perform screening analysis on the cutter hob eccentric wear feature set to obtain a target eccentric wear feature set; an eccentric wear identification module, which is used to use the target eccentric wear feature set as input information of a cutter hob eccentric wear fault identification model and obtain output information through the cutter hob eccentric wear fault identification model; a failure rate judgment module, which is used to judge whether the target eccentric wear failure rate in the output information reaches a predetermined failure rate threshold; an eccentric wear warning module, which is used to issue an eccentric wear fault warning instruction if the target eccentric wear failure rate reaches the predetermined failure rate threshold; a tool change execution module, which is used to perform a tool change process on the target TBM based on the eccentric wear fault warning instruction.

[0008] In a third aspect, the present application also provides a TBM hob uneven wear recognition device based on fault simulation, which is used to execute the TBM hob uneven wear recognition method based on fault simulation as described in the first aspect. Among them, the TBM hob uneven wear recognition device based on fault simulation includes: a cutter head, on which a plurality of grooves are distributed; a cutter holder, which is installed on the plurality of grooves of the cutter head through bolts; and a hob, which is installed on the cutter holder.

[0009] One or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0010] By obtaining preset test conditions and simulating the hob uneven wear fault of the target TBM under the constraint of the preset test conditions, a target simulation record is obtained; analyzing the target simulation record to obtain an uneven wear time series signal, and performing multi-domain feature extraction on the uneven wear time series signal to obtain a hob uneven wear feature set; introducing a feature screening mechanism to screen and analyze the hob uneven wear feature set to obtain a target uneven wear feature set; using the target uneven wear feature set as the input information of the hob uneven wear fault recognition model, and obtaining output information through the hob uneven wear fault recognition model; judging whether the target uneven wear failure rate in the output information reaches a predetermined failure rate threshold; if the target uneven wear failure rate reaches the predetermined failure rate threshold, issuing an uneven wear fault warning instruction; and performing cutter replacement processing on the target TBM based on the uneven wear fault warning instruction. That is to say, by simulating the hob uneven wear fault of the TBM under preset conditions, analyzing the simulation results to obtain an uneven wear time series signal, performing multi-domain feature extraction and screening on it, and inputting the screened features into the recognition model. If the failure rate output by the model reaches the threshold, a warning is issued and the cutter is replaced. By accurately simulating the hob uneven wear fault and effectively collecting and extracting the hob uneven wear fault feature data, and then machine learning and training to obtain an uneven wear fault recognition model, it helps to accurately judge whether the hob has an uneven wear fault and the degree of the uneven wear fault, so as to timely replace the damaged hob. Further, after accurately identifying the hob uneven wear fault, it is possible to avoid problems such as increased wear of adjacent hobs, premature failure of hobs, and even endangerment to the overall structure of the cutter head caused by the failure to detect the uneven wear fault in time, thereby prolonging the service life of the hobs and improving the utilization rate of the hobs. Furthermore, it reduces the construction interruption and equipment maintenance time caused by hob faults, significantly improves the working efficiency and construction safety of the roadheader, and provides important support for the intelligent development of TBM construction.

[0011] The above description is only an overview of the technical solution of this application. In order to understand the technical means of this application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of this application more obvious and understandable, the following specifically illustrates the specific implementation manners of this application. It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of this application, nor is it used to limit the scope of this application. Other features of this application will become easily understandable through the following specification. Brief Description of the Drawings

[0012] In order to more clearly illustrate the technical solutions in this application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only exemplary, and for those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0013] Figure 1 It is a schematic flow diagram of the method for identifying the partial wear of TBM cutters based on fault simulation in this application;

[0014] Figure 2 It is a schematic structural diagram of the system for identifying the partial wear of TBM cutters based on fault simulation in this application.

[0015] Description of the Reference Numerals:

[0016] Partial wear simulation module 11, feature extraction module 12, feature screening module 13, partial wear identification module 14, failure rate judgment module 15, partial wear warning module 16, cutter change execution module 17. Detailed Description of the Embodiments

[0017] The present application provides a method, system and device for identifying TBM cutter hob eccentric wear based on fault simulation, which solves the problems in existing research that mostly focuses on the interaction between a single eccentric wear cutter hob and rock mass, making it difficult to comprehensively consider the comprehensive influence of the number, distribution and degree of eccentric wear of cutter hobs on the overall load of TBM, and the indoor cutting test results are difficult to accurately reflect the characteristics of multi-factor coupling in actual construction conditions. In addition, since the monitoring and recording of eccentric wear faults at the construction site have not yet formed a systematic standard, the relevant data samples under actual working conditions are extremely scarce, which greatly limits the development of cutter hob fault detection and identification technologies based on on-site data. By accurately simulating cutter hob eccentric wear faults and effectively collecting and extracting cutter hob eccentric wear fault characteristic data, and then machine learning and training to obtain an eccentric wear fault identification model, it helps to accurately judge whether a cutter hob has an eccentric wear fault and the degree of the eccentric wear fault, so as to replace the damaged cutter hob in time. Further, after accurately identifying the cutter hob eccentric wear fault, it is possible to avoid problems such as increased wear of adjacent cutter hobs, premature failure of cutter hobs and even endangerment to the overall structure of the cutter head caused by failure to detect the eccentric wear fault in time, thereby prolonging the service life of cutter hobs and improving the utilization rate of cutter hobs. Furthermore, it reduces construction interruptions and equipment maintenance time caused by cutter hob faults, significantly improves the working efficiency and construction safety of roadheaders, and provides important support for the intelligent development of TBM construction.

[0018] Next, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application. Additionally, it should be noted that for the sake of description, only the parts related to the present application are shown in the drawings rather than all of them.

[0019] Embodiment 1, please refer to the attached Figure 1 , the present application provides a method for identifying TBM cutter hob eccentric wear based on fault simulation. Among them, the method is applied to a system for identifying TBM cutter hob eccentric wear based on fault simulation. The method for identifying TBM cutter hob eccentric wear based on fault simulation specifically includes the following steps:

[0020] Step P10: Obtain preset test conditions, and simulate the cutter hob eccentric wear fault of the target TBM under the constraint of the preset test conditions to obtain a target simulation record;

[0021] Specifically, the preset test conditions are first determined. These conditions set a controllable environment for the simulation experiment, including geological conditions, tunneling parameters, and the initial state of the cutters, etc. Under these conditions, the cutter eccentric wear fault simulation is carried out on the target TBM. During the simulation process, the cutters operate according to the set working conditions to reproduce the working state in actual construction. Through the monitoring equipment and data acquisition system, the running parameters and performance of the cutters during the simulation process are recorded in real time to form the target simulation record. These records contain in detail information such as cutter force, wear, and tunneling speed, providing a basis for subsequent analysis. Analyzing these simulation records can evaluate the performance of the cutters under different working conditions, and then optimize the TBM design and construction plan to improve construction efficiency and safety.

[0022] Step P20: Analyze the target simulation record to obtain the eccentric wear time series signal, and perform multi-domain feature extraction on the eccentric wear time series signal to obtain the cutter eccentric wear feature set;

[0023] Specifically, analyzing the target simulation record to obtain the eccentric wear time series signal and performing multi-domain feature extraction on this time series to obtain the eccentric wear feature set. This process aims to deeply mine the key information in the simulation data and provide a solid foundation for subsequent fault identification.

[0024] First, carefully analyze the target simulation record, which contains various operation data of the cutters during the simulation process. Through professional data analysis methods, the time series signal related to cutter eccentric wear is extracted from these data. This time series reflects the change of the eccentric wear state of the cutters during the simulation process and is the key clue for analyzing the cutter performance. Then, perform multi-domain feature extraction on the extracted eccentric wear time series signal. This means not only analyzing the features such as amplitude and frequency of the signal from the time domain perspective, but also expanding to multiple dimensions such as the frequency domain and time-frequency domain to comprehensively capture the characteristics of the signal. In the time domain, the periodic and trend changes of the signal can be observed; in the frequency domain, the frequency distribution of the signal can be analyzed to understand the proportion of different frequency components; time-frequency domain analysis can consider both time and frequency factors to more accurately describe the dynamic characteristics of the signal. Through these multi-domain analyses, a series of features related to eccentric wear are collected, and these features together form the eccentric wear feature set. Among them, each feature reflects the characteristics of cutter eccentric wear from different aspects and provides rich input information for the subsequent fault identification model. Therefore, this process is a key step in achieving accurate fault identification. By comprehensively collecting multi-domain features, the state of cutter eccentric wear can be more accurately characterized, improving the accuracy and reliability of fault identification.

[0025] To sum up, by analyzing the target simulation record to obtain the eccentric wear time series signal and performing multi-domain feature extraction to form the cutter eccentric wear feature set, it lays a foundation for the accurate identification of cutter eccentric wear faults, which helps to improve the safety and efficiency of TBM construction.

[0026] Step P30: Introduce a feature screening mechanism to screen and analyze the hob wear feature set to obtain a target wear feature set;

[0027] Specifically, after obtaining a wear feature set containing rich information, in order to improve the efficiency and accuracy of subsequent analysis, it is necessary to screen these features. This process is achieved by introducing a feature screening mechanism, aiming to select the most representative and influential features for hob wear fault recognition from a large number of features, so as to form a target wear feature set and provide key inputs for establishing an efficient and accurate fault recognition model.

[0028] First, conduct a comprehensive review of the collected wear feature set. These feature sets cover various parameters obtained from multi-domain analysis. Although theoretically all related to hob wear, in practical applications, some features may contribute little to fault recognition and may even introduce noise, affecting the performance of the model. Therefore, an effective screening mechanism is needed to distinguish the importance of these features. Then, use methods such as statistical analysis and machine learning algorithms to screen and analyze the wear feature set. For example, the correlation coefficient between features and wear faults can be calculated to screen out features highly correlated with faults; or feature importance evaluation methods, such as the feature importance ranking based on decision tree models, can be used to determine which features play a key role in predicting wear faults. In addition, dimensionality reduction techniques, such as principal component analysis (PCA), can be adopted to remove redundant features and retain the main components, further streamlining the feature set. Through these screening and analysis steps, irrelevant or redundant features can be effectively removed, leaving features truly valuable for fault recognition. Finally, the screened features are combined to form a target wear feature set. This feature set is more refined, highlighting features closely related to hob wear faults, providing high-quality data support for subsequent construction of a fault recognition model, and helping to improve the recognition accuracy and generalization ability of the model.

[0029] In summary, introducing a feature screening mechanism to screen and analyze the wear feature set to obtain a target wear feature set is crucial for improving the efficiency and accuracy of hob wear fault recognition, ensuring that the fault recognition model is more reliable and effective in practical applications.

[0030] Step P40: Use the target wear feature set as the input information of the hob wear fault recognition model and obtain output information through the hob wear fault recognition model;

[0031] Specifically, inputting the target partial wear feature set into the hob partial wear fault recognition model to obtain the output information of the model is a key link in realizing the accurate diagnosis of hob partial wear faults. The aim is to transform the feature set into a result with clear diagnostic significance through the analysis and processing of the model, so as to provide decision-making support for the maintenance and operation of the TBM.

[0032] First, prepare the screened target partial wear feature set. These feature sets are refined through in-depth analysis of simulation records and multi-domain feature extraction in the early stage, and then through a feature screening mechanism. They contain information crucial for hob partial wear fault recognition. Then, take these feature sets as input information and input them into a specially constructed hob partial wear fault recognition model. This model is trained based on a large amount of sample data and can recognize and understand the complex relationship between the input features and hob partial wear faults. Inside the model, the input feature set will go through a series of calculation and analysis processes, including but not limited to the evaluation of feature weights, pattern matching, and calculation of fault probabilities, etc. Through these complex operations, the model can comprehensively evaluate the partial wear state of the hob. Next, the model will generate output information, which is usually the diagnostic result of the hob partial wear fault, and may include the probability of the fault occurring, the severity level of the fault, or the specific fault type, etc. These output information are the interpretation and judgment of the model on the input feature set and provide a direct basis for subsequent decision-making. Therefore, this process is a key step in transforming abstract feature data into specific and practical diagnostic results, making fault recognition more scientific and accurate through the intelligent analysis of the model.

[0033] To sum up, by inputting the target partial wear feature set into the hob partial wear fault recognition model and obtaining the output information, the transformation from feature data to fault diagnostic results is realized, providing strong technical support for the maintenance and management of TBM hobs, helping to detect and handle hob partial wear faults in a timely manner, and ensuring the smooth progress of tunnel boring projects.

[0034] Step P50: Judge whether the target partial wear fault rate in the output information reaches a predetermined fault rate threshold;

[0035] Step P60: If the target partial wear fault rate reaches the predetermined fault rate threshold, issue a partial wear fault warning instruction;

[0036] Step P70: Perform cutter replacement processing on the target TBM based on the partial wear fault warning instruction.

[0037] Specifically, in the hob eccentric wear fault identification process, first, the target eccentric wear failure rate in the output information of the hob eccentric wear fault identification model is judged to see if it reaches the pre-set failure rate threshold. This threshold is determined based on actual engineering experience and safety standards and is used to distinguish between normal wear and fault states of the hob. If the target eccentric wear failure rate does not reach the threshold, it is considered that the hob is currently within an acceptable wear range and no special measures need to be taken; conversely, if it reaches or exceeds the threshold, it means that the hob may have a relatively serious eccentric wear fault and further processing is required. Then, when it is confirmed that the target eccentric wear failure rate reaches the pre-set failure rate threshold, an eccentric wear fault warning instruction is issued. This warning instruction is a signal used to notify the operator or the automated control system that there is a potential fault risk with the hob. The issuance of the warning instruction is the starting signal for the fault handling process, triggering a series of subsequent maintenance operations to ensure that the fault can be responded to in a timely manner. Finally, based on the eccentric wear fault warning instruction, tool changing treatment is carried out on the target TBM. Tool changing treatment is an actual maintenance operation, the purpose of which is to replace the severely worn or faulty hob with a new one to restore the normal tunneling ability of the TBM. This treatment measure is part of preventive maintenance. By replacing the faulty hob in a timely manner, the further expansion of the fault can be avoided, the impact on the overall performance of the TBM can be reduced, and the continuity and safety of the tunnel boring project can be ensured.

[0038] In summary, from the judgment of the failure rate to the issuance of the warning instruction and then to the final tool changing treatment, each link is closely connected, ensuring the stable operation of the TBM during tunneling and effectively reducing the construction risks caused by hob eccentric wear faults.

[0039] Furthermore, the present application includes the following steps:

[0040] Take sandy soil as the upper stratum;

[0041] Take broken soft rock as the lower stratum;

[0042] Form a composite stratum based on the upper stratum and the lower stratum, and take the composite stratum as the preset geological condition;

[0043] Obtain the preset fault ratio;

[0044] Obtain the preset tunneling mode;

[0045] Form the preset test condition based on the preset geological condition, the preset fault ratio, and the preset tunneling mode.

[0046] Specifically, first, sandy soil is set as the upper stratum, which is a common stratum material with certain water permeability and low bearing capacity. Then, broken soft rock is used as the lower stratum. This kind of rock is relatively easy to break and, together with the sandy soil layer, constitutes a stratum structure with specific engineering characteristics. Then, a composite stratum is constructed based on these two strata. This composite stratum simulates the complex geological environment that may be encountered in actual tunnel boring and is used as the preset geological condition, providing a background close to the actual working conditions for subsequent tests. For example, the geological condition for the eccentric wear fault simulation test is a composite stratum with sandy soil in the upper layer and broken soft rock in the lower layer. The data acquisition system records parameters such as cutterhead speed, propulsion speed, cutterhead torque, and thrust at a sampling frequency of 1 Hz. The specific process of the test includes: first, filling the upper soft and lower hard composite stratum in the preparatory rock box, and then adjusting the test bench and assembling the cutterhead. When assembling the cutterhead, faulty hob cutters are arranged according to different set fault ratios. After completing these steps, the console selects the full-face tunneling mode and sets the cutterhead speed and propulsion speed.

[0047] Obtain the preset fault ratio, which is pre-determined according to research purposes or engineering experience and is used to simulate different degrees of hob eccentric wear fault conditions in the test to study the influence of different fault degrees on the tunneling performance of the TBM. For example, the test groups of the simulation test include different eccentric wear fault rates, with ratios of 0%, 25%, 50%, and 75% respectively. At the same time, obtain the preset tunneling mode. The tunneling mode includes the setting of parameters such as cutterhead speed and propulsion speed, which directly affect the tunneling efficiency of the TBM and the force on the hob cutters.

[0048] Finally, comprehensively considering the preset geological condition, preset fault ratio, and preset tunneling mode, complete preset test conditions are formed. These conditions together constitute a control variable, ensuring that the test is carried out under a unified standard, making the test results comparable and reliable, and providing an accurate test basis for studying hob eccentric wear faults and their influence on the tunneling performance of the TBM.

[0049] Furthermore, the present application includes the following steps:

[0050] Extract the thrust simulation time series signal from the target simulation record and use the thrust simulation time series signal as the eccentric wear time series signal;

[0051] Read the predetermined time domain features and perform time domain feature extraction on the eccentric wear time series signal based on the predetermined time domain features to obtain the first time domain feature set;

[0052] Read the predetermined frequency domain features and perform frequency domain feature extraction on the first eccentric wear signal spectrum obtained by performing fast Fourier transform on the eccentric wear time series signal based on the predetermined frequency domain features to obtain the first frequency domain feature set;

[0053] Retrieve a predetermined time-frequency domain feature extraction strategy, and obtain a first time-frequency domain feature set of the partial wear time series signal according to the predetermined time-frequency domain feature extraction strategy;

[0054] Based on the first time domain feature set, the first frequency domain feature set, and the first time-frequency domain feature set, form the hob partial wear feature set.

[0055] Specifically, first, extract the thrust simulation time series signal from the target simulation record and use it as a representative of the partial wear time series signal. This is because the thrust signal can directly reflect the force condition of the hob during tunneling and is an important basis for evaluating the partial wear state of the hob. Then, read the predetermined time domain features. The time domain features include mean value, standard deviation, root mean square, root amplitude, peak value, skewness, kurtosis, waveform factor, peak factor, pulse factor, margin factor, and energy, which can provide information about the overall trend, dispersion degree, periodicity, and morphology of the parameters, that is, can reflect the basic characteristics of the signal in the time dimension. Based on these predetermined time domain features, perform time domain feature extraction on the partial wear time series signal to obtain the first time domain feature set, which provides key information in the time series for subsequent analysis. Then, read the predetermined frequency domain features, perform a fast Fourier transform on the partial wear time series signal to convert it from the time domain to the frequency domain, and obtain the first partial wear signal spectrum. Based on the predetermined frequency domain features, perform frequency domain feature extraction on this spectrum to obtain the first frequency domain feature set, providing data support for analyzing the frequency characteristics of the signal. Exemplarily, the frequency domain features include center frequency, average frequency, frequency standard deviation, and root mean square frequency, which help to understand the frequency distribution of the signal.

[0056] In addition, retrieve a predetermined time-frequency domain feature extraction strategy. This strategy comprehensively considers the correlation characteristics of the signal in the time domain and the frequency domain. According to this strategy, obtain the first time-frequency domain feature set of the partial wear time series signal. This feature set integrates information in both the time and frequency dimensions and can more comprehensively describe the characteristics of the signal. Exemplarily, the time-frequency domain features are obtained by performing two-layer wavelet packet decomposition on the tunneling parameters using the db4 wavelet basis function, obtaining four sub-band signals, and extracting the average energy of the wavelet packet coefficients of each sub-band as feature parameters, and then obtaining four time-frequency domain features.

[0057] Finally, based on the first time domain feature set, the first frequency domain feature set, and the first time-frequency domain feature set, integrate the features in these three sets to form a complete partial wear feature set. This partial wear feature set contains features extracted from different angles, providing a rich and comprehensive data basis for subsequent fault identification and analysis, and helping to more accurately identify and evaluate the partial wear fault state of the hob.

[0058] Furthermore, the present application includes the following steps:

[0059] Perform two - layer wavelet packet decomposition on the partial wear time - series signal based on the db4 wavelet basis function to obtain the decomposition result of the partial wear signal, where the decomposition result of the partial wear signal includes four sub - band signals;

[0060] Extract any sub - band signal from the four sub - band signals, and obtain the average energy of any wavelet packet coefficients of the any sub - band signal;

[0061] Construct the first time - domain and frequency - domain feature set based on the average energy of the any wavelet packet coefficients.

[0062] Specifically, first, perform two - layer wavelet packet decomposition on the partial wear time - series signal based on the db4 wavelet basis function. This process decomposes the signal into multiple sub - band signals, specifically obtaining four sub - band signals. These four sub - band signals respectively represent the characteristics of the signal in different frequency bands, providing a basis for subsequent feature extraction. Then, extract any one sub - band signal from these four sub - band signals, and calculate the average energy of any wavelet packet coefficients of this sub - band signal. The average energy of wavelet packet coefficients is obtained by calculating the sum of the squares of each wavelet packet coefficient in the sub - band signal and then taking its average, which reflects the energy distribution of the signal in this frequency band. Finally, construct the first time - domain and frequency - domain feature set based on the calculated average energy of any wavelet packet coefficients. This feature set integrates time - domain and frequency - domain information, can more comprehensively describe the characteristics of the partial wear signal, and provides important data support for subsequent fault identification and analysis.

[0063] Furthermore, this application includes the following steps:

[0064] Extract the torque analog time - series signal in the target analog record, and use the torque analog time - series signal as the partial wear time - series signal;

[0065] Obtain the second time - domain feature set of the partial wear time - series signal based on the predetermined time - domain feature extraction;

[0066] Obtain the second frequency - domain feature set of the second partial wear signal spectrum obtained by performing fast Fourier transform on the partial wear time - series signal based on the predetermined frequency - domain feature extraction;

[0067] Obtain the second time - domain and frequency - domain feature set of the partial wear time - series signal according to the predetermined time - domain and frequency - domain feature extraction strategy;

[0068] Add the second time - domain feature set, the second frequency - domain feature set, and the second time - domain and frequency - domain feature set to the hob partial wear feature set.

[0069] Specifically, first, extract the torque simulation time series signal from the target simulation record and use it as another important data source for the partial wear time series signal. The torque time series signal can reflect the change of the torsional moment borne by the hob during tunneling and is one of the key indicators for evaluating the partial wear state of the hob. Next, based on the predetermined time-domain features, extract the time-domain features of the torque simulation time series signal. These features may include the mean, variance, peak value, etc. of the signal, so as to obtain the second time-domain feature set, which provides detailed information in the time series for analyzing the partial wear state of the hob. Then, based on the predetermined frequency-domain features, perform a fast Fourier transform on the torque simulation time series signal to convert it into a frequency-domain signal and obtain the second partial wear signal spectrum. By extracting the frequency-domain features of this spectrum, such as calculating the center frequency and bandwidth of the spectrum, the second frequency-domain feature set is obtained, which further enriches the description of the partial wear state of the hob, especially the information in terms of frequency distribution. In addition, according to the predetermined time-domain and frequency-domain feature extraction strategy, perform a comprehensive analysis on the torque simulation time series signal to obtain the second time-domain and frequency-domain feature set. This feature set combines the features of the time domain and the frequency domain and can more comprehensively reflect the changes of the torque signal at different times and frequencies, providing richer data dimensions for fault identification.

[0070] Finally, integrate the second time-domain feature set, the second frequency-domain feature set, and the second time-domain and frequency-domain feature set into the existing partial wear feature set. In this way, the partial wear feature set not only contains the relevant features of the thrust signal but also adds the multi-dimensional features of the torque signal, making the entire feature set more comprehensive and rich, providing more powerful data support for the subsequent partial wear fault identification and analysis of the hob, and helping to improve the accuracy and reliability of fault diagnosis.

[0071] Furthermore, the present application includes the following steps: The feature screening mechanism refers to the mechanism for performing correlation screening on the partial wear feature set based on a predetermined correlation analysis method, and the predetermined correlation analysis method includes the pearson correlation analysis method and the copula entropy theory analysis method.

[0072] Specifically, first, the core of the feature screening mechanism lies in using the predetermined correlation analysis method to screen the partial wear feature set. This process aims to identify and retain the features most relevant to the partial wear fault diagnosis of the hob, thereby improving the efficiency and accuracy of the fault identification model. Specifically, the predetermined correlation analysis method covers the pearson correlation analysis method and the copula entropy theory analysis method, and these two methods evaluate the degree of association between features and faults from different perspectives.

[0073] Next, the Pearson correlation analysis method measures the linear correlation between features and the partial wear failure rate by calculating the Pearson correlation coefficient between them. The value range of the correlation coefficient is between -1 and 1. The closer the value is to 1 or -1, the stronger the linear relationship between the feature and the failure rate; the closer the value is to 0, the weaker the linear relationship. Through this method, features with a strong linear correlation with the partial wear failure rate can be screened out, providing direct linear clues for fault diagnosis. Among them, Pearson correlation analysis calculates the Pearson correlation coefficient between the extracted features and the partial wear failure rate. The results show that there is a strong linear correlation between the frequency domain and time-frequency domain features and the fault partial wear rate, and the correlation coefficient values all exceed 0.5. However, for 12 time domain features such as skewness, kurtosis, waveform factor, peak factor, margin factor, and impulse factor, the absolute value of their correlation coefficient is less than 0.2 or close to 0.2, indicating that the linear relationship between these features and the partial wear failure rate is weak or even non-existent.

[0074] In addition, the copula entropy theory analysis method focuses on evaluating the non-linear correlation between features and faults. It reveals the complex non-linear relationship between the two by calculating the copula value of the feature and the fault state. This method can capture non-linear associations that may be overlooked in Pearson correlation analysis, thus providing a more comprehensive understanding of the internal connection between features and faults. Among them, Copula entropy theory calculates the Copula value of the extracted features and the partial wear failure rate. The results show that there is a complex non-linear relationship between the features in the frequency domain, time-frequency domain, and time domain and the fault state, and only the waveform factor, a time domain feature, still shows a very low level of non-linear correlation. Therefore, the finally screened effective features can be used to identify the hob partial wear failure rate.

[0075] Finally, by comprehensively applying these two predetermined correlation analysis methods, a comprehensive screening of the partial wear feature set is carried out. Retain those features that show a strong linear correlation in Pearson correlation analysis and significant non-linear correlation in copula entropy theory analysis to form the target partial wear feature set. This target feature set is refined and efficient, laying a solid foundation for constructing an accurate hob partial wear fault recognition model and helping to improve the sensitivity and accuracy of the model for fault diagnosis.

[0076] Furthermore, this application includes the following steps:

[0077] Match the partial wear failure rate corresponding to the partial wear time series signal in the target simulation record;

[0078] Based on the Stacking ensemble learning principle, perform ensemble learning training on the training data set formed according to the partial wear feature set and the partial wear failure rate to obtain the hob partial wear fault recognition model.

[0079] Specifically, first, the partial wear failure rate corresponding to the partial wear timing signal is accurately matched in the target simulation record. This process is achieved by analyzing the data in the simulation record to find the failure rate index that matches the partial wear timing signal, which directly reflects the degree of hob partial wear. Then, based on the Stacking ensemble learning principle, which is an advanced machine learning method that combines the advantages of multiple different models to improve the accuracy and stability of prediction. Specifically, the partial wear feature set and the corresponding partial wear failure rate are used to form training data groups, which contain rich feature information and accurate fault labels, providing a solid foundation for model training. Then, these training data groups are trained using ensemble learning. During the training process, the Stacking method first trains multiple base models, such as decision trees, support vector machines, etc., and each model learns the patterns in the data from different perspectives. The prediction results of these base models are then used as new features and input into a meta-model, which is usually a simple model such as logistic regression. It is responsible for integrating the prediction results of the base models and outputting the final fault identification result. In this way, Stacking ensemble learning can make full use of the advantages of each base model while reducing the bias and variance problems that may exist in a single model.

[0080] Finally, after the ensemble learning training, a hob partial wear fault identification model is obtained. This model can accurately predict the partial wear failure rate of the hob based on the input partial wear feature set, providing scientific decision-making support for the maintenance and operation of the TBM. Through this training method based on the Stacking ensemble learning principle, the obtained model not only has high prediction accuracy but also can better adapt to different working conditions and data changes, improving the practicality and reliability of the model.

[0081] Furthermore, the present application includes the following steps:

[0082] Extract the first base model from the predetermined base models;

[0083] Perform machine learning on the first data group in the training data group using the first base model optimized by hyperparameters to obtain the first target base model;

[0084] Obtain a predetermined meta-model and integrate it with the first target base model to obtain the hob partial wear fault identification model;

[0085] Wherein, the predetermined base models at least include a K-nearest neighbor model, a random forest model, a support vector classification model, and a multi-layer perceptron model.

[0086] Specifically, first, extract the first base model from a predetermined set of base models. These base models are the foundation for constructing an ensemble learning model, including K-nearest neighbor models, random forest models, support vector classification models, and multi-layer perceptron models, etc. Each of these models has its own unique algorithms and advantages, and can analyze and predict data from different perspectives.

[0087] Next, perform hyperparameter optimization on the extracted first base model. Hyperparameter optimization is to adjust the parameter settings of the model, such as the number of neighbors in the K-nearest neighbor model, the number and depth of trees in the random forest model, etc., to improve the performance of the model. The optimized first base model can better adapt to the training data and improve the accuracy of prediction. Then, use the optimized first base model to perform machine learning on the first data set in the training data set. The training process includes hyperparameter optimization of the base model. The hyperparameters of the KNN model only include one parameter, that is, the number of neighbors (n_neighbors). When the number of neighbors is 7, the classification accuracy of the KNN model reaches the highest. For the RF model, the best performance is obtained by setting the number of decision trees (n_estimators) to 40 and the maximum depth of each tree (max_depth) to 4. For the SVC model, the best combination is found by adjusting the penalty parameter (C) and kernel width parameter (gamma) of the Gaussian kernel. When the penalty parameter is set to 1000 and the kernel width parameter is set to 0.001, the accuracy of the SVC model reaches the optimum. In the configuration of the MLP model with a single-layer and double-layer hidden layer, when the number of neurons is set to 60 and Adam is selected as the solver, the accuracy reaches 88.16%. On the other hand, in the double-hidden layer configuration, when the number of neurons is set to (40, 40) and the Adam solver is used, the accuracy slightly decreases to 88.05%, slightly lower than the accuracy of the single-hidden layer configuration. Logistic regression has one hyperparameter, that is, the penalty coefficient (CLR). When the value of CLR is set to 1.0, the model achieves the best classification performance, and the accuracy reaches 91.69%. This process involves the model's learning and training of the data, aiming to enable the model to identify patterns and regularities in the data, so as to accurately predict unknown data. Through this learning process, the first target base model is obtained, and this model already has the ability to predict a specific data set.

[0088] Next, obtain a predetermined meta-model. The meta-model is an advanced model in ensemble learning, responsible for integrating the outputs of each base model and making the final decision. For example, the meta-model is selected to use a logistic regression model (LR). The selection and design of the meta-model are crucial for the effect of ensemble learning.

[0089] Finally, combine the first target base model with the meta-model to integrate and obtain the final hob wear fault identification model. This model integrates the advantages of multiple base models and, through the coordination of the meta-model, achieves high-precision identification of hob wear faults. This integration method not only improves the model's prediction ability but also enhances the model's stability and generalization ability, enabling it to be better applied to actual engineering scenarios and providing reliable fault warning and diagnostic support for the maintenance and operation of TBMs.

[0090] In summary, the TBM hob wear identification method based on fault simulation provided by this application has the following technical effects:

[0091] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0092] By obtaining preset test conditions and simulating the hob wear fault of the target TBM under the constraints of the preset test conditions to obtain a target simulation record; analyzing the target simulation record to obtain a wear time series signal, and performing multi-domain feature extraction on the wear time series signal to obtain a hob wear feature set; introducing a feature screening mechanism to screen and analyze the hob wear feature set to obtain a target wear feature set; using the target wear feature set as the input information of the hob wear fault identification model, and obtaining output information through the hob wear fault identification model; determining whether the target wear failure rate in the output information reaches a predetermined failure rate threshold; if the target wear failure rate reaches the predetermined failure rate threshold, issuing a wear fault warning instruction; performing cutter replacement processing on the target TBM based on the wear fault warning instruction. That is to say, by simulating the hob wear fault of the TBM under preset conditions, analyzing the simulation results to obtain a wear time series signal, performing multi-domain feature extraction and screening on it, inputting the screened features into the identification model, if the failure rate output by the model reaches the threshold, then issuing a warning and replacing the cutter. By accurately simulating the hob wear fault and effectively collecting and extracting the hob wear fault feature data, and then machine learning and training to obtain a wear fault identification model, it helps to accurately judge whether the hob has a wear fault and the degree of the wear fault, so as to replace the damaged hob in time. Further, after accurately identifying the hob wear fault, it is possible to avoid problems such as increased wear of adjacent hobs, premature failure of hobs, and even endangerment of the overall structure of the cutterhead caused by the failure to detect the wear fault in time, thereby extending the service life of the hob and improving the utilization rate of the hob. Furthermore, it reduces the construction interruption and equipment maintenance time caused by hob faults, significantly improves the working efficiency and construction safety of the roadheader, and provides important support for the intelligent development of TBM construction.

[0093] Embodiment 2. Based on the same inventive concept as the TBM cutter head eccentric wear identification method based on fault simulation in the foregoing embodiment, the present application further provides a TBM cutter head eccentric wear identification system based on fault simulation. Please refer to the appendix Figure 2 , the TBM cutter head eccentric wear identification system based on fault simulation includes:

[0094] An eccentric wear simulation module 11, configured to obtain preset test conditions, and perform a cutter head eccentric wear fault simulation on a target TBM under the constraint of the preset test conditions to obtain a target simulation record;

[0095] A feature extraction module 12, configured to analyze the target simulation record to obtain an eccentric wear time series signal, and perform multi-domain feature extraction on the eccentric wear time series signal to obtain a cutter head eccentric wear feature set;

[0096] A feature screening module 13, configured to introduce a feature screening mechanism to perform screening analysis on the cutter head eccentric wear feature set to obtain a target eccentric wear feature set;

[0097] An eccentric wear identification module 14, configured to use the target eccentric wear feature set as input information of a cutter head eccentric wear fault identification model, and obtain output information through the cutter head eccentric wear fault identification model;

[0098] A failure rate judgment module 15, configured to judge whether a target eccentric wear failure rate in the output information reaches a predetermined failure rate threshold;

[0099] An eccentric wear warning module 16, configured to issue an eccentric wear fault warning instruction if the target eccentric wear failure rate reaches the predetermined failure rate threshold;

[0100] A tool change execution module 17, configured to perform a tool change process on the target TBM based on the eccentric wear fault warning instruction.

[0101] Furthermore, the eccentric wear simulation module 11 in the TBM cutter head eccentric wear identification system based on fault simulation is further configured to:

[0102] Use sandy soil as the upper formation;

[0103] Use fragmented soft rock as the lower formation;

[0104] Form a composite formation based on the upper formation and the lower formation, and use the composite formation as the preset geological condition;

[0105] Obtain a preset fault ratio;

[0106] Obtain a preset tunneling mode;

[0107] Form the preset test conditions based on the preset geological condition, the preset fault ratio, and the preset tunneling mode.

[0108] Furthermore, the feature extraction module 12 in the TBM hob wear recognition system based on fault simulation is further configured to:

[0109] Extract the thrust simulation time series signal from the target simulation record, and use the thrust simulation time series signal as the wear time series signal;

[0110] Read the predetermined time domain features, and perform time domain feature extraction on the wear time series signal based on the predetermined time domain features to obtain a first time domain feature set;

[0111] Read the predetermined frequency domain features, and perform frequency domain feature extraction on the first wear signal spectrum obtained by performing fast Fourier transform on the wear time series signal based on the predetermined frequency domain features to obtain a first frequency domain feature set;

[0112] Retrieve the predetermined time domain and frequency domain feature extraction strategy, and obtain a first time domain and frequency domain feature set of the wear time series signal according to the predetermined time domain and frequency domain feature extraction strategy;

[0113] Based on the first time domain feature set, the first frequency domain feature set and the first time domain and frequency domain feature set, form the hob wear feature set.

[0114] Furthermore, the feature extraction module 12 in the TBM hob wear recognition system based on fault simulation is further configured to:

[0115] Perform two-layer wavelet packet decomposition on the wear time series signal based on the db4 wavelet basis function to obtain a wear signal decomposition result, where the wear signal decomposition result includes four sub-band signals;

[0116] Extract any sub-band signal from the four sub-band signals, and obtain the average energy of any wavelet packet coefficient of the any sub-band signal;

[0117] Form the first time domain and frequency domain feature set based on the average energy of the any wavelet packet coefficient.

[0118] Furthermore, the feature extraction module 12 in the TBM hob wear recognition system based on fault simulation is further configured to:

[0119] Extract the torque simulation time series signal from the target simulation record, and use the torque simulation time series signal as the wear time series signal;

[0120] Obtain a second time domain feature set of the wear time series signal based on the extraction of the predetermined time domain features;

[0121] Obtain a second frequency domain feature set of the second wear signal spectrum obtained by performing fast Fourier transform on the wear time series signal based on the extraction of the predetermined frequency domain features;

[0122] Obtain the second time-frequency domain feature set of the partial wear time series signal according to the predetermined time-frequency domain feature extraction strategy;

[0123] Add the second time domain feature set, the second frequency domain feature set, and the second time-frequency domain feature set to the hob partial wear feature set.

[0124] Furthermore, the feature screening module 13 in the TBM hob partial wear recognition system based on fault simulation is further configured to, the feature screening mechanism refers to a mechanism for performing correlation screening on the hob partial wear feature set based on a predetermined correlation analysis method, and the predetermined correlation analysis method includes the pearson correlation analysis method and the copula entropy theory analysis method.

[0125] Furthermore, the partial wear recognition module 14 in the TBM hob partial wear recognition system based on fault simulation is further configured to:

[0126] Match the partial wear failure rate corresponding to the partial wear time series signal in the target simulation record;

[0127] Perform ensemble learning training on the training data group formed according to the partial wear feature set and the partial wear failure rate based on the Stacking ensemble learning principle to obtain the hob partial wear failure recognition model.

[0128] Furthermore, the partial wear recognition module 14 in the TBM hob partial wear recognition system based on fault simulation is further configured to:

[0129] Extract the first base model in the predetermined base model;

[0130] Perform machine learning on the first data group in the training data group through the first base model after hyperparameter optimization to obtain the first target base model;

[0131] Obtain a predetermined meta-model, and integrate it with the first target base model to obtain the hob partial wear failure recognition model;

[0132] Wherein, the predetermined base model at least includes a K-nearest neighbor model, a random forest model, a support vector classification model, and a multi-layer perceptron model.

[0133] The various embodiments in this specification are described in a progressive manner, and the key point of each embodiment is the difference from other embodiments. The foregoing Figure 1The method and specific example for identifying TBM cutter hob eccentric wear based on fault simulation in Embodiment 1 are equally applicable to the system for identifying TBM cutter hob eccentric wear based on fault simulation in this embodiment. Through the detailed description of the method for identifying TBM cutter hob eccentric wear based on fault simulation above, those skilled in the art can clearly know the system for identifying TBM cutter hob eccentric wear based on fault simulation in this embodiment. Therefore, for the sake of brevity of the specification, it will not be described in detail here. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0134] Embodiment 3. Based on the same inventive concept as the method for identifying TBM cutter hob eccentric wear based on fault simulation in the foregoing embodiments, the present application also provides a device for identifying TBM cutter hob eccentric wear based on fault simulation, which is used to execute the steps of the method for identifying TBM cutter hob eccentric wear based on fault simulation described in any one of Embodiment 1 above. Among them, the device for identifying TBM cutter hob eccentric wear based on fault simulation includes: a cutter head, on which a plurality of grooves are distributed; a cutter holder, which is installed on the plurality of grooves of the cutter head through bolts; a hob, which is installed on the cutter holder. There are 15 hob cutter holder grooves distributed on the TBM cutter head. 12 cutter holders inside install 12 main hobs, arranged at a 60-degree angle. 3 cutter holders on the outer ring of the cutter head install 3 side hobs, arranged at a 120-degree angle and inclined 10 degrees to the horizontal plane. The installation radii of different hobs are different. For the convenience of description, the 15 hobs on the cutter head are numbered. The numbers 1 to 12 are main hobs, and the numbers 13 to 15 are side hobs. In addition, the cutter head is also equipped with 3 groups of scrapers, arranged at a 120-degree angle, with 8 in each group, and installed on the edge area of the hob opening through bolts. In addition, the cylindrical end of the hob shaft of a normal hob is designed as a square structure, and the corresponding cutter holder of the hob is also modified accordingly to realize the trapped state of the hob. When the hob just enters the trapped state, the cutter ring has not been worn. On the basis of the trapped hob, different degrees of cutter ring thickness are gradually cut off to simulate different eccentric wear states.

[0135] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0136] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is also intended to include these changes and variations.

Claims

1. A TBM cutter eccentric wear identification method based on fault simulation, characterized in that: The method comprises: Acquire preset test conditions, and perform a hob cutter eccentric wear fault simulation on a target TBM under the constraints of the preset test conditions to obtain a target simulation record; Analyze the target simulation record to obtain an eccentric wear timing signal, and perform multi-domain feature extraction on the eccentric wear timing signal to obtain a hob eccentric wear feature set; A feature screening mechanism is introduced to screen and analyze the hob eccentric wear feature set to obtain a target eccentric wear feature set; Using the target eccentric wear feature set as input information of a roller cutter eccentric wear fault identification model, and obtaining output information through the roller cutter eccentric wear fault identification model; Determining whether the target eccentric wear failure rate in the output information reaches a predetermined failure rate threshold; If the target eccentric wear failure rate reaches the predetermined failure rate threshold, issuing an eccentric wear failure warning instruction; Performing a tool change process on the target TBM based on the eccentric wear fault warning instruction; The target eccentric wear feature set is used as input information of a roller cutter eccentric wear fault identification model, and output information is obtained through the roller cutter eccentric wear fault identification model, including: Matching the eccentric wear failure rate corresponding to the eccentric wear timing signal in the target simulation record; Based on the Stacking ensemble learning principle, ensemble learning training is performed on the training data group formed according to the eccentric wear feature set and the eccentric wear failure rate to obtain the hob eccentric wear fault identification model; Wherein, based on the Stacking ensemble learning principle, an ensemble learning training is performed on the training data group formed according to the eccentric wear feature set and the eccentric wear failure rate to obtain the hob eccentric wear fault identification model, including: extracting a first base model from predetermined base models; Performing machine learning on the first data group in the training data group using the first base model after hyperparameter optimization to obtain a first target base model; Acquire a predetermined meta-model, and integrate it with the first target base model to obtain the hob cutter eccentric wear fault identification model; The predetermined base models include at least a K-nearest neighbor model, a random forest model, a support vector classification model and a multivariate perceptron model.

2. The method according to claim 1, characterized in that Obtaining preset test conditions, and simulating the hob cutter eccentric wear fault on the target TBM under the constraints of the preset test conditions, and obtaining a target simulation record, including: Use sand as the upper stratum; The crushed soft rock is taken as the underlying stratum; forming a composite stratum based on the upper stratum and the lower stratum, and using the composite stratum as a preset geological condition; Get the preset fault ratio; Get the preset excavation mode; The preset test conditions are formed based on the preset geological conditions, the preset fault ratio and the preset excavation mode.

3. The method according to claim 1, characterized in that The target simulation record is analyzed to obtain an eccentric wear timing signal, and multi-domain feature extraction is performed on the eccentric wear timing signal to obtain a hob eccentric wear feature set, including: Extracting a thrust simulation timing signal from the target simulation record, and using the thrust simulation timing signal as the eccentric wear timing signal; Reading a predetermined time domain feature, and extracting a time domain feature of the eccentric wear timing signal based on the predetermined time domain feature to obtain a first time domain feature set; Reading a predetermined frequency domain feature, and extracting a frequency domain feature of a first eccentric wear signal spectrum obtained after fast Fourier transforming the eccentric wear time series signal based on the predetermined frequency domain feature, to obtain a first frequency domain feature set; Retrieving a predetermined time-domain and frequency-domain feature extraction strategy, and obtaining a first time-domain and frequency-domain feature set of the eccentric wear timing signal according to the predetermined time-domain and frequency-domain feature extraction strategy; The roller cutter eccentric wear feature set is constructed based on the first time domain feature set, the first frequency domain feature set and the first time domain and frequency domain feature set.

4. The method according to claim 3, characterized in that Retrieving a predetermined time domain and frequency domain feature extraction strategy, and obtaining a first time domain and frequency domain feature set of the eccentric wear timing signal according to the predetermined time domain and frequency domain feature extraction strategy, including: Performing a 2-layer wavelet packet decomposition on the eccentric wear time series signal based on the db4 wavelet basis function to obtain an eccentric wear signal decomposition result, wherein the eccentric wear signal decomposition result includes four sub-band signals; Extracting any sub-band signal from the four sub-band signals, and obtaining average energy of any wavelet packet coefficients of the arbitrary sub-band signal; The first time-domain and frequency-domain feature set is established based on the average energy of the arbitrary wavelet packet coefficients.

5. The method according to claim 4, characterized in that After the hob cutter eccentric wear feature set is formed based on the first time domain feature set, the first frequency domain feature set and the first time domain and frequency domain feature set, the method further includes: Extracting a torque simulation timing signal from the target simulation record, and using the torque simulation timing signal as the eccentric wear timing signal; Extracting and obtaining a second time domain feature set of the eccentric wear timing signal based on the predetermined time domain feature; A second frequency domain feature set of a second eccentric wear signal spectrum obtained by fast Fourier transforming the eccentric wear time series signal is extracted based on the predetermined frequency domain feature; Obtaining a second time-domain and frequency-domain feature set of the eccentric wear timing signal according to the predetermined time-domain and frequency-domain feature extraction strategy; The second time domain feature set, the second frequency domain feature set and the second time domain and frequency domain feature set are added to the roller cutter eccentric wear feature set.

6. The method according to claim 1, characterized in that The feature screening mechanism refers to a mechanism for performing correlation screening on the hob cutter eccentric wear feature set based on a predetermined correlation analysis method, wherein the predetermined correlation analysis method includes a Pearson correlation analysis method and a copula entropy theory analysis method.

7. The TBM cutter eccentric wear identification system based on fault simulation is characterized by: The system is used to perform the method according to any one of claims 1 to 6, and the system comprises: An eccentric wear simulation module is used to obtain preset test conditions and simulate the eccentric wear fault of the cutter on the target TBM under the constraints of the preset test conditions to obtain a target simulation record; A feature extraction module, used for analyzing the target simulation record to obtain an eccentric wear timing signal, and performing multi-domain feature extraction on the eccentric wear timing signal to obtain a hob eccentric wear feature set; A feature screening module is used to introduce a feature screening mechanism to screen and analyze the hob eccentric wear feature set to obtain a target eccentric wear feature set; An eccentric wear identification module, used for taking the target eccentric wear feature set as input information of a roller cutter eccentric wear fault identification model, and obtaining output information through the roller cutter eccentric wear fault identification model; A failure rate judgment module, used to judge whether the target eccentric wear failure rate in the output information reaches a predetermined failure rate threshold; An eccentric wear warning module is used to issue an eccentric wear fault warning instruction if the target eccentric wear failure rate reaches the predetermined failure rate threshold; The tool change execution module is used to perform tool change processing on the target TBM based on the eccentric wear fault warning instruction.

8. A TBM cutter eccentric wear identification device based on fault simulation, characterized in that: The device is used to perform the method according to any one of claims 1 to 6, and the device comprises: A cutter disc, wherein a plurality of grooves are distributed on the cutter disc; A knife seat, the knife seat is mounted on the plurality of grooves of the knife disc by bolts; A hob is installed on the cutter seat.

Images