Intelligent Evaluation System and Method for Cutter Head State of Shield Machine

Abstract

This application relates to the technical field of shield machine cutter head evaluation. It discloses an intelligent evaluation system and method for the state of a shield machine cutter head. By collecting multi-source data from torque, current, and vibration sensors, and using the sliding window feature extraction technology, rich statistical features of the operating state are obtained. Further, temporal context encoding is performed on the time queue of the statistical features of the cutter head operating state to obtain the temporal context encoding features of the cutter head operating state, so as to capture the deep-level information of the cutter head operating state. Finally, the encoded temporal context encoding features of the cutter head operating state are input into the trained GBRT model to achieve accurate prediction of the tool wear degree. This method not only improves the accuracy and comprehensiveness of state monitoring, but also enhances the early warning ability for potential faults, providing guarantee for the safe and stable operation of the shield machine.

Description

Technical Field

[0001] This application relates to the technical field of shield machine cutter head evaluation, and more specifically, to an intelligent evaluation system and method for the state of a shield machine cutter head. Background Art

[0002] During the construction process of a shield machine, as one of the key components, the state of the cutter head directly affects the progress and cost of the entire project. Traditional evaluation of the cutter head state mainly relies on manual inspection and regular maintenance. This method is not only inefficient but also difficult to detect potential problems in a timely manner, easily leading to equipment failures or shutdowns, thus causing huge economic losses. In addition, due to the complex and changeable underground construction environment, it is very difficult for traditional methods to meet the precise monitoring requirements under different geological conditions.

[0003] Although there are already some methods for monitoring based on sensor data in the prior art, these methods are often limited to single - type sensor data, such as only using torque or current data to judge the working state of the cutter head. However, single - type sensor data cannot comprehensively reflect the actual operating conditions of the cutter head. Especially when facing complex geological structures, changes in the pressure and wear degree of the cutter head may not be directly reflected in the changes in torque or current, which limits the accuracy and reliability of the monitoring system. At the same time, there are also deficiencies in data analysis in the prior art. Usually, only a simple threshold alarm mechanism is set, lacking the effective utilization of historical data and the ability to predict future trends. This processing method is difficult to capture the early minor changes of the equipment, greatly reducing the effectiveness of the early warning system.

[0004] Therefore, an optimized intelligent evaluation scheme for the state of a shield machine cutter head is expected. Summary of the Invention

[0005] To solve the above - mentioned technical problems, this application is proposed. The embodiments of this application provide an intelligent evaluation system and method for the state of a shield machine cutter head, which not only improves the accuracy and comprehensiveness of state monitoring but also enhances the early warning ability for potential failures, providing guarantee for the safe and stable operation of the shield machine.

[0006] According to one aspect of the present application, an intelligent evaluation method for the cutter head state of a shield machine is provided, including: collecting data from a torque sensor, a current sensor, and a vibration sensor at a predetermined sampling frequency to obtain a time queue of torque data, a time queue of motor current data, and a time queue of vibration data; performing sliding window feature extraction on the time queue of torque data, the time queue of motor current data, and the time queue of vibration data to obtain a time queue of statistical features of the cutter head operation state, where the statistical features of the cutter head operation state include torque mean, torque variance, motor current mean, motor current peak value, vibration energy, vibration kurtosis, and torque-current correlation coefficient; performing temporal context encoding on the time queue of the statistical features of the cutter head operation state to obtain temporal context encoding features of the cutter head operation state; and inputting the temporal context encoding features of the cutter head operation state into a trained GBRT tool wear degree prediction model to obtain an estimated value of the tool wear degree.

[0007] In the above intelligent evaluation method for the cutter head state of a shield machine, performing sliding window feature extraction on the time queue of torque data, the time queue of motor current data, and the time queue of vibration data to obtain a time queue of statistical features of the cutter head operation state includes: using a sliding window to perform window sampling on the time queue of torque data, the time queue of motor current data, and the time queue of vibration data to obtain a torque data window, a motor current data window, and a vibration data window; and calculating the Pearson correlation coefficient between the torque data window and the vibration data window as the torque-current correlation coefficient.

[0008] In the above intelligent evaluation method for the cutter head state of a shield machine, the size of the sliding window is 1 minute, and its sliding step is 10 seconds.

[0009] In the above intelligent evaluation method for the cutter head state of a shield machine, performing temporal context encoding on the time queue of the statistical features of the cutter head operation state to obtain temporal context encoding features of the cutter head operation state includes: calculating the end constraint association degree factor of the cutter head operation state for each statistical feature of the cutter head operation state in the time queue of the statistical features of the cutter head operation state; calculating the axial position correlation factor of the cutter head operation state for each statistical feature of the cutter head operation state in the time queue of the statistical features of the cutter head operation state; calculating the multi-dimensional constraint fusion factor for each statistical feature of the cutter head operation state based on the end constraint association degree factor and the axial position correlation factor of the cutter head operation state for each statistical feature of the cutter head operation state in the time queue of the statistical features of the cutter head operation state; and performing adaptive constraint propagation encoding on the time queue of the statistical features of the cutter head operation state based on the multi-dimensional constraint fusion factor for each statistical feature of the cutter head operation state to obtain the temporal context encoding features of the cutter head operation state.

[0010] In the above intelligent evaluation method for the cutter head state of a shield machine, calculating the cutter head running state end constraint correlation factor of each cutter head running state statistical feature in the time queue of the cutter head running state statistical features includes: extracting the cutter head running state end time series state representation from the time queue of the cutter head running state statistical features as the cutter head running state spatio-temporal propagation end point feature coding vector; calculating the cutter head running state end constraint correlation factor of each cutter head running state statistical feature in the time queue of the cutter head running state statistical features relative to the cutter head running state spatio-temporal propagation end point feature coding vector.

[0011] In the above intelligent evaluation method for the cutter head state of a shield machine, calculating the cutter head running state axial position correlation factor of each cutter head running state statistical feature in the time queue of the cutter head running state statistical features includes: performing clustering analysis on the time queue of the cutter head running state statistical features to obtain the cutter head running state spatio-temporal propagation main axis feature coding vector; calculating the cutter head running state axial position correlation factor of each cutter head running state statistical feature in the time queue of the cutter head running state statistical features relative to the cutter head running state spatio-temporal propagation main axis feature coding vector.

[0012] In the above intelligent evaluation method for the cutter head state of a shield machine, based on the multi-dimensional constraint fusion factor of each cutter head running state statistical feature, performing adaptive constraint propagation coding on the time queue of the cutter head running state statistical features to obtain the cutter head running state time series context coding feature includes: using the multi-dimensional constraint fusion factor of each cutter head running state statistical feature as a weight, performing weighted fusion on the time queue of the cutter head running state statistical features to obtain the cutter head running state time series context coding feature.

[0013] In the above intelligent evaluation method for the cutter head state of a shield machine, calculating the cutter head running state end constraint correlation factor of each cutter head running state statistical feature in the time queue of the cutter head running state statistical features relative to the cutter head running state spatio-temporal propagation end point feature coding vector includes: calculating the initial cutter head running state end constraint correlation factor; calculating the axial propagation domain space divergence factor of each cutter head running state statistical feature relative to the cutter head running state spatio-temporal propagation end point feature coding vector and the cutter head running state spatio-temporal propagation main axis feature coding vector; constructing a global message propagation equilibrium factor based on the cutter head running state spatio-temporal propagation end point feature coding vector and the cutter head running state spatio-temporal propagation main axis feature coding vector; optimizing the initial cutter head running state end constraint correlation factor based on the axial propagation domain space divergence factor and the global message propagation equilibrium factor to obtain the cutter head running state end constraint correlation factor.

[0014] In the above intelligent evaluation method for the cutterhead state of a shield machine, calculating the cutterhead running state axial position correlation factor of each cutterhead running state statistical feature in the time queue of the cutterhead running state statistical features relative to the cutterhead running state spatio-temporal propagation main axis feature coding vector includes: calculating the initial cutterhead running state axial position correlation factor; and optimizing the initial cutterhead running state axial position correlation factor based on the axial propagation domain spatial divergence factor and the global message propagation equilibrium factor to obtain the cutterhead running state axial position correlation factor.

[0015] According to another aspect of the present application, there is also provided an intelligent evaluation system for the cutterhead state of a shield machine, including: a cutterhead data acquisition module, configured to collect data from a torque sensor, a current sensor, and a vibration sensor at a predetermined sampling frequency to obtain a time queue of torque data, a time queue of motor current data, and a time queue of vibration data; a cutterhead running state statistical feature extraction module, configured to perform sliding window feature extraction on the time queue of the torque data, the time queue of the motor current data, and the time queue of the vibration data to obtain a time queue of cutterhead running state statistical features, where the cutterhead running state statistical features include torque mean, torque variance, motor current mean, motor current peak value, vibration energy, vibration kurtosis, and torque-current correlation coefficient; a cutterhead running state context encoding module, configured to perform temporal context encoding on the time queue of the cutterhead running state statistical features to obtain cutterhead running state temporal context encoding features; and a tool wear degree prediction module, configured to input the cutterhead running state temporal context encoding features into a trained GBRT tool wear degree prediction model to obtain an estimated value of the tool wear degree.

[0016] Compared with the prior art, the intelligent evaluation system and method for the cutterhead state of a shield machine provided by the present application collect multi-source data from torque, current, and vibration sensors, and use the sliding window feature extraction technology to obtain rich running state statistical features. Further, temporal context encoding is performed on the time queue of the cutterhead running state statistical features to obtain cutterhead running state temporal context encoding features to capture the deep information of the cutterhead running state. Finally, the encoded cutterhead running state temporal context encoding features are input into a trained GBRT model to achieve accurate prediction of the tool wear degree. This method not only improves the accuracy and comprehensiveness of state monitoring, but also enhances the early warning ability for potential faults, providing guarantee for the safe and stable operation of the shield machine. Description of the Drawings

[0017] The above and other objects, features, and advantages of the present application will become more apparent by describing the embodiments of the present application in more detail with reference to the accompanying drawings. The drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation to the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0018] Figure 1 It is a schematic flowchart of the intelligent evaluation method for the cutter head state of a shield machine according to an embodiment of the present application.

[0019] Figure 2 It is a schematic diagram of data flow of the intelligent evaluation method for the cutter head state of a shield machine according to an embodiment of the present application.

[0020] Figure 3 It is a schematic flowchart of step S2 in the intelligent evaluation method for the cutter head state of a shield machine according to an embodiment of the present application.

[0021] Figure 4 It is a schematic flowchart of step S3 in the intelligent evaluation method for the cutter head state of a shield machine according to an embodiment of the present application.

[0022] Figure 5 It is a schematic flowchart of step S31 in the intelligent evaluation method for the cutter head state of a shield machine according to an embodiment of the present application.

[0023] Figure 6 It is a schematic flowchart of step S32 in the intelligent evaluation method for the cutter head state of a shield machine according to an embodiment of the present application.

[0024] Figure 7 It is a schematic block diagram of the intelligent evaluation system for the cutter head state of a shield machine according to an embodiment of the present application.

[0025] Figure 8 It is a schematic block diagram of an electronic device according to an embodiment of the present application. Detailed implementation manners

[0026] Next, exemplary embodiments according to the present application will be described in detail with reference to the 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.

[0027] Figure 1 It is a schematic flowchart of the intelligent evaluation method for the cutter head state of a shield machine according to an embodiment of the present application. Figure 2 It is a schematic diagram of data flow of the intelligent evaluation method for the cutter head state of a shield machine according to an embodiment of the present application. As Figure 1 and Figure 2As shown, the intelligent evaluation method for the cutterhead state of the shield machine includes: S1, collecting data from a torque sensor, a current sensor, and a vibration sensor at a predetermined sampling frequency to obtain a time queue of torque data, a time queue of motor current data, and a time queue of vibration data; S2, performing sliding window feature extraction on the time queue of torque data, the time queue of motor current data, and the time queue of vibration data to obtain a time queue of statistical features of the cutterhead operation state, where the statistical features of the cutterhead operation state include torque mean, torque variance, motor current mean, motor current peak value, vibration energy, vibration kurtosis, and torque-current correlation coefficient; S3, performing time series context encoding on the time queue of statistical features of the cutterhead operation state to obtain time series context encoding features of the cutterhead operation state; S4, inputting the time series context encoding features of the cutterhead operation state into a trained GBRT tool wear degree prediction model to obtain an estimated value of the tool wear degree.

[0028] Specifically, in step S1, data is collected from a torque sensor, a current sensor, and a vibration sensor at a predetermined sampling frequency to obtain a time queue of torque data, a time queue of motor current data, and a time queue of vibration data. It should be understood that during the operation of the shield machine, as one of the key components, the working state of the cutterhead directly affects the safety and efficiency of the entire project. In order to ensure that the cutterhead can operate stably under complex geological conditions and detect potential faults in a timely manner, it is particularly important to adopt effective monitoring means. Collecting data from a torque sensor, a current sensor, and a vibration sensor at a predetermined sampling frequency, and then obtaining a time queue of torque data, a time queue of motor current data, and a time queue of vibration data, is the basic step to achieve this goal.

[0029] Specifically, the reason for selecting these three types of sensors for data collection is that they can respectively reflect the operation conditions of the cutterhead in different aspects. The torque sensor is mainly used to measure the torque magnitude borne by the cutterhead during rotation, which is a key indicator for evaluating the cutting force of the cutterhead; the current sensor indirectly reflects the working load of the cutterhead by monitoring the current change of the driving motor; while the vibration sensor can capture the mechanical vibration information of the cutterhead during operation, which is crucial for identifying problems such as cutterhead wear and bearing damage. Since these sensors each have unique physical meanings, their combined use can provide a more comprehensive and accurate basis for evaluating the cutterhead state.

[0030] In specific implementation, an appropriate sampling frequency needs to be set according to actual requirements. In a specific embodiment, the sampling frequency is set to collect data once per second, which means that a new set of readings will be obtained from the above three sensors every second. Considering that the shield machine may encounter various complex geological conditions during tunneling, such as hard rock formations or soft soil layers, etc., the load on the cutter head caused by different geological environments varies greatly. Therefore, a relatively high sampling frequency is required to capture these rapidly changing situations. Of course, it can be adjusted according to the actual situation, and no specific limitation is made in this embodiment.

[0031] In a specific embodiment, before the start of a shield machine engineering project, technicians first install and calibrate the required sensor devices to ensure that they can work properly and the output data is accurate. Then, while the shield machine starts tunneling operations, the preset data acquisition system automatically triggers the sensors to collect data according to the established sampling frequency (such as once per second). Each piece of data collected (including torque values, current intensity, and vibration amplitude, etc.) is marked with the corresponding timestamp and then stored in the central database to form a time series data queue. As time goes by, these time series data gradually accumulate, providing a solid data foundation for subsequent feature extraction, model training, and predictive analysis.

[0032] Specifically, in step S2, a sliding window feature extraction is performed on the time queue of the torque data, the time queue of the motor current data, and the time queue of the vibration data to obtain the time queue of the statistical features of the cutter head operation state. The statistical features of the cutter head operation state include torque mean, torque variance, motor current mean, motor current peak, vibration energy, vibration kurtosis, and torque-current correlation coefficient. It should be understood that considering the huge and complex amount of raw data directly obtained from the sensors, which contains a large amount of noise and unnecessary information, through the sliding window feature extraction technology, key statistical features can be extracted from these raw data, thus more effectively reflecting the real operation state of the cutter head. For example, the torque mean can show the average working load of the cutter head over a period of time; the torque variance reflects the torque fluctuation situation, indirectly reflecting the change of geological conditions or the degree of tool wear; the motor current mean and peak can reveal the load condition of the drive system; the vibration energy and kurtosis can capture the abnormal vibration of the mechanical structure, which is particularly crucial for early detection of equipment failures; finally, the torque-current correlation coefficient provides a method to measure the dynamic relationship between the two, helping to identify potential problems inside the system.

[0033] In an embodiment, as Figure 3As shown, in step S2, sliding window feature extraction is performed on the time queue of the torque data, the time queue of the motor current data, and the time queue of the vibration data to obtain the time queue of the cutter head operation state statistical features, including: S21, using a sliding window to perform window sampling on the time queue of the torque data, the time queue of the motor current data, and the time queue of the vibration data to obtain a torque data window, a motor current data window, and a vibration data window; S22, calculating the Pearson correlation coefficient between the torque data window and the vibration data window as the torque-current correlation coefficient.

[0034] In one embodiment, the size of the sliding window is 1 minute, and its sliding step is 10 seconds. This means that a new data window will be generated every 10 seconds, and this window contains all the data points within the most recent 1 minute. In the specific example of this embodiment, during a continuous monitoring process, assuming the current time is the T moment, then the system will collect the torque, current, and vibration data from T - 60 seconds to the T moment to form the first window. Next, at T + 10 seconds, the system will collect the data from T - 50 seconds to T + 10 seconds again to form the second window, and so on. For each newly generated data window, the system will calculate statistical features such as the torque mean, torque variance, motor current mean, motor current peak, vibration energy, and vibration kurtosis in it, and store these features together with the corresponding timestamps to form the time queue of the cutter head operation state statistical features.

[0035] In addition, calculating the Pearson correlation coefficient between the torque data window and the vibration data window as the torque-current correlation coefficient is also one of the key steps to achieve accurate evaluation. The Pearson correlation coefficient is a standard statistic that measures the strength of the linear relationship between two variables, and its value ranges from -1 to +1. In this context, by calculating the Pearson correlation coefficient between torque and vibration, the degree of mutual influence between these two parameters can be quantified. If it is found that this coefficient significantly deviates from the normal range during a certain period, it may indicate that there are abnormal conditions with the cutter head, such as increased tool wear or encountering special geological conditions, etc. This method based on sliding window feature extraction not only improves the data processing efficiency but also enhances the sensitivity to changes in the cutter head state, enabling maintenance personnel to take actions in a timely manner according to the real-time analysis results and avoid the occurrence of major accidents.

[0036] Specifically, in step S3, a temporal context encoding is performed on the time queue of the statistical features of the cutter head operation status to obtain the temporal context encoding features of the cutter head operation status. It should be understood that the cutter head of a shield machine operates under complex and variable geological conditions, and its operation status is affected by various factors, including but not limited to formation properties, tunneling speed, and the condition of the equipment itself. These factors result in the operation status of the cutter head showing highly non-linear and time series characteristics. Therefore, it is difficult to comprehensively reflect the actual working status and its changing trend of the cutter head only relying on the statistical features at a single time point. By performing a temporal context encoding on the time queue of the statistical features of the cutter head operation status, the patterns and rules of the evolution of these features over time can be captured, thus providing a richer information basis for subsequent analysis.

[0037] Moreover, traditional methods based on single or multiple statistical features often ignore the internal relationships between data. For example, in a shield machine operation scenario, the average torque may remain relatively stable within a short time period, but if the observation scope is extended to a longer time scale, it may be found that this indicator has periodic fluctuations, which may be caused by factors such as tool wear and geological condition changes. By introducing a temporal context encoding mechanism, the long-term dependence relationships and patterns hidden behind the data can be effectively identified, enabling the system to not only recognize the current status but also predict the future development trend based on historical data.

[0038] In one embodiment, as Figure 4 shown, in step S3, performing a temporal context encoding on the time queue of the statistical features of the cutter head operation status to obtain the temporal context encoding features of the cutter head operation status includes: S31, calculating the cutter head operation status end constraint correlation degree factor of each cutter head operation status statistical feature in the time queue of the cutter head operation status statistical features; S32, calculating the cutter head operation status axial position correlation factor of each cutter head operation status statistical feature in the time queue of the cutter head operation status statistical features; S33, calculating the multi-dimensional constraint fusion factor of each cutter head operation status statistical feature based on the cutter head operation status end constraint correlation degree factor and the cutter head operation status axial position correlation factor of each cutter head operation status statistical feature in the time queue of the cutter head operation status statistical features; S34, performing an adaptive constraint propagation encoding on the time queue of the cutter head operation status statistical features based on the multi-dimensional constraint fusion factor of each cutter head operation status statistical feature to obtain the temporal context encoding features of the cutter head operation status.

[0039] In one embodiment, as Figure 5As shown, in step S31, calculate the end constraint correlation degree factor of each cutter head operation state statistical feature in the time queue of the cutter head operation state statistical features, including: S311, extract the end time series state representation of the cutter head operation state from the time queue of the cutter head operation state statistical features as the end feature coding vector of the cutter head operation state spatio-temporal propagation; S312, calculate the end constraint correlation degree factor of each cutter head operation state statistical feature in the time queue of the cutter head operation state statistical features relative to the end feature coding vector of the cutter head operation state spatio-temporal propagation. Specifically, this process can be expressed by the formula: ; where is the time series of the cutter head operation state statistical feature, and are the 1st, 2nd, th, and th cutter head operation state statistical features in the time series of the cutter head operation state statistical feature respectively. Here, for those skilled in the art, the th cutter head operation state statistical feature is the end time series state representation of the cutter head operation state. Therefore, is used as the end feature coding vector of the cutter head operation state spatio-temporal propagation, is the th feature value at the th position of the th cutter head operation state statistical feature, represents the natural constant, is the th feature value at the th position of the th cutter head operation state statistical feature, is the th cutter head operation state statistical feature,

[0040] It should be understood that by extracting the end - time sequence state representation of the cutter head operation state from the time queue of the cutter head operation state statistical features as the feature encoding vector of the spatio - temporal propagation end point of the cutter head operation state, a key reference point can be provided for the entire sequence. This strategy is similar to setting boundary conditions in a dynamic system. It not only provides a fixed reference for subsequent information processing but also enhances the model's understanding and utilization of the end - point information of the time series. Specifically, this approach enables us to more accurately capture the overall state of the cutter head running to the current moment, which is crucial for identifying potential failure modes or wear levels. Then, calculate the cutter head operation state end - point constraint correlation factor of each cutter head operation state statistical feature relative to this feature encoding vector of the spatio - temporal propagation end point of the cutter head operation state. In fact, it is to quantify the similarity or distance between the feature at each time point and the overall end state. This step helps to adjust the importance weights of the features at each time point, ensuring that the features more relevant to the final state can receive more attention during the analysis process. In this way, not only can we focus on the features that have the greatest impact on the prediction results, but also effectively filter out the information that may introduce noise or be misleading, thereby improving the accuracy and reliability of the model prediction.

[0041] In one embodiment, as Figure 6 shown, in step S32, calculate the cutter head operation state axial position correlation factor of each cutter head operation state statistical feature in the time queue of the cutter head operation state statistical features, including: S321, perform clustering analysis on the time queue of the cutter head operation state statistical features to obtain the feature encoding vector of the main axis of spatio - temporal propagation of the cutter head operation state. Specifically, this process can be represented by the formula: ; where and respectively take the maximum and minimum values of the th cutter head operation state statistical feature, is the adjustment hyperparameter. Among them, according to experience, the adjustment hyperparameter is preset to 0.75, which can be adjusted and optimized according to the data situation. This embodiment does not make specific limitations. is the th cutter head operation state statistical reference benchmark value, is the normalization function, is the feature encoding vector of the main axis of spatio - temporal propagation of the cutter head operation state.

[0042] S322, calculate the cutter head operation state axial position correlation factor of each cutter head operation state statistical feature in the time queue of the cutter head operation state statistical features relative to the feature encoding vector of the main axis of spatio - temporal propagation of the cutter head operation state. Specifically, this process can be represented by the formula: ; where To calculate the L2 norm of a vector, is the inverse hyperbolic cosine function, represents the amplitude axial node membership factor corresponding to the statistical feature of the nth cutter head operating state.

[0043] It should be understood that by performing clustering analysis on the time queue of the statistical features of the cutter head operating state, the spatio-temporal propagation main axis feature coding vector representing the main pattern of the entire sequence can be refined. This process is similar to identifying the dominant structure or trend in a dataset, which provides a global perspective for understanding the behavior of the cutter head throughout the working cycle. Such a main axis not only reflects the main dynamics of the cutter head operation but also serves as a framework that enables the features at each individual time point to be more accurately interpreted and located under its guidance. Further calculating the axial position correlation factor of each cutter head operating state statistical feature relative to this spatio-temporal propagation main axis feature coding vector of the cutter head operating state is essentially measuring the degree of association between the feature at each time point and the overall main axis. In this way, the relative importance and position of each feature in the global structure can be determined, ensuring consistency with the main axis during the information transmission process. This helps to emphasize the data features that conform to the overall trend while suppressing the outliers that may deviate from the normal pattern.

[0044] Specifically, for the spatio-temporal propagation end feature coding vector of the cutter head operating state define the end signal transmission limit condition, and for the spatio-temporal propagation main axis feature coding vector of the cutter head operating state define the directional propagation path limit condition. To implement the global calibration mechanism of the projection weight under the global structure constraint, it is necessary to quantify the propagation path correlation effect of each cutter head operating state statistical feature during the signal transmission process along the sequence axial channel.

[0045] Therefore, in a preferred embodiment, first, calculate the initial cutter head operating state end constraint correlation factor and the initial cutter head operating state axial position correlation factor. Specifically, use the cutter head operating state end constraint correlation factor and the cutter head operating state axial position correlation factor calculated in the previous embodiment as the initial cutter head operating state end constraint correlation factor and the initial cutter head operating state axial position correlation factor.

[0046] Next, calculate the axial propagation domain spatial divergence factor of each cutter head operating state statistical feature relative to the spatio-temporal propagation end feature coding vector of the cutter head operating state and the spatio-temporal propagation main axis feature coding vector of the cutter head operating state : ; wherein, represents the axial propagation domain spatial vector, represents the axial propagation domain spatial divergence factor.

[0047] It is represented by measuring the propagation attenuation of the spatial distribution during axial propagation with respect to the end and the main axis.

[0048] Then, a regularization equilibrium under the global message passing structure is established. Based on the spatio-temporal propagation end feature encoding vector of the cutter head operating state and the spatio-temporal propagation main axis feature encoding vector of the cutter head operating state, a global message propagation equilibrium factor is constructed: ; wherein, represents the global message propagation equilibrium factor, represents the first weighted hyperparameter. Among them, the first term measures the degree of deviation of the propagation end from the corresponding subspace constraint through the difference norm, while the second term uses the global difference space response regularization to eliminate the non-smooth coupling during the propagation process, jointly constituting the transfer equilibrium mechanism under the global structure.

[0049] Finally, the weighted sums of and are used to correct the end constraint correlation factor of the cutter head operating state and the axial position correlation factor of the cutter head operating state, which is expressed as: ; wherein, represents the initial end constraint correlation factor of the cutter head operating state, represents the initial axial position correlation factor of the cutter head operating state, represents the second weighted hyperparameter, represents the third weighted hyperparameter, represents the fourth weighted hyperparameter, represents the fifth weighted hyperparameter, represents the end constraint correlation factor of the cutter head operating state, represents the axial position correlation factor of the cutter head operating state. Here, those skilled in the art should know that the weighted hyperparameters can be set by experience or grid search, and the optimal weight hyperparameters can be determined through cross-validation techniques.

[0050] The dynamic distribution characteristics of the global constraint weight and the tail constraint weight in the feature sequence transmission channel field can be balanced for the local cutter head operating state statistical features Moreover, the sequence propagation consistency of the end constraint correlation factor of the cutter head operating state and the axial position correlation factor of the cutter head operating state is improved through local-global regularization equilibrium.

[0051] In one embodiment, calculating the end constraint correlation degree factor of each cutter head operation state statistical feature in the time queue of the cutter head operation state statistical features with respect to the cutter head operation state spatio-temporal propagation end feature coding vector includes: calculating the initial end constraint correlation degree factor of the cutter head operation state; calculating the axial propagation domain space divergence factor of each cutter head operation state statistical feature with respect to the cutter head operation state spatio-temporal propagation end feature coding vector and the cutter head operation state spatio-temporal propagation main axis feature coding vector; constructing a global message propagation equilibrium factor based on the cutter head operation state spatio-temporal propagation end feature coding vector and the cutter head operation state spatio-temporal propagation main axis feature coding vector; and optimizing the initial end constraint correlation degree factor of the cutter head operation state based on the axial propagation domain space divergence factor and the global message propagation equilibrium factor to obtain the end constraint correlation degree factor of the cutter head operation state.

[0052] In one embodiment, calculating the cutter head operation state axial position correlation factor of each cutter head operation state statistical feature in the time queue of the cutter head operation state statistical features with respect to the cutter head operation state spatio-temporal propagation main axis feature coding vector includes: calculating the initial cutter head operation state axial position correlation factor; and optimizing the initial cutter head operation state axial position correlation factor based on the axial propagation domain space divergence factor and the global message propagation equilibrium factor to obtain the cutter head operation state axial position correlation factor.

[0053] In one embodiment, based on the end constraint correlation degree factor and the cutter head operation state axial position correlation factor of each cutter head operation state statistical feature in the time queue of the cutter head operation state statistical features, calculating the multi-dimensional constraint fusion factor of each cutter head operation state statistical feature. Specifically, this process can be expressed by the formula: ; where and are weighted hyperparameters respectively, is a normalization function, is the th multi-dimensional constraint fusion factor corresponding to the cutter head operation state statistical feature.

[0054] It should be understood that by comprehensively considering the end-constraint correlation factor and the axial position correlation factor of the cutter head running state in the time queue of each cutter head running state statistical feature, the multi-dimensional constraint fusion factor of each cutter head running state statistical feature can be calculated. This process is like skillfully combining local characteristics and global trends to form a more comprehensive understanding framework. Specifically, the end-constraint correlation factor emphasizes the connection between the features at each time point and the end state of the sequence, highlighting those critical moments that have a significant impact on the final state; while the axial position correlation factor focuses on evaluating the role and importance of each feature in the overall dynamic pattern. When these two factors are fused into a multi-dimensional constraint fusion factor, in fact, a comprehensive weight that can reflect the relative status and role of each feature in the entire sequence is assigned to each feature. This fusion is not just a simple mathematical superposition, but a deep-level information integration process, which enables the model to not only identify the features closely related to the final state, but also take into account the distribution of these features in the entire data structure. In this way, whether it is capturing long-term dependencies or understanding short-term fluctuations, the model can handle them with ease.

[0055] In one embodiment, based on the multi-dimensional constraint fusion factor of the cutter head running state statistical features, performing adaptive constraint propagation encoding on the time queue of the cutter head running state statistical features to obtain the cutter head running state time-series context encoding feature, including: using the multi-dimensional constraint fusion factor of the cutter head running state statistical features as weights, performing weighted fusion on the time queue of the cutter head running state statistical features to obtain the cutter head running state time-series context encoding feature. Specifically, this process can be represented by the formula: ; where represents the cutter head running state time-series context encoding feature.

[0056] It should be understood that by using the multi-dimensional constraint fusion factor of the statistical features of the operation status of each cutter head to perform adaptive constraint propagation encoding on the time queue of the statistical features of the operation status of the cutter head, a time-series context encoding feature reflecting the deep information of the operation status of the cutter head can be generated. This process is like filtering and adjusting the features at each time point through a carefully designed filter, so that the final obtained information can not only accurately reflect the current state, but also capture the dynamic change patterns in the entire sequence. Specifically, the multi-dimensional constraint fusion factor plays a key role in this process. It assigns a comprehensive weight to the features at each time point. This weight not only considers the correlation between the feature and the end state of the sequence, but also measures its position and importance in the entire data structure. In this way, the adaptive constraint propagation encoding can dynamically adjust the influence of each feature in the information transmission process, ensuring that the features crucial to the prediction result are given due attention, while suppressing the information that may introduce noise or be misleading.

[0057] Specifically, in step S4, the time-series context encoding feature of the cutter head operation status is input into the trained GBRT tool wear degree prediction model to obtain an estimated value of the tool wear degree. It should be understood that the reason for choosing GBRT as the prediction model lies in its strong non-linear fitting ability and high processing speed. GBRT is an ensemble learning method that constructs a strong learner by combining multiple weak learners, thereby improving the overall performance of the model. Compared with traditional single decision trees, GBRT can capture complex patterns in the data more precisely and has high robustness to outliers and noise. In addition, GBRT also supports feature importance analysis, which helps to identify which features contribute the most to the prediction result, thus providing guidance for subsequent data collection and feature engineering.

[0058] In the implementation process, in order to effectively apply the time-series context encoding feature of the cutter head operation status to the GBRT model for tool wear degree prediction, a series of carefully designed steps are required. In a specific embodiment, a large amount of historical data needs to be collected, including sensor data such as torque, current, and vibration under different geological conditions, as well as the corresponding tool wear conditions. These data form the basis of the training set for training the GBRT model. Then, using the above-mentioned sliding window feature extraction technology and time-series context encoding method, the time-series context encoding feature of the cutter head operation status is extracted from the original sensor data. Then, the time-series context encoding feature of the cutter head operation status is used as an input variable and fed into the trained GBRT model, and the model calculates the output according to the pre-learned knowledge, that is, the estimated value of the tool wear degree.

[0059] Specifically, the GBRT model is trained based on a large amount of historical data and learns how to predict the wear degree of the cutting tool according to the given feature vector. Inside the model, there are multiple regression trees, and each tree tries to correct the errors of the previous tree, finally forming a highly accurate prediction model. When the time-series context encoding features of the new cutterhead operation state are fed into the model, the GBRT calculates the response values of each node layer by layer until it reaches the leaf node, thereby obtaining the prediction result of the cutting tool wear degree. If the prediction result shows that the cutting tool wear is close to the critical value, the system will automatically send a warning notice to the maintenance personnel to prepare for replacing the cutting tool in advance, avoiding equipment failures or safety accidents caused by excessive cutting tool wear.

[0060] In summary, the intelligent evaluation method for the cutterhead state of the shield machine provided in this application has been clarified. It collects multi-source data from torque sensors, current sensors, and vibration sensors and uses the sliding window feature extraction technology to obtain rich statistical features of the operation state. Further, the time series of the statistical features of the cutterhead operation state is encoded in the time context to obtain the time-series context encoding features of the cutterhead operation state, so as to capture the deep information of the cutterhead operation state. Finally, the encoded time-series context encoding features of the cutterhead operation state are input into the trained GBRT model to achieve accurate prediction of the cutting tool wear degree. This method not only improves the accuracy and comprehensiveness of state monitoring but also enhances the early warning ability for potential failures, providing a guarantee for the safe and stable operation of the shield machine.

[0061] This application also provides an intelligent evaluation system for the cutterhead state of a shield machine, as Figure 7 shown. The intelligent evaluation system 600 for the cutterhead state of the shield machine includes: a cutterhead data acquisition module 610, configured to collect data from a torque sensor, a current sensor, and a vibration sensor at a predetermined sampling frequency to obtain a time queue of torque data, a time queue of motor current data, and a time queue of vibration data; a cutterhead operation state statistical feature extraction module 620, configured to perform sliding window feature extraction on the time queue of the torque data, the time queue of the motor current data, and the time queue of the vibration data to obtain a time queue of cutterhead operation state statistical features, where the cutterhead operation state statistical features include torque mean, torque variance, motor current mean, motor current peak value, vibration energy, vibration kurtosis, and torque-current correlation coefficient; a cutterhead operation state context encoding module 630, configured to perform time-series context encoding on the time queue of the cutterhead operation state statistical features to obtain time-series context encoding features of the cutterhead operation state; and a cutting tool wear degree prediction module 640, configured to input the time-series context encoding features of the cutterhead operation state into a trained GBRT cutting tool wear degree prediction model to obtain an estimated value of the cutting tool wear degree.

[0062] This embodiment of the application also provides an electronic device, as Figure 8As shown, the electronic device 10 includes one or more processors 11 and a memory 12. The processor 11 may be a central processing unit (CPU) or other forms of processing units with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 10 to perform desired functions. The memory 12 may include one or more computer program products, and the computer program products may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory, etc. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage media, and the processor 11 may run the program instructions to implement the intelligent leakage protection of the electric water heater and / or other desired functions of the various embodiments of the present application described above. Various contents such as the acquired leakage current signals for a predetermined period of time may also be stored in the computer-readable storage media. In one example, the electronic device 10 may further include: an input device 13 and an output device 14, and these components are interconnected through a bus system and / or other forms of connection mechanisms (not shown).

[0063] In addition to the above methods, systems, and electronic devices, the embodiments of the present application may also be computer program products, which include computer program instructions that, when run by a processor, cause the processor to execute the corresponding methods provided above.

[0064] The computer program products may be written in any combination of one or more programming languages to write program codes for performing the operations of the embodiments of the present application. The programming languages include object-oriented programming languages such as Java, C++, etc., and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program codes may be executed entirely on the user computing device, partially on the user device, executed as an independent software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0065] Furthermore, the embodiments of the present application may also be computer-readable storage media, on which computer program instructions are stored, and the computer program instructions, when run by a processor, cause the processor to execute the corresponding methods provided above.

[0066] The computer-readable storage medium may adopt any combination of one or more readable media. The readable media may be a readable signal medium or a readable storage medium. The readable storage medium may include, for example, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0067] The basic principles of the present application have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, benefits, effects, etc. mentioned in the present application are only examples and not limitations. It cannot be considered that these advantages, benefits, effects, etc. are essential for each embodiment of the present application. Additionally, the above-disclosed specific details are only for illustrative and easy-to-understand purposes and are not limitations. The above details do not limit the present application to necessarily adopt the above specific details for implementation.

[0068] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to the broadest scope consistent with the principles and novel features disclosed herein. The above description has been given for purposes of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions, and sub-combinations thereof.

Claims

1. An intelligent evaluation method for the cutter head state of a shield machine, characterized in that, Including: Collecting data from a torque sensor, a current sensor, and a vibration sensor at a predetermined sampling frequency to obtain a time queue of torque data, a time queue of motor current data, and a time queue of vibration data; Performing sliding window feature extraction on the time queue of the torque data, the time queue of the motor current data, and the time queue of the vibration data to obtain a time queue of cutter head operation state statistical features, where the cutter head operation state statistical features include torque mean, torque variance, motor current mean, motor current peak value, vibration energy, vibration kurtosis, and torque-current correlation coefficient; Performing temporal context encoding on the time queue of the cutter head operation state statistical features to obtain cutter head operation state temporal context encoding features; Inputting the cutter head operation state temporal context encoding features into a trained GBRT tool wear degree prediction model to obtain an estimated value of the tool wear degree; Among them, performing temporal context encoding on the time queue of the cutter head operation state statistical features to obtain cutter head operation state temporal context encoding features includes: calculating the cutter head operation state end constraint correlation degree factor of each cutter head operation state statistical feature in the time queue of the cutter head operation state statistical features; calculating the cutter head operation state axial position correlation factor of each cutter head operation state statistical feature in the time queue of the cutter head operation state statistical features; calculating the multi-dimensional constraint fusion factor of each cutter head operation state statistical feature based on the cutter head operation state end constraint correlation degree factor and the cutter head operation state axial position correlation factor of each cutter head operation state statistical feature in the time queue of the cutter head operation state statistical features; performing adaptive constraint propagation encoding on the time queue of the cutter head operation state statistical features based on the multi-dimensional constraint fusion factor of each cutter head operation state statistical feature to obtain the cutter head operation state temporal context encoding features.

2. The intelligent evaluation method for the cutter head state of a shield machine according to claim 1, characterized in that Performing sliding window feature extraction on the time queue of the torque data, the time queue of the motor current data, and the time queue of the vibration data to obtain a time queue of cutter head operation state statistical features includes: using a sliding window to perform window sampling on the time queue of the torque data, the time queue of the motor current data, and the time queue of the vibration data to obtain a torque data window, a motor current data window, and a vibration data window; calculating the Pearson correlation coefficient between the torque data window and the vibration data window as the torque-current correlation coefficient.

3. The intelligent evaluation method for the cutter head state of a shield machine according to claim 2, characterized in that The size of the sliding window is 1 minute, and its sliding step is 10 seconds.

4. The intelligent evaluation method for the cutter head state of a shield machine according to claim 1, characterized in that Calculating the end-constraint correlation degree factors of each cutter head operation state statistical feature in the time queue of the cutter head operation state statistical features, including: extracting the end-time sequence state representation of the cutter head operation state from the time queue of the cutter head operation state statistical features as the end feature coding vector of the spatio-temporal propagation of the cutter head operation state; calculating the end-constraint correlation degree factors of each cutter head operation state statistical feature in the time queue of the cutter head operation state statistical features relative to the end feature coding vector of the spatio-temporal propagation of the cutter head operation state.

5. The intelligent evaluation method for the cutter head state of a shield machine according to claim 4, characterized in that Calculating the axial position correlation factors of each cutter head operation state statistical feature in the time queue of the cutter head operation state statistical features, including: performing clustering analysis on the time queue of the cutter head operation state statistical features to obtain the main axis feature coding vector of the spatio-temporal propagation of the cutter head operation state; calculating the axial position correlation factors of each cutter head operation state statistical feature in the time queue of the cutter head operation state statistical features relative to the main axis feature coding vector of the spatio-temporal propagation of the cutter head operation state.

6. The intelligent evaluation method for the cutter head state of a shield machine according to claim 5, wherein, Calculating the end-constraint correlation degree factors of each cutter head operation state statistical feature in the time queue of the cutter head operation state statistical features relative to the end feature coding vector of the spatio-temporal propagation of the cutter head operation state, including: calculating the initial end-constraint correlation degree factor of the cutter head operation state; calculating the axial propagation domain space divergence factor of each cutter head operation state statistical feature relative to the end feature coding vector of the spatio-temporal propagation of the cutter head operation state and the main axis feature coding vector of the spatio-temporal propagation of the cutter head operation state; constructing a global message propagation equilibrium factor based on the end feature coding vector of the spatio-temporal propagation of the cutter head operation state and the main axis feature coding vector of the spatio-temporal propagation of the cutter head operation state; optimizing the initial end-constraint correlation degree factor of the cutter head operation state based on the axial propagation domain space divergence factor and the global message propagation equilibrium factor to obtain the end-constraint correlation degree factor of the cutter head operation state.

7. The intelligent evaluation method for the cutter head state of a shield machine according to claim 6, characterized in that, Calculating the axial position correlation factors of each cutter head operation state statistical feature in the time queue of the cutter head operation state statistical features relative to the main axis feature coding vector of the spatio-temporal propagation of the cutter head operation state, including: calculating the initial axial position correlation factor of the cutter head operation state; optimizing the initial axial position correlation factor of the cutter head operation state based on the axial propagation domain space divergence factor and the global message propagation equilibrium factor to obtain the axial position correlation factor of the cutter head operation state.

8. The intelligent evaluation method for the cutter head state of a shield machine according to claim 7, wherein, Based on the multi-dimensional constraint fusion factors of each cutter head operation state statistical feature, performing adaptive constraint propagation coding on the time queue of the cutter head operation state statistical features to obtain the time-sequence context coding feature of the cutter head operation state, including: using the multi-dimensional constraint fusion factors of each cutter head operation state statistical feature as weights to perform weighted fusion on the time queue of the cutter head operation state statistical features to obtain the time-sequence context coding feature of the cutter head operation state.

9. An intelligent evaluation system for the cutter head state of a shield machine, which is used to execute the intelligent evaluation method for the cutter head state of the shield machine according to any one of claims 1-8, characterized in that, Including: The cutter head data acquisition module is used to collect data from the torque sensor, current sensor, and vibration sensor at a predetermined sampling frequency to obtain the time queues of torque data, motor current data, and vibration data; the cutter head operation state statistical feature extraction module is used to perform sliding window feature extraction on the time queues of the torque data, the motor current data, and the vibration data to obtain the time queue of the cutter head operation state statistical features, and the cutter head operation state statistical features include torque mean, torque variance, motor current mean, motor current peak value, vibration energy, vibration kurtosis, and torque-current correlation coefficient; The cutter head operation state context encoding module is used to perform temporal context encoding on the time queue of the cutter head operation state statistical features to obtain the cutter head operation state temporal context encoding features; The tool wear degree prediction module is used to input the cutter head operation state temporal context encoding features into the trained GBRT tool wear degree prediction model to obtain the tool wear degree estimation value.

Images