Shield tunneling machine operation data processing method and device for hob abrasion prediction, computer equipment, readable storage medium and program product

Abstract

The invention relates to a shield tunneling machine operation data processing method and device for hob abrasion prediction, computer equipment, a computer readable storage medium and a computer program product, relates to the technical field of tunnel excavation, and can improve the prediction precision of a hob abrasion prediction model for the hob abrasion loss. The method comprises the following steps: acquiring operation data and geological exploration data of the shield tunneling machine; determining a decomposition basis function and a decomposition frequency band number, and decomposing the operation data according to the decomposition basis function and the decomposition frequency band number to obtain a plurality of operation sub-data; in the operation sub-data, keeping the operation sub-data reflecting the working characteristics of the shield tunneling machine, and performing data reconstruction on the kept operation sub-data to obtain candidate operation data; and determining the data compression size according to the geological exploration data, performing feature extraction on the candidate operation data according to the data compression size to obtain target operation data, and inputting the target operation data into the hob wear prediction model.

Description

Technical Field

[0001] The present application relates to the technical field of tunnel excavation, and in particular to a shield machine operation data processing method, device, computer equipment, computer-readable storage medium and computer program product for predicting cutter wear. Background Art

[0002] The shield method is the mainstream means of tunnel construction. The disc cutters are important components of the shield machine cutterhead in the process of rock breaking and excavation. Their overall condition directly affects the efficiency and quality of the shield. Too frequent shutdowns for maintenance will increase economic and time costs, but too long a cutter change distance will reduce the efficiency of the shield and even pose a risk of damaging the cutterhead. Therefore, accurate monitoring of the wear condition of the disc cutters is crucial for safe, sustainable construction and cost management of tunnels. At present, machine learning technology has been applied to tunnel excavation projects, providing an effective way to explore more complex nonlinear relationships between disc cutter wear, geological conditions and shield operating parameters.

[0003] However, there are some problems with the current data processing methods during prediction. Since the shield operation parameter signals are recorded by industrial sensors and programmable logic controllers (PLCs), the working conditions are complex and the collected data will inevitably be interfered with or lost. On the other hand, the geological information of the tunnel is obtained from the measurement results of a series of exploration boreholes. These data are discrete and sparse, which makes it impossible to avoid the problem of how to effectively combine the high-frequency shield parameter acquisition data (30-60 seconds / time, about 6,300 times / hundred meters) with the sparse geological parameters (4 times (groups) / hundred meters) when training the model. Therefore, it is urgent to perform appropriate data processing before model prediction, scientifically solve the huge gap in data volume, and improve the model's prediction effect on roller cutter wear. Summary of the invention

[0004] Based on this, it is necessary to provide a shield machine operation data processing method, device, computer equipment, computer readable storage medium and computer program product for predicting roller cutter wear, which can improve the prediction effect of the model for roller cutter wear, in order to address the above technical problems.

[0005] In a first aspect, the present application provides a shield machine operation data processing method for predicting cutter wear, comprising:

[0006] Obtain shield machine operation data and geological exploration data;

[0007] Determine a decomposition basis function and the number of decomposition frequency bands, and decompose the operation data according to the decomposition basis function and the number of decomposition frequency bands to obtain a plurality of operation sub-data;

[0008] Among the operation sub-data, the operation sub-data reflecting the working characteristics of the shield machine are retained, and the retained operation sub-data are reconstructed to obtain candidate operation data;

[0009] The data compression size is determined according to the geological exploration data, and the candidate operation data is feature extracted according to the data compression size to obtain target operation data, and the target operation data is input into the hob wear prediction model.

[0010] In one embodiment, the extracting features of the candidate operating data according to the data compression size to obtain the target operating data includes:

[0011] According to the data compression size, the candidate running data is segmented and sorted to obtain a plurality of segmented sub-data arranged in sequence;

[0012] The mean value of each segmented sub-data is calculated to obtain the characteristic operation data corresponding to each segmented sub-data, and the characteristic operation data are combined in the order to obtain the target operation data.

[0013] In one embodiment, the operation sub-data reflecting the working characteristics of the shield machine are retained, including:

[0014] Determining the decomposition coefficient of the running sub-data in the corresponding decomposition frequency band;

[0015] Perform median absolute deviation calculation on the decomposition coefficient corresponding to the highest decomposition frequency band to obtain a calculation result, and use the calculation result as a conditional threshold for retaining the running sub-data;

[0016] If the absolute value of the decomposition coefficient is greater than the conditional threshold, the sub-data corresponding to the decomposition coefficient is retained.

[0017] In one embodiment, before inputting the target operating data into the hob wear prediction model, the method further includes:

[0018] Performing normalization operation on the target operation data to obtain normalized operation data;

[0019] Determining the data size of the normalized running data, and increasing the data size according to a preset ratio to obtain a target data size;

[0020] The normalized operating data is expanded, and when the size of the expanded data reaches the target data size, the expansion is stopped to obtain the expanded target operating data.

[0021] In one embodiment, the hob wear prediction model is obtained by the following steps:

[0022] Inputting the historical target operation data into the hob wear prediction model to be trained, the hob wear prediction model determines the hidden information at each time point in the historical target operation data, and performs weighted processing on the hidden information to obtain the operation characteristics at each time point; and obtaining the wear amount prediction result according to the operation characteristics at each time point;

[0023] According to the wear prediction result and the hob wear label, the model loss value is determined, and the model parameters of the hob wear prediction model are adjusted. When the training end condition is met, the hob wear prediction model is obtained.

[0024] In one embodiment, determining the data compression size according to the geological exploration data includes:

[0025] Determine the data compression size matching the operation data according to the excavation length of the geological exploration data; or,

[0026] Determine the data compression size matching the operation data according to the number of tunneling rings of the geological exploration data; or,

[0027] The data compression size matching the operation data is determined according to the number of exploration points of the geological exploration data.

[0028] In a second aspect, the present application also provides a shield machine operation data processing device for predicting cutter wear, comprising:

[0029] Operation data acquisition and geological exploration data acquisition module, used to acquire shield machine operation data and geological exploration data;

[0030] An operation data decomposition and sub-data extraction module is used to determine a decomposition basis function and the number of decomposition frequency bands, and decompose the operation data according to the decomposition basis function and the number of decomposition frequency bands to obtain a plurality of operation sub-data;

[0031] An operation data reconstruction and candidate data generation module, used for retaining the operation sub-data reflecting the working characteristics of the shield machine in the operation sub-data, and reconstructing the retained operation sub-data to obtain candidate operation data;

[0032] The feature extraction and target data generation module is used to determine the data compression size according to the geological exploration data, perform feature extraction on the candidate operation data according to the data compression size, obtain target operation data, and input the target operation data into the hob wear prediction model.

[0033] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0034] Obtain shield machine operation data and geological exploration data;

[0035] Determine a decomposition basis function and the number of decomposition frequency bands, and decompose the operation data according to the decomposition basis function and the number of decomposition frequency bands to obtain a plurality of operation sub-data;

[0036] Among the operation sub-data, the operation sub-data reflecting the working characteristics of the shield machine are retained, and the retained operation sub-data are reconstructed to obtain candidate operation data;

[0037] The data compression size is determined according to the geological exploration data, and the candidate operation data is feature extracted according to the data compression size to obtain target operation data, and the target operation data is input into the hob wear prediction model.

[0038] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0039] Obtain shield machine operation data and geological exploration data;

[0040] Determine a decomposition basis function and the number of decomposition frequency bands, and decompose the operation data according to the decomposition basis function and the number of decomposition frequency bands to obtain a plurality of operation sub-data;

[0041] Among the operation sub-data, the operation sub-data reflecting the working characteristics of the shield machine are retained, and the retained operation sub-data are reconstructed to obtain candidate operation data;

[0042] The data compression size is determined according to the geological exploration data, and the candidate operation data is feature extracted according to the data compression size to obtain target operation data, and the target operation data is input into the hob wear prediction model.

[0043] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0044] Obtain shield machine operation data and geological exploration data;

[0045] Determine a decomposition basis function and the number of decomposition frequency bands, and decompose the operation data according to the decomposition basis function and the number of decomposition frequency bands to obtain a plurality of operation sub-data;

[0046] Among the operation sub-data, the operation sub-data reflecting the working characteristics of the shield machine are retained, and the retained operation sub-data are reconstructed to obtain candidate operation data;

[0047] The data compression size is determined according to the geological exploration data, and the candidate operation data is feature extracted according to the data compression size to obtain target operation data, and the target operation data is input into the hob wear prediction model.

[0048] The shield machine operation data processing method, device, computer equipment, computer readable storage medium and computer program product for roller cutter wear prediction described above obtain the operation data and geological exploration data of the shield machine; determine the decomposition basis function and the number of decomposition frequency bands, decompose the operation data according to the decomposition basis function and the number of decomposition frequency bands, and obtain multiple operation sub-data; in the operation sub-data, the operation sub-data reflecting the working characteristics of the shield machine are retained, and the retained operation sub-data are reconstructed to obtain candidate operation data; determine the data compression size according to the geological exploration data, extract features from the candidate operation data according to the data compression size, obtain target operation data, and input the target operation data into the roller cutter wear prediction model. In the present application, by determining the decomposition basis function and the number of decomposition frequency bands, the shield machine operation data is decomposed, which can effectively separate noise and effective signals and realize data cleaning; determine the data compression size according to the geological data, so that the target operation data obtained after the feature extraction of the candidate operation data can match the geological exploration data in magnitude, thereby improving the effect of model prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0050] Figure 1 An application environment diagram of a shield machine operation data processing method for roller cutter wear prediction in one embodiment;

[0051] Figure 2 A schematic flow chart of a shield machine operation data processing method for predicting cutter wear in one embodiment;

[0052] Figure 3 A schematic diagram of a hob wear prediction model according to an embodiment;

[0053] Figure 4A schematic flow chart of a shield machine operation data processing method for predicting cutter wear in another embodiment;

[0054] Figure 5 It is a schematic diagram of the result of hob wear prediction based on wavelet transform denoising and ring averaging method in one embodiment;

[0055] Figure 6 It is a schematic diagram of the result of hob wear prediction based on Fourier transform denoising and segmented aggregation averaging method in one embodiment;

[0056] Figure 7 It is a schematic diagram of the results of hob wear prediction based on a combination of a ring averaging method, a segmented aggregation averaging method, wavelet transform denoising, Fourier transform denoising, smoothing average denoising and autoencoder denoising in one embodiment;

[0057] Figure 8 It is a structural block diagram of a shield machine operation data processing device for predicting cutter wear in one embodiment;

[0058] Fig. 9 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0060] The shield machine operation data processing method for predicting cutter wear provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. The terminal communicates with the server through the network. The data storage system can store the data that the server needs to process. The data storage system can be integrated on the server, or placed on the cloud or other network servers. The terminal can be, but is not limited to, various personal computers, laptops, smart phones, and tablets. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services.

[0061] In an exemplary embodiment, Figure 2 As shown in the figure, a shield machine operation data processing method for predicting cutter wear is provided. Figure 1 The terminal in is taken as an example to illustrate, including the following steps S202 to S208. Among them:

[0062] Step S202, obtaining the operation data and geological exploration data of the shield machine.

[0063] Among them, shield machine operation data refers to the parameters related to equipment operation collected in real time by sensors during the operation of the shield machine, including but not limited to thrust, torque, propulsion speed, excavation depth, etc. Shield machine operation data can be stored in the form of time series to reflect the current working status and historical operation of the shield machine. Geological exploration data refers to relevant information describing underground soil layers and geological structures obtained through drilling, sampling or other exploration methods, including but not limited to soil layer classification, formation hardness, groundwater distribution, rock layer thickness, etc.

[0064] For example, by deploying various types of sensors (such as thrust sensors, torque sensors, etc.) on the shield machine, the real-time operating parameters of the equipment are collected during the tunneling operation of the shield machine, and the operating data is archived as multi-dimensional time series data, where each dimension corresponds to a specific operating parameter collected by a sensor. The geological attribute data of the pre-deployed geological detection equipment is obtained, and the collected geological exploration data is marked according to the construction section.

[0065] Step S204, determining the decomposition basis function and the number of decomposition frequency bands, and decomposing the operation data according to the decomposition basis function and the number of decomposition frequency bands to obtain a plurality of operation sub-data.

[0066] The decomposition basis function is a set of mathematical functions used to decompose and reconstruct signals, which can decompose complex signals into different frequency bands or characteristic components in a specific decomposition method. The decomposition basis function includes but is not limited to wavelet basis function or Fourier basis function. The selection of the decomposition basis function is determined by the characteristics of the running data, and the terminal can select the corresponding decomposition basis function according to the data type.

[0067] The number of decomposed frequency bands refers to the number of frequency bands into which the data is divided during the signal decomposition process. The terminal can dynamically adjust the number of decomposed frequency bands according to the decomposition task (such as extracting the operation characteristics of the shield machine) to ensure the optimal processing result.

[0068] Exemplarily, the terminal determines the frequency domain characteristics in the operating data, and selects an adaptive wavelet basis function in combination with the non-stationarity of the shield machine operating data. The number of decomposition frequency bands is set according to the spectral range of the signal and the analysis requirements. Optionally, when the shield machine operating data contains high-frequency vibrations and low-frequency thrust changes, the decomposition layer is set to 6 layers. Optionally, if the accuracy requirements for a specific frequency band are high, the number of decomposition layers can be dynamically adjusted to increase the degree of refinement of the frequency band. The terminal executes a wavelet decomposition algorithm to decompose the shield machine's operating data into several operating sub-data. Each sub-data represents the component of the operating data in a certain frequency band.

[0069] Step S206, in the operation sub-data, the operation sub-data reflecting the working characteristics of the shield machine are retained, and the retained operation sub-data are reconstructed to obtain candidate operation data.

[0070] Among them, the operation sub-data with working characteristics is a subset of the operation sub-data, which contains information directly related to the key operation status of the shield machine. Retaining these operation sub-data can help accurately capture the operation status of the shield machine and filter out irrelevant information related to noise.

[0071] Data reconstruction refers to overlapping the retained running sub-data according to the time sequence or frequency of the original signal to restore the main characteristics of the signal. Through data reconstruction, the decomposed useful signal is recombined into candidate data that can be used for analysis, while removing irrelevant or noise components.

[0072] Exemplarily, the terminal calculates the eigenvalues ​​of each decomposed operation sub-data. According to the preset feature retention conditions, the sub-data reflecting the working characteristics of the shield machine are screened out. For the operation sub-data that do not meet the retention conditions, the terminal marks them as irrelevant signals and discards them. The terminal performs an inverse transformation on the retained operation sub-data and recombines all retained signal components into complete time series data. During the data reconstruction process, the terminal device superimposes or splices each sub-data to ensure the consistency of the timing and amplitude of the candidate operation data.

[0073] Optionally, if the operation sub-data contains multiple frequency band signals, the terminal uses the variance contribution rate of the variance of each frequency band component relative to the total signal variance as the retention basis. The operation sub-data with a variance contribution rate higher than a certain threshold is considered as a working characteristic signal, and the remaining operation sub-data is discarded.

[0074] Step S208, determining the data compression size according to the geological exploration data, performing feature extraction on the candidate operation data according to the data compression size, obtaining target operation data, and inputting the target operation data into the hob wear prediction model.

[0075] Data compression size refers to the data reduction ratio when the candidate operation data is reduced in dimension or simplified. Optionally, the terminal can dynamically adjust the compression size according to the complexity of the geological data and the target resolution. Feature extraction is the process of extracting representative statistics or signal characteristics from the candidate operation data to generate more concise target operation data.

[0076] Exemplarily, the terminal reads relevant data of the excavation section from the geological exploration database, including the excavation length, stratum hardness and groundwater distribution, and determines the data compression size. In one embodiment, if the stratum is soft and the excavation length is long, a higher compression ratio is set to reduce unnecessary calculations. The terminal segments the candidate operation data according to the compression size to generate a number of data segments. Statistical features are calculated for each segment of data, and the obtained statistical features are combined into target operation data in chronological order. The statistical features include but are not limited to time domain features, frequency domain features and energy features. The terminal inputs the generated target operation data into the cutter wear prediction model in the form of a time series, triggers the model to perform calculations, and obtains wear prediction results.

[0077] Optionally, in addition to statistical feature extraction, the terminal may also use a principal component analysis method to reduce the dimension of the compressed data segment to several principal components as target operating data, and the amount of information extracted from each principal component is controlled by the cumulative contribution rate.

[0078] In the shield machine operation data processing method for roller cutter wear prediction, the operation data and geological exploration data of the shield machine are obtained; the decomposition basis function and the number of decomposition frequency bands are determined, and the operation data is decomposed according to the decomposition basis function and the number of decomposition frequency bands to obtain multiple operation sub-data; in the operation sub-data, the operation sub-data reflecting the working characteristics of the shield machine are retained, and the retained operation sub-data are reconstructed to obtain candidate operation data; the data compression size is determined according to the geological exploration data, and the candidate operation data is feature extracted according to the data compression size to obtain target operation data, and the target operation data is input into the roller cutter wear prediction model. In this embodiment, by determining the decomposition basis function and the number of decomposition frequency bands, the shield machine operation data is decomposed, which can effectively separate noise and effective signals and realize data cleaning; the data compression size is determined according to the geological data, so that the target operation data obtained after the candidate operation data is feature extracted can match the geological exploration data in magnitude, thereby improving the effect of model prediction.

[0079] In an exemplary embodiment, in step S208, feature extraction is performed on the candidate operating data according to the data compression size to obtain target operating data, including: segmenting and sorting the candidate operating data according to the data compression size to obtain a plurality of segmented sub-data arranged in sequence; performing mean calculation on each segmented sub-data to obtain feature operating data corresponding to each segmented sub-data, and combining the feature operating data in sequence to obtain the target operating data.

[0080] The segmented sub-data are candidate operation data segments divided according to the compression size, and each segment of sub-data is a subset of the candidate operation data in the time dimension.

[0081] Specifically, the candidate running data sequence group P={P1,P2,...,P k}, for each sequence P k According to the set data compression size n, the data is segmented. The length of each sequence is L. The segment size is determined as size=L / / n (integer division). The mean of each segment of data is calculated in turn. If the sequence length L cannot divide the segment size (i.e. L%n≠0), the remaining data points are treated as a separate segment. The segment means are combined in order to generate the compressed sequence P PAAk For each input sequence P k According to the above steps, the compressed sequence group P is finally generated. PAA ={P PAA1 ,P PAA2 ,...,P PAAk}, and determine the obtained ordered sequence as the target running data.

[0082] In this embodiment, by segmenting and averaging according to the data compression size, the size of the candidate operation data is effectively compressed, the amount of redundant information is reduced, and the overall trend of the signal is retained. The compression ratio is dynamically adjusted according to the geological conditions, and the sequence key points are extracted. The key points can reflect the working characteristics of the shield machine at this stage. The extraction is performed on the basis of retaining the operation data characteristics, making the subsequent calculation more convenient, efficient and accurate.

[0083] In an exemplary embodiment, in step S204, the operating sub-data reflecting the working characteristics of the shield machine are retained, including: determining the decomposition coefficient of the operating sub-data in the corresponding decomposition frequency band; performing median absolute deviation calculation on the decomposition coefficient corresponding to the highest decomposition frequency band to obtain a calculation result, and using the calculation result as a conditional threshold for retaining the operating sub-data; if the absolute value of the decomposition coefficient is greater than the conditional threshold, retaining the sub-data corresponding to the decomposition coefficient.

[0084] Specifically, the shield machine operation data is decomposed by multi-layer wavelet transform using formula (1) to generate multiple operation sub-data, each of which corresponds to a specific decomposition frequency band, where DWT is discrete wavelet transform, P is shield machine operation data, is the decomposition coefficient corresponding to the decomposition frequency band, is the number of decomposition levels, the wavelet basis function is wavele=db6, and the decomposition level is set to level=6.

[0085]

[0086] Select the decomposition coefficient corresponding to the highest decomposition frequency band, calculate the median absolute deviation of the decomposition coefficient of this frequency band by formula (2), and obtain the conditional threshold for retaining the running sub-data ,in indicates the median calculation, and 06745 is the normalization factor used to convert the median of the decomposition coefficients into an estimate of the standard deviation of the noise.

[0087]

[0088] Based on the condition threshold of running child data retention , according to formula (3), the decomposition coefficient of each frequency band Filter, among which is the decomposition coefficient after screening, when the original decomposition coefficient The absolute value of is greater than the conditional threshold When , keep the original value; otherwise, set the coefficient to zero.

[0089]

[0090] The denoised candidate running data is obtained by using formula (4) using the screened decomposition coefficients and inverse discrete wavelet reconstruction (IDWT) .

[0091]

[0092] In this embodiment, the decomposition coefficients of the decomposition frequency band are screened by using the median absolute deviation calculation condition threshold, which successfully eliminates background noise and irrelevant signals, and retains only key sub-data closely related to the working characteristics of the shield machine, providing a concise and high-quality signal basis for subsequent data reconstruction and model prediction.

[0093] In an exemplary embodiment, before inputting the target operating data into the hob wear prediction model, it also includes: normalizing the target operating data to obtain normalized operating data; determining the data size of the normalized operating data, increasing the data size according to a preset expansion ratio to obtain the target data size; expanding the normalized operating data, and stopping the expansion when the expanded data size reaches the target data size to obtain the expanded target operating data.

[0094] Exemplarily, the terminal reads the target operating data, maps the target operating data to the interval [0,1] using the maximum and minimum normalization formula, obtains the normalized operating data, and calculates the current data size of the current normalized operating data. The target data size is set according to the model input requirements, and the normalized data is expanded by setting the expansion ratio. Check whether the expanded data size reaches the target data size. If the expanded data size reaches the target data size, stop expanding and generate the final expanded target operating data.

[0095] Specifically, before inputting the cutter wear prediction model, the data must first be normalized. Each ring of shield excavation is regarded as a time step, and each time step corresponds to the key points of shield operation parameters, geological information and cutterhead geometry information of the target operation data. In order to ensure that the input data of the model has a reasonable time step length, the maximum time step can be determined optionally based on the cutter change records in the same project. Specifically, the maximum time step can be estimated by selecting the longest number of excavation rings experienced by the cutter with the longest continuous shield distance in the cutter replacement record in the shield project and multiplying it by 110%. In a specific embodiment, if the longest number of excavation rings in all cutter replacement records in project A is 181 rings, the maximum time step of the model can be set to: 181×110%≈200 rings. To ensure that the cutter wear prediction model can cover the complete excavation history of the cutter. For cutters with an excavation distance of less than 200 rings, the end of the input sequence should be filled with -1 to ensure that the length of all input sequences is consistent. Finally, the expanded target operation data is input into Figure 3 In the hob wear prediction model shown, this filling method can effectively avoid the data misalignment problem caused by different time step lengths, ensuring that the model can be trained and predicted at a unified time step scale.

[0096] In this embodiment, after normalization, the size of the target operating data is determined according to the input requirements of the disc cutter wear prediction model. This can avoid the situation where too little target operating data is set and the actual excavation process of each disc cutter cannot be covered, thereby affecting the prediction accuracy of the model. It can also avoid the situation where too much target operating data causes the learning task of the disc cutter wear prediction model to be too complicated, thereby reducing its ability to extract key information, thereby ensuring the model prediction effect.

[0097] In an exemplary embodiment, a hob wear prediction model is obtained by the following steps: inputting historical target operating data into the hob wear prediction model to be trained, and using the hob wear prediction model to determine the hidden information at each time point in the historical target operating data, and weighting the hidden information to obtain the operating characteristics at each time point; obtaining the wear amount prediction result based on the operating characteristics at each time point; determining the model loss value based on the wear amount prediction result and the hob wear amount label, and adjusting the model parameters of the hob wear prediction model, and obtaining the hob wear prediction model when the training end conditions are met.

[0098] Specifically, the target operation data is read from the historical database and input into the hob wear prediction model to be trained. The hob wear prediction model extracts the hidden information of the historical target operation data through the long short-term memory network (LSTM). LSTM can learn the time dependency in the data through the memory and forget gate mechanism, extract the long-term and short-term features between time points, and obtain the hidden information of the historical target operation data. The attention mechanism (AM) is used to weight the hidden information and calculate the operation features at each time point. The wear amount of the operation features is calculated through the fully connected layer to obtain the wear amount prediction result. The error between the prediction result and the actual wear amount label is calculated to obtain the loss value. When the loss value is lower than the preset threshold or the number of training rounds reaches the upper limit, the training is terminated and the hob wear prediction model is generated.

[0099] In this embodiment, the dynamic features between time points are extracted through the hob wear prediction model, and a weighting mechanism is used to ensure that important features contribute more to the model output. The model parameters are adjusted through loss value feedback, and the model gradually converges to the optimal state, ensuring training efficiency and prediction performance.

[0100] In an exemplary embodiment, in step S208, the data compression size is determined based on the geological exploration data, including: determining the data compression size of the matching operation data based on the excavation length of the geological exploration data; or, determining the data compression size of the matching operation data based on the number of excavation rings of the geological exploration data; or, determining the data compression size of the matching operation data based on the number of exploration points of the geological exploration data.

[0101] In a specific embodiment, a 1000-meter shield project (625 rings) collected about 50,000 data sampling points of shield parameters, and there were about 50 geological exploration test data along the shield. Before extracting the operation characteristics, the data compression size can be determined as 1000 according to the excavation length of 1000 meters, so that the size of the segment window is determined to be 50 in the operation feature extraction; the data compression size can be determined as 625 according to the number of excavation rings of 625, so that the size of the segment window is determined to be 80 in the operation feature extraction; the data compression size can be determined as 50 according to the number of exploration points of 50, so that the size of the segment window is determined to be 1000 in the operation feature extraction. Optionally, the data compression size can also be adjusted according to the field data and the structural characteristics of the neural network.

[0102] In this embodiment, the data compression size is dynamically adjusted according to the excavation length, number of excavation rings or number of exploration points of the geological exploration data, so that the scale of the operating data matches the complexity of the working conditions, thereby highlighting the application of geological exploration data in the model and improving the prediction accuracy of the cutter wear prediction model.

[0103] In order to enable those skilled in the art to better understand the above steps, the embodiment of the present application is illustrated below by using an example, but it should be understood that the embodiment of the present application is not limited to this.

[0104] like Figure 4 As shown in the figure, the terminal obtains operation data from the shield machine in real time or in batches. These data include various operating parameters such as thrust, torque, feed speed, cutter head speed, shield displacement, etc. At the same time, corresponding geological information is extracted from the geological exploration report, such as excavation length, number of excavation rings, and exploration point distribution data, to guide the compression and feature extraction of operation data. These data are connected to the terminal through the interface and stored in a local or cloud database as the basis for subsequent data processing.

[0105] The acquired shield machine operation data is denoised. Specifically, the terminal uses the wavelet transform method to decompose the operation data into sub-signals of multiple frequency bands, and uses the hard threshold method to remove high-frequency noise and invalid signals, retaining the main working characteristic signals. The threshold is calculated based on the median absolute deviation of the decomposition coefficient, and can adaptively set appropriate noise removal conditions to improve the quality of the data. The denoised operation data is recombined and output, providing a higher signal-to-noise ratio for subsequent decomposition and feature extraction.

[0106] The terminal uses the Piecewise Aggregate Averaging (PAA) method to compress the denoised operation data. According to the characteristics of geological exploration data (such as excavation length, number of excavation rings, exploration point distribution, etc.), the data compression size is flexibly set and the size of the data window is calculated: when the excavation length is used as the unit, the excavation length is used as the data compression size and the excavation distance is segmented according to a fixed excavation distance; when the excavation ring is used as the unit, the excavation ring is used as the data compression size and the excavation ring is segmented according to a fixed excavation ring; when the exploration point is used as the unit, the excavation ring is used as the data compression size and the excavation ring is segmented according to a fixed excavation ring. The segmented data is averaged to extract the operation characteristics corresponding to the operation data and obtain the target operation data.

[0107] Before inputting the target running data, the terminal performs normalization operations on it to eliminate the dimensional differences between different eigenvalues ​​and ensure that the input data has better numerical stability during model training. The terminal calculates the current size of the normalized running data and determines the target data size based on the model's requirements for the input data length. When the normalized running data is not enough to meet the target data size, the terminal performs an expansion operation on the data. When the expanded data size reaches the target data size, the expansion stops and a complete expanded data set is generated.

[0108] The hob wear prediction model is built based on LSTM and attention mechanism (AM), which aims to extract time-dependent features and weights of key feature points in the operation data. The historical target operation data is used as input, and the hidden information is extracted through LSTM units in turn. AM assigns weights to the hidden information at each time point and generates the final operation features by weight. According to the operation features, the model generates the prediction results of the hob wear. The wear prediction results are compared with the actual hob wear labels to calculate the loss value of the model. The model calculates the gradient of the loss value through the back propagation algorithm and gradually optimizes the parameters of the model. To prevent overfitting, the model uses Dropout technology to randomly block some neuron connections. In addition, the model parameters such as learning rate, number of LSTM units and Attention dimension are dynamically adjusted through hyperparameter optimization methods. When the prediction accuracy on the validation set reaches the preset requirements, the training is terminated and the hob wear prediction model is generated.

[0109] Finally, the target operation data is input into the trained hob wear prediction model to predict the hob wear amount.

[0110] Similarly, the ring-based summation averaging (RSA) and Fourier transform denoising (FTD), rolling average denoising (RAD), autoencoder denoising (AED), wavelet transform denoising (WTD) and other denoising methods are used to process the operation data and input it into the model of cutter wear prediction. Among them, RSA is an algorithm for compressing and reconstructing shield machine operation data. The algorithm reduces the redundancy and noise of the data by grouping the operation data of each ring and calculating the mean of each ring, thereby achieving data simplification and denoising.

[0111] In a specific embodiment, the WTD+RSA method processes the operation data to obtain the following Figure 5 The results are shown in Figure 2. The FTD+PAA method processes the operating data and obtains the following Figure 6 The results shown are from Figure 5 , 6It can be seen from the local enlarged images of (a) and (b) that the blue key points extracted by PAA reconstruction fit the original signal better than the key points after RSA processing. This is because PAA is processed on the basis of retaining the phase relationship of the original sequence. Therefore, the key points better track the changes in the signal. The blue key points extracted at this location fit the original signal better than the key points after RSA processing.

[0112] In a specific embodiment, RSA and PAA are combined with four denoising methods, namely WTD, FTD, RAD and AED, to process the operation data, and the processed results are input into the hob wear prediction model of LTSM+AM, and the following can be obtained: Figure 7 The results shown in Figure 2 show that the effects of RAD and AED treatments are not ideal, and the effect of combining with RSA is closer to the effect of direct compression of Origin data. ) is about 0.87~0.88, even lower than the combination of Origin+PAA. Compared with the above two methods, the data processed by WTD and FTD significantly improve the data quality; the effect of WTD is slightly better than FTD. FTD has an outstanding ability to process stationary signals, but the shield operation signal contains non-stationary characteristics, and WTD can perform high-resolution analysis in both the time domain and the frequency domain, which makes it more advantageous in identifying and removing such local noise. On the other hand, it can be seen from the line graph in the figure that the various types of data compressed by PAA are used to train the model, and the mean squared error (MSE) performance is better than that of the data set compressed by RSA. It can be obtained through the above experiments that the shield data processed by WTD+PAA, combined with the serialized stratigraphic information obtained by stratigraphic slicing method, can achieve good results in predicting the wear of the roller cutter used for training the LSTM+AM model.

[0113] In this embodiment, the operation data is processed by PAA and WTD, and customized extraction of data compression size can be performed according to the geological exploration characteristics of each project. The processed data is combined with the LSTM-AM model, which can not only faithfully reflect the continuously changing process of shield wear, but also match the volume of geological exploration data. In model training, it can take into account the influence of geological parameters and shield operation parameters on cutter wear at the same time, which can meet the high requirements for data volume when using deep models for training and prediction.

[0114] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0115] Based on the same inventive concept, the embodiment of the present application also provides a shield machine operation data processing device for roller cutter wear prediction for implementing the shield machine operation data processing method for roller cutter wear prediction involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more embodiments of the shield machine operation data processing device for roller cutter wear prediction provided below can refer to the limitations of the shield machine operation data processing method for roller cutter wear prediction above, and will not be repeated here.

[0116] In an exemplary embodiment, Figure 8 As shown, a shield machine operation data processing device for roller cutter wear prediction is provided, comprising: an operation data acquisition and geological exploration data acquisition module 810, an operation data decomposition and sub-data extraction module 820, an operation data reconstruction and candidate data generation module 830, and a feature extraction and target data generation module 840, wherein:

[0117] The operation data acquisition and geological exploration data acquisition module 810 is used to acquire the operation data and geological exploration data of the shield machine;

[0118] The operation data decomposition and sub-data extraction module 820 is used to determine the decomposition basis function and the number of decomposition frequency bands, and decompose the operation data according to the decomposition basis function and the number of decomposition frequency bands to obtain a plurality of operation sub-data;

[0119] An operation data reconstruction and candidate data generation module 830 is used to retain the operation sub-data reflecting the working characteristics of the shield machine in the operation sub-data, and reconstruct the retained operation sub-data to obtain candidate operation data;

[0120] The feature extraction and target data generation module 840 is used to determine the data compression size according to the geological exploration data, perform feature extraction on the candidate operation data according to the data compression size, obtain target operation data, and input the target operation data into the hob wear prediction model.

[0121] In an exemplary embodiment, the feature extraction and target data generation module 840 is also used to perform feature extraction on the candidate operating data according to the data compression size to obtain the target operating data, including: segmenting and sorting the candidate operating data according to the data compression size to obtain a plurality of segmented sub-data arranged in sequence; performing mean calculation on each segmented sub-data to obtain feature operating data corresponding to each segmented sub-data, and combining the feature operating data in the order to obtain the target operating data.

[0122] In an exemplary embodiment, the operation data decomposition and sub-data extraction module 820 is also used to retain the operation sub-data that will reflect the working characteristics of the shield machine, including: determining the decomposition coefficient of the operation sub-data in the corresponding decomposition frequency band; performing median absolute deviation calculation on the decomposition coefficient corresponding to the highest decomposition frequency band to obtain a calculation result, and using the calculation result as a conditional threshold for retaining the operation sub-data; if the absolute value of the decomposition coefficient is greater than the conditional threshold, retaining the sub-data corresponding to the decomposition coefficient.

[0123] In an exemplary embodiment, the feature extraction and target data generation module 840 is also used for inputting the target operating data into the hob wear prediction model before the target operating data is input, and also includes: normalizing the target operating data to obtain normalized operating data; determining the data size of the normalized operating data, increasing the data size according to a preset ratio to obtain the target data size; expanding the normalized operating data, and stopping the expansion when the expanded data size reaches the target data size to obtain the expanded target operating data.

[0124] In an exemplary embodiment, the feature extraction and target data generation module 840 is also used to obtain the hob wear prediction model through the following steps: inputting historical target operation data into the hob wear prediction model to be trained, and the hob wear prediction model determines the hidden information at each time point in the historical target operation data, and performs weighted processing on the hidden information to obtain the operation characteristics at each time point; obtaining the wear amount prediction result based on the operation characteristics at each time point; determining the model loss value based on the wear amount prediction result and the hob wear amount label, and adjusting the model parameters of the hob wear prediction model, and obtaining the hob wear prediction model when the training end conditions are met.

[0125] In an exemplary embodiment, the feature extraction and target data generation module 840 is also used to determine the data compression size based on the geological exploration data, including: determining the data compression size that matches the operating data based on the excavation length of the geological exploration data; or, determining the data compression size that matches the operating data based on the number of excavation rings of the geological exploration data; or, determining the data compression size that matches the operating data based on the number of exploration points of the geological exploration data.

[0126] Each module in the shield machine operation data processing device for roller cutter wear prediction can be implemented in whole or in part by software, hardware and a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0127] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Fig. 9 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface and the input device are connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (NFC) or other technologies. When the computer program is executed by the processor, a shield machine operation data processing method for roller cutter wear prediction is implemented.

[0128] Those skilled in the art will understand that Fig. 9 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0129] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0130] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0131] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0132] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0133] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0134] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0135] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A shield machine operation data processing method for predicting cutter wear, characterized in that: The method comprises: Obtain shield machine operation data and geological exploration data; Determine a decomposition basis function and the number of decomposition frequency bands, and decompose the operation data according to the decomposition basis function and the number of decomposition frequency bands to obtain a plurality of operation sub-data; Among the operation sub-data, the operation sub-data reflecting the working characteristics of the shield machine are retained, and the retained operation sub-data are reconstructed to obtain candidate operation data; The data compression size is determined according to the geological exploration data, and the candidate operation data is feature extracted according to the data compression size to obtain target operation data, and the target operation data is input into the hob wear prediction model.

2. The method according to claim 1, characterized in that: The step of extracting features from the candidate operating data according to the data compression size to obtain target operating data includes: According to the data compression size, the candidate running data is segmented and sorted to obtain a plurality of segmented sub-data arranged in sequence; The mean value of each segmented sub-data is calculated to obtain the characteristic operation data corresponding to each segmented sub-data, and the characteristic operation data are combined in the order to obtain the target operation data.

3. The method according to claim 1, characterized in that The operation sub-data reflecting the working characteristics of the shield machine are retained, including: Determining the decomposition coefficient of the running sub-data in the corresponding decomposition frequency band; Perform median absolute deviation calculation on the decomposition coefficient corresponding to the highest decomposition frequency band to obtain a calculation result, and use the calculation result as a conditional threshold for retaining the running sub-data; If the absolute value of the decomposition coefficient is greater than the conditional threshold, the sub-data corresponding to the decomposition coefficient is retained.

4. The method according to claim 1, characterized in that Before inputting the target operation data into the hob wear prediction model, the method further includes: Performing normalization operation on the target operation data to obtain normalized operation data; Determining the data size of the normalized running data, and increasing the data size according to a preset ratio to obtain a target data size; The normalized operating data is expanded, and when the size of the expanded data reaches the target data size, the expansion is stopped to obtain the expanded target operating data.

5. The method according to claim 1, characterized in that The hob wear prediction model is obtained through the following steps: Inputting the historical target operation data into the hob wear prediction model to be trained, the hob wear prediction model determines the hidden information at each time point in the historical target operation data, and performs weighted processing on the hidden information to obtain the operation characteristics at each time point; and obtaining the wear amount prediction result according to the operation characteristics at each time point; According to the wear prediction result and the hob wear label, the model loss value is determined, and the model parameters of the hob wear prediction model are adjusted. When the training end condition is met, the hob wear prediction model is obtained.

6. The method according to claim 1, characterized in that Determining the data compression size according to the geological exploration data includes: Determine the data compression size matching the operation data according to the excavation length of the geological exploration data; or, Determine the data compression size matching the operation data according to the number of tunneling rings of the geological exploration data; or, The data compression size matching the operation data is determined according to the number of exploration points of the geological exploration data.

7. A shield machine operation data processing device for predicting cutter wear, characterized in that: The device comprises: Operation data acquisition and geological exploration data acquisition module, used to acquire shield machine operation data and geological exploration data; An operation data decomposition and sub-data extraction module is used to determine a decomposition basis function and the number of decomposition frequency bands, and decompose the operation data according to the decomposition basis function and the number of decomposition frequency bands to obtain a plurality of operation sub-data; An operation data reconstruction and candidate data generation module, used for retaining the operation sub-data reflecting the working characteristics of the shield machine in the operation sub-data, and reconstructing the retained operation sub-data to obtain candidate operation data; The feature extraction and target data generation module is used to determine the data compression size according to the geological exploration data, perform feature extraction on the candidate operation data according to the data compression size, obtain target operation data, and input the target operation data into the hob wear prediction model.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images