Model training method, geological type identification method, device and electronic equipment
By using feature extraction and geological type identification models, the geological type in front of the TBM is automatically identified using the working parameters of the tunneling equipment, which solves the problem of low accuracy in existing technologies and achieves higher identification accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA MOBILE SHANGHAI ICT CO LTD
- Filing Date
- 2021-05-17
- Publication Date
- 2026-06-23
AI Technical Summary
In the existing technology, full-face tunnel boring machines (TBMs) have a problem with low accuracy in judging the geological type in front of the tunnel face.
By acquiring the operating parameters of the tunneling equipment, the geological type in front of the TBM is automatically identified using a feature extraction model and a geological type recognition model. The feature extraction model is trained using a sparse autoencoder to extract useful features, while the geological type recognition model is trained using a network model to achieve accurate identification of the geological type.
It improved the accuracy of geological type identification, ensured the effective use of TBM working parameters, and enhanced the accuracy of geological type judgment.
Smart Images
Figure CN115439704B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of model training, in particular to a model training method, a geological type identification method, a device and an electronic device. BACKGROUND
[0002] When a tunnel boring machine (TBM) is excavating a tunnel, it is usually necessary to determine the geological type in front of the TBM's working face and adjust the working parameters of the TBM based on the determined geological type.
[0003] In the prior art, the geological type in front of the TBM's working face is mainly determined by experience, however, this way of determining the geological type is prone to misjudgment. It can be seen that in the prior art, when determining the geological type in front of the TBM's working face, the accuracy of the determination result is low. SUMMARY
[0004] The model training method, the geological type identification method, the device and the electronic device provided by the present application can alleviate the problem of low accuracy of the determination result when determining the geological type in front of the TBM's working face in the prior art.
[0005] To solve the above technical problems, the present application is implemented as follows:
[0006] In a first aspect, the embodiments of the present application provide a model training method, comprising:
[0007] obtaining first training data, wherein the first training data comprises first working parameters of a tunneling device and a geological type corresponding to the first working parameters;
[0008] inputting the first training data into a pre-trained feature extraction model to obtain second training data output by the feature extraction model, wherein the second training data comprises first target working parameters and a geological type corresponding to the first target working parameters, and the first target working parameters are parameters obtained by the feature extraction model by performing feature extraction on the first working parameters;
[0009] training the network model by inputting the second training data into the pre-constructed network model to obtain a geological type identification model.
[0010] In a second aspect, the embodiments of the present application further provide a geological type identification method, comprising:
[0011] obtaining first working parameters;
[0012] inputting the first working parameters into a feature extraction model to obtain target working parameters;
[0013] The target working parameters are input into the geological type identification model to obtain the target geological type;
[0014] The geological type identification model is obtained by inputting the second training data into a pre-constructed network model for training.
[0015] The second training data includes a first target working parameter and a geological type corresponding to the first target working parameter. The first target working parameter is a parameter obtained by the feature extraction model from the working parameters of the tunneling equipment.
[0016] Thirdly, embodiments of this application also provide a model training apparatus, including:
[0017] The first acquisition module is used to acquire first training data, wherein the first training data includes first operating parameters of the tunneling equipment and geological types corresponding to the first operating parameters;
[0018] The first feature extraction module is used to input the first training data into a pre-trained feature extraction model and obtain the second training data output by the feature extraction model. The second training data includes a first target working parameter and a geological type corresponding to the first target working parameter. The first target working parameter is a parameter obtained by the feature extraction model from the first working parameter.
[0019] The first training module is used to input the second training data into a pre-built network model for training, thereby obtaining a geological type identification model.
[0020] Fourthly, embodiments of this application also provide a geological type identification device, including:
[0021] The third acquisition module is used to acquire the first working parameters;
[0022] The second feature extraction module is used to input the first working parameters into the feature extraction model to obtain the target working parameters;
[0023] The identification module is used to input the target working parameters into the geological type identification model to obtain the target geological type;
[0024] The geological type identification model is obtained by inputting the second training data into a pre-constructed network model for training.
[0025] The second training data includes a first target working parameter and a geological type corresponding to the first target working parameter. The first target working parameter is a parameter obtained by the feature extraction model from the working parameters of the tunneling equipment.
[0026] Fifthly, embodiments of this application provide an electronic device, including: a processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the methods described in the first and second aspects above.
[0027] In a sixth aspect, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the methods described in the first and second aspects above.
[0028] In this embodiment, a feature extraction model and a geological type identification model are trained. Subsequently, the operating parameters of the tunneling equipment can be used as input to the feature extraction model, and the output of the feature extraction model can be used as input to the geological type identification model, thus enabling automatic identification of the geological type in front of the tunneling equipment. Since the feature extraction model can fully utilize the features in the operating parameters of the tunneling equipment, the accuracy of geological type identification can be improved. Attached Figure Description
[0029] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0030] Figure 1 This is one of the flowcharts of a model training method provided in the embodiments of this application;
[0031] Figure 2 This is the second flowchart of a model training method provided in the embodiments of this application;
[0032] Figure 3 This is a flowchart of a geological type identification method provided in an embodiment of this application;
[0033] Figure 4 This is a schematic diagram of the structure of a model training device provided in an embodiment of this application;
[0034] Figure 5 This is a schematic diagram of the structure of a geological type identification device provided in an embodiment of this application;
[0035] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0036] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0037] Please see Figure 1 , Figure 1 A model training method provided in this application includes:
[0038] Step S101: Obtain first training data, wherein the first training data includes first operating parameters of the tunneling equipment and geological types corresponding to the first operating parameters;
[0039] Step S102: Input the first training data into the pre-trained feature extraction model and obtain the second training data output by the feature extraction model. The second training data includes a first target working parameter and a geological type corresponding to the first target working parameter. The first target working parameter is a parameter obtained by the feature extraction model from the first working parameter.
[0040] Step S103: Input the second training data into the pre-built network model for training to obtain the geological type recognition model.
[0041] The aforementioned tunneling equipment can be various tunnel excavation equipment, such as a tunnel boring machine (TBM). The first operating parameter can include the parameter values of multiple monitored variables. For example, multiple monitored variables can be pre-determined from the historical drilling information of the tunneling equipment, and a set of parameter values corresponding to these multiple monitored variables under the same condition can be used as the first operating parameter. It is understood that the parameter values of the monitored variables will differ depending on the geological type of the excavation. These multiple monitored variables can include: torque, cutterhead rotation speed, thrust, average advance speed, NO.1-NO.6 soil chamber pressure, screw conveyor pump outlet oil pressure, water tank temperature, oil tank temperature, cutterhead drive seal temperature, NO.1-NO.2 main bearing lubrication flow rate, pinion bearing lubrication flow rate, system voltage, system current, and system power, etc., during the tunnel excavation process.
[0042] The aforementioned first training data can be obtained from the historical drilling information of the tunneling equipment. This historical drilling information typically records the ring number, the corresponding working parameters, and the corresponding geological type during the tunneling process. Therefore, by obtaining the working parameters and geological type corresponding to the same ring number, a set of first training data can be generated. The ring number is used to indicate the drilling position of the tunneling equipment, or to indicate the drilling depth.
[0043] The aforementioned feature extraction model can be a feature extraction model pre-trained based on the working parameters of the tunneling equipment. The feature extraction model can be used to extract useful features from the first working parameters. For example, when the first working parameters include the first working parameters corresponding to 20 monitoring variables, after inputting the first working parameters into the feature extraction model, the feature extraction model can output 10 first target working parameters, thereby realizing the extraction of useful features from the first working parameters and maximizing the utilization of the working parameters of the tunneling equipment.
[0044] After extracting features from the first working parameters to obtain the first target working parameters, the geological type corresponding to the first working parameters can be determined as the geological type corresponding to the first target working parameters, thus obtaining the second training data. Then, the network model is trained using the second training data, enabling the network model to learn the correspondence between the first target working parameters and the geological types, thereby obtaining the geological type recognition model.
[0045] It is understood that the aforementioned first training data may include multiple sets of training samples, and each set of training samples includes a set of first working parameters and the geological type corresponding to the first working parameters. By inputting multiple sets of training samples into the feature extraction model, multiple sets of first target training samples can be obtained. Then, based on the multiple sets of first target training samples, the network model can be trained.
[0046] The specific process for obtaining the first training data can be as follows: Obtain historical drilling information of the tunneling equipment, and acquire multiple sets of initial data, each set corresponding to a different ring number, with each set including the working parameters and geological type corresponding to the same ring number. Then, anomaly detection can be performed on the multiple sets of initial data using the isolated forest algorithm, and initial data with abnormal working parameters can be removed to obtain multiple sets of intermediate data. Then, based on the rated values of the monitoring variables, the parameter values of the monitoring variables in each set of intermediate data can be normalized to obtain the multiple sets of training samples, which can be used as the first training data.
[0047] Each set of intermediate data may include m working parameters corresponding to m monitored variables. The parameter values of the monitored variables in each set of intermediate data can be normalized based on the following formula:
[0048]
[0049] Among them, X i X is the real-time monitoring value of the i-th (i = 1, 2, 3, ..., m) of the m monitored variables. i_min Let X be the nominal minimum value of the i-th monitored variable. i_max This represents the rated maximum value of the i-th monitored variable.
[0050] In this implementation, a feature extraction model and a geological type identification model are trained. Subsequently, the operating parameters of the tunneling equipment can be used as input to the feature extraction model, and the output of the feature extraction model can be used as input to the geological type identification model, thus enabling automatic identification of the geological type in front of the tunneling equipment. Since the feature extraction model can fully utilize the features in the operating parameters of the tunneling equipment, the accuracy of geological type identification can be improved.
[0051] Optionally, before inputting the first training data into the pre-trained feature extraction model, the method further includes:
[0052] Acquire third training data, the third training data including the second operating parameters of the tunneling equipment;
[0053] The pre-constructed sparse autoencoder is trained based on the third training data to obtain the feature extraction model, wherein the feature extraction model is a neural network model obtained after training the sparse autoencoder.
[0054] The second working parameter mentioned above can be the same as the first working parameter, or it can be different from the first working parameter. It should be noted that both the first and second working parameters correspond to the same monitored variables. For example, m monitored variables can be preset. Then, the parameter values corresponding to the m monitored variables are obtained from the historical drilling information of the tunneling equipment, thereby obtaining the first and second working parameters. Specifically, the first working parameter can be m-dimensional data composed of the parameter values of the m monitored variables, and correspondingly, the second working parameter can also be m-dimensional data composed of the parameter values of the m monitored variables.
[0055] The process of obtaining the second working parameter can be the same as the process of obtaining the first working parameter, that is, both have gone through the outlier removal process and the normalization process.
[0056] It is understood that the aforementioned third training data may include multiple second working parameters. Since the sparse autoencoder is an unsupervised machine learning algorithm, the encoder parameters are continuously adjusted by calculating the error between the autoencoder's output and the original input, ultimately training the model. The encoder can be used to compress input information and extract useful input features.
[0057] The aforementioned sparse autoencoder includes an encoder and a decoder. The training process of the sparse autoencoder is explained below with a specific embodiment: n m-dimensional data points are acquired and input into the encoder. The encoder performs dimensionality reduction processing on each m-dimensional data point to obtain n s-dimensional data points, where m is an integer greater than s, s is an integer greater than 1, and n is greater than or equal to 1. Then, the n s-dimensional data points are input into the decoder for dimensionality increase processing to obtain n m-dimensional data points. By comparing the encoder's input with the decoder's output, and adjusting the parameters of the encoder and decoder based on the comparison results, the training process of the sparse autoencoder is completed. Finally, the trained encoder is used as the feature extraction model. The subsequent feature extraction model can perform dimensionality reduction processing on the received working parameters and output useful features, thereby maximizing the utilization of the working parameters of the tunneling equipment.
[0058] In this implementation, an unsupervised algorithm is used to train the feature extraction model, and a supervised algorithm is used to train the geological type identification model. This allows for the reuse of labeled and unlabeled data to improve the accuracy and reliability of geological type identification.
[0059] Optionally, the third training data includes n first samples, where n is an integer greater than or equal to 1, and each first sample includes a set of second working parameters. The step of training the pre-constructed sparse autoencoder based on the third training data to obtain the feature extraction model includes:
[0060] The pre-constructed sparse autoencoder is trained based on the third training data to obtain the trained encoder.
[0061] Obtain n sets of second working parameters from the n first samples;
[0062] The n sets of second working parameters are respectively input into the trained encoder to obtain n sets of second target working parameters, and the n sets of second target working parameters correspond one-to-one with the n sets of second working parameters;
[0063] Clustering is performed on the n groups of second working parameters to obtain a first clustering result, and clustering is performed on the n groups of second target working parameters to obtain a second clustering result;
[0064] If the first clustering result matches the second clustering result, the trained encoder is output as the feature extraction model.
[0065] If the first clustering result does not match the second clustering result, the sparse autoencoder is retrained.
[0066] In this embodiment of the application, after training the sparse autoencoder based on the third training data, the trained encoder is further verified based on the n sets of second working parameters to determine whether the trained encoder is effective. Only when the trained encoder is effective will it be output as the feature extraction model. In this way, it can be ensured that the trained feature extraction model can accurately identify useful features in the data to be identified, thereby improving the accuracy of subsequent geological type identification.
[0067] Specifically, since each second working parameter includes parameter values of multiple monitored variables, each second working parameter can be considered as multidimensional data. Accordingly, the n sets of second working parameters are input into the trained encoder, which performs dimensionality reduction processing on each set of n second working parameters to obtain n sets of second target working parameters. These second target working parameters can also be considered as multidimensional data, with their dimension being smaller than that of the second working parameters. This allows for clustering of the n sets of second working parameters to obtain a first clustering result including the distribution of the n sets of second working parameters. Similarly, clustering can be performed on the n sets of second target working parameters to obtain a second clustering result including the distribution of the n sets of second target working parameters. By comparing the distributions of the first and second clustering results, it can be determined whether the first and second clustering results match. For example, if their clustering characteristics are similar, they are considered a match; conversely, if their clustering characteristics are significantly different, they are considered a mismatch.
[0068] The aforementioned retraining of the sparse autoencoder when the first clustering result and the second clustering result do not match can refer to: re-acquiring fourth training data, which includes the third operating parameters of the tunneling equipment. The process of acquiring the fourth training data is similar to that of acquiring the third training data, and the training process of the sparse autoencoder based on the fourth training data is similar to that based on the third training data. To avoid repetition, it will not be described in detail here.
[0069] Optionally, the second working parameters include m-dimensional data composed of m first sub-parameters, and the second target working parameters include s-dimensional data composed of s second sub-parameters, where m is greater than s and s is an integer greater than or equal to 1. The step of clustering the n sets of second working parameters to obtain a first clustering result, and clustering the n sets of second target working parameters to obtain a second clustering result, includes:
[0070] The n sets of second working parameters are dimensionality reduced based on a preset algorithm to obtain n sets of k-dimensional first data. The n sets of second target working parameters are dimensionality reduced based on the preset algorithm to obtain n sets of k-dimensional second data, where k is a positive integer less than or equal to 3.
[0071] Cluster the n k-dimensional first data to obtain the first clustering result, and cluster the n k-dimensional second data to obtain the second clustering result.
[0072] The aforementioned preset algorithm can be the t-SNE algorithm, and k can be 1, 2, or 3. The following example uses k equal to 3 to illustrate the clustering process in this embodiment. Correspondingly, m can be greater than 3, and s can also be greater than 3.
[0073] In this implementation, n sets of first data are obtained by dimensionality reduction of n sets of second working parameters, and n sets of second data are obtained by dimensionality reduction of n sets of second target working parameters. Both the first and second data are coordinates of 3D data points. Thus, plotting the n coordinate points corresponding to the n sets of first data in a coordinate system yields the first clustering result; similarly, plotting the n coordinate points corresponding to the n sets of second data in a coordinate system yields the second clustering result. The first and second clustering results can be plotted on the same 3D coordinate system to determine if they match. This matching is determined by whether the distance between the cluster centers in the first and second clustering results is within a preset range. If the distance is within the preset range, the first and second clustering results are considered a match; conversely, if the distance is not within the preset range, the first and second clustering results are considered a mismatch.
[0074] Optionally, the step of inputting the second training data into a pre-built network model for training to obtain a geological type identification model includes:
[0075] The second training data is input into the pre-built network model for training to obtain the trained network model.
[0076] The first working parameters are input into the feature extraction model to obtain the third target working parameters;
[0077] The third target working parameters are input into the trained network model to obtain the first recognition result;
[0078] The accuracy of the first identification result is determined based on the geological type corresponding to the first working parameter;
[0079] If the accuracy is greater than a preset value, the trained network model will be output as the geological type identification model.
[0080] If the accuracy is less than or equal to the preset value, the network model is retrained; wherein the preset value can be any value greater than 80%.
[0081] In this embodiment of the application, after training the network model based on the second training data, the accuracy of the first identification result is verified based on the geological type corresponding to the first working parameter. That is, the geological type identified by the model is verified based on the real geological type of the first working parameter to determine the accuracy of the model identification. Only when the accuracy is higher than the preset value will the trained network model be output as the geological type identification model. In this way, the accuracy of subsequent geological type identification can be improved.
[0082] Specifically, since the geological type corresponding to the first working parameter is data obtained from the historical drilling information of the tunneling equipment, the geological type corresponding to the first working parameter can be regarded as the actual geological type corresponding to the first working parameter. The first identification result is the identification result obtained by inputting the first working parameter into a feature extraction model and using the output of the feature extraction model as the input to the trained network model; that is, the first identification result model obtains the identification result based on the first working parameter for geological type identification. Therefore, by comparing the geological type corresponding to the first working parameter with the first identification result model, the accuracy of the identification by the trained network model can be determined.
[0083] The aforementioned retraining of the network model when the accuracy is less than or equal to the preset value can refer to: reacquiring fifth training data, which includes the fifth operating parameters of the tunneling equipment and the geological type corresponding to the fifth operating parameters. The process of acquiring the fifth training data is similar to that of acquiring the second training data, and the process of training the network model based on the fifth training data is similar to that based on the second training data. To avoid repetition, it will not be described in detail here.
[0084] Optionally, determining the accuracy of the first identification result based on the geological type corresponding to the first working parameter includes:
[0085] Based on the first identification result and the geological type corresponding to the first working parameter, a confusion matrix is generated;
[0086] The accuracy is determined based on the confusion matrix.
[0087] The confusion matrix, also known as the error matrix, is a standard format for representing accuracy evaluation, expressed as an n x n matrix. Specific evaluation metrics include overall accuracy, mapping accuracy, and user accuracy, which reflect the accuracy of image classification from different perspectives. In artificial intelligence, the confusion matrix is a visualization tool, particularly used in supervised learning; in unsupervised learning, it is generally called a matching matrix. In image accuracy evaluation, it is mainly used to compare classification results with actual measured values, displaying the accuracy of the classification results within a confusion matrix. The confusion matrix is calculated by comparing the position and classification of each measured pixel with the corresponding position and classification in the classified image.
[0088] In this implementation, a confusion matrix is generated based on the actual results and predicted results of the geological type corresponding to the first working parameter. The accuracy of the first identification result is determined by using the precision, recall, and F1 score calculated from the confusion matrix. The F1 score is a comprehensive evaluation of the classification accuracy.
[0089] Please see Figure 2 The following section provides a specific example to further illustrate the above model training process, such as... Figure 2 As shown, the above model training process may include the following steps: First, initial data is obtained, including the operating parameters of the tunneling equipment. Then, the initial data is preprocessed, primarily including outlier removal and normalization. The specific processing steps can be found in the above embodiments. After preprocessing, the operating parameters are obtained. Based on these parameters, an unlabeled dataset (i.e., the third training data in the above embodiments) is generated. Based on the operating parameters and the corresponding geological type labels, a labeled dataset (i.e., the first training data in the above embodiments) is generated. The encoder and decoder are trained based on the unlabeled dataset. The encoder and decoder are respectively sparse autoencoders, and their specific training processes can be found in the above embodiments. After training, the labeled dataset is input into the encoder to obtain a labeled feature set (i.e., the second training data in the above embodiments). Then, a pre-built network model is trained based on the second training data to obtain a geological type recognition model. This network model can be a deep learning model. This completes the model training process.
[0090] Please see Figure 3 This application also provides a geological type identification method, the method comprising:
[0091] Step S301: Obtain the first working parameters;
[0092] Step S302: Input the first working parameters into the feature extraction model to obtain the target working parameters;
[0093] Step S303: Input the target working parameters into the geological type identification model to obtain the target geological type;
[0094] The second training data includes a first target working parameter and a geological type corresponding to the first target working parameter. The first target working parameter is a parameter obtained by the feature extraction model from the working parameters of the tunneling equipment.
[0095] The feature extraction model can be the one obtained by training a sparse autoencoder based on the third training data in the above embodiments. The geological type identification model can be the one obtained by training a pre-built network model based on the second training data in the above embodiments. The specific training process is similar to that in the above embodiments and will not be repeated here to avoid repetition.
[0096] In this embodiment, by inputting the first working parameter into the feature extraction model and using the output of the feature extraction model as the input to the geological type identification model, the geological type corresponding to the working parameters of the tunneling equipment can be automatically identified based on the working parameters of the tunneling equipment. Thus, by simply acquiring the working parameters of the tunneling equipment in real time, the geological type in front of the tunneling equipment's face can be identified in real time.
[0097] Please see Figure 4 , Figure 4 This application provides a model training device 400, which includes:
[0098] The first acquisition module 401 is used to acquire first training data, wherein the first training data includes first operating parameters of the tunneling equipment and geological types corresponding to the first operating parameters;
[0099] The first feature extraction module 402 is used to input the first training data into a pre-trained feature extraction model and obtain the second training data output by the feature extraction model. The second training data includes a first target working parameter and a geological type corresponding to the first target working parameter. The first target working parameter is a parameter obtained by the feature extraction model from the first working parameter.
[0100] The first training module 403 is used to input the second training data into a pre-built network model for training to obtain a geological type identification model.
[0101] Optionally, the device further includes:
[0102] The second acquisition module is used to acquire third training data, the third training data including the second operating parameters of the tunneling equipment;
[0103] The second training module is used to train the pre-constructed sparse autoencoder based on the third training data to obtain the feature extraction model, wherein the feature extraction model is a neural network model obtained after training the sparse autoencoder.
[0104] Optionally, the third training data includes n first samples, where n is an integer greater than or equal to 1, and each first sample includes a set of second working parameters. The second training module includes:
[0105] The first training submodule is used to train the pre-constructed sparse autoencoder based on the third training data to obtain the trained encoder.
[0106] The first acquisition submodule is used to acquire n sets of second working parameters from the n first samples;
[0107] The second acquisition submodule is used to input the n sets of second working parameters into the trained encoder to obtain n sets of second target working parameters, wherein the n sets of second target working parameters correspond one-to-one with the n sets of second working parameters.
[0108] The clustering submodule is used to perform clustering processing on the n groups of second working parameters to obtain a first clustering result, and to perform clustering processing on the n groups of second target working parameters to obtain a second clustering result;
[0109] The first output submodule is used to output the trained encoder as the feature extraction model when the first clustering result matches the second clustering result.
[0110] The first training submodule is further configured to retrain the sparse autoencoder if the first clustering result does not match the second clustering result.
[0111] Optionally, the second working parameter includes m-dimensional data composed of m first sub-parameters, and the second target working parameter includes s-dimensional data composed of s second sub-parameters, where m is greater than s and s is an integer greater than or equal to 1. The clustering submodule includes:
[0112] The processing unit is configured to perform dimensionality reduction processing on the n sets of second working parameters based on a preset algorithm to obtain n sets of k-dimensional first data, and to perform dimensionality reduction processing on the n sets of second target working parameters based on the preset algorithm to obtain n sets of k-dimensional second data, where k is a positive integer less than or equal to 3;
[0113] A clustering unit is used to cluster the n k-dimensional first data to obtain the first clustering result, and to cluster the n k-dimensional second data to obtain the second clustering result.
[0114] Optionally, the first training module 403 includes:
[0115] The second training submodule is used to input the second training data into a pre-built network model for training, and obtain the trained network model.
[0116] The third acquisition submodule is used to input the first working parameters into the feature extraction model to obtain the third target working parameters;
[0117] The fourth acquisition submodule is used to input the third target working parameters into the trained network model to obtain the first recognition result;
[0118] The determination submodule is used to determine the accuracy of the first identification result based on the geological type corresponding to the first working parameter;
[0119] The second output submodule is used to output the trained network model as the geological type identification model when the accuracy is greater than a preset value.
[0120] The second training submodule is further configured to retrain the network model if the accuracy is less than or equal to the preset value.
[0121] Optionally, the determining submodule includes:
[0122] The generation unit is used to generate a confusion matrix based on the geological type corresponding to the first identification result and the first working parameters;
[0123] A determining unit is used to determine the accuracy based on the confusion matrix.
[0124] The aforementioned model training device 400 can achieve Figure 1 To avoid repetition, the various processes in the method embodiments shown will not be described again here.
[0125] Please see Figure 5 , Figure 5 A geological type identification device 500 provided in this application embodiment includes:
[0126] The third acquisition module 501 is used to acquire the first working parameters;
[0127] The second feature extraction module 502 is used to input the first working parameters into the feature extraction model to obtain the target working parameters;
[0128] The identification module 503 is used to input the target working parameters into the geological type identification model to obtain the target geological type;
[0129] The geological type identification model is obtained by inputting the second training data into a pre-constructed network model for training.
[0130] The second training data includes a first target working parameter and a geological type corresponding to the first target working parameter. The first target working parameter is a parameter obtained by the feature extraction model from the working parameters of the tunneling equipment.
[0131] The aforementioned geological type identification device 500 can achieve Figure 2 To avoid repetition, the various processes in the method embodiments shown will not be described again here.
[0132] This application also provides an electronic device, including: a processor, a memory, and a program stored in the memory and executable on the processor. When the program is executed by the processor, it implements the various processes of the above-described model training method and geological type identification method embodiments and achieves the same technical effect. To avoid repetition, it will not be described again here.
[0133] See Figure 6 As shown in the figure, this application embodiment also provides an electronic device, including a bus 601, a transceiver 602, an antenna 603, a bus interface 604, a processor 605, and a memory 606.
[0134] exist Figure 6In this document, a bus architecture (represented by bus 601) is used. Bus 601 can include any number of interconnected buses and bridges, linking various circuits including one or more processors represented by processor 605 and memory represented by memory 606. Bus 601 can also link various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. Bus interface 604 provides an interface between bus 601 and transceiver 602. Transceiver 602 can be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by processor 605 is transmitted over a wireless medium via antenna 603, which further receives data and transmits data to processor 605.
[0135] Processor 605 manages bus 601 and general processing, and also provides various functions, including timing, peripheral interface, voltage regulation, power management, and other control functions. Memory 606 can be used to store data used by processor 605 during operation.
[0136] Optionally, the processor 605 can be a CPU, ASIC, FPGA, or CPLD.
[0137] This application also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the various processes of the above-described method embodiments and achieves the same technical effects. To avoid repetition, it will not be described again here. The computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.
[0138] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.
[0139] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or second terminal device, etc.) to execute the methods described in the various embodiments of this application.
[0140] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.
Claims
1. A model training method, characterized in that, include: Obtain third training data, which includes the second operating parameters of the tunneling equipment. The third training data includes n first samples, where n is an integer greater than or equal to 1, and each first sample includes a set of the second operating parameters. The pre-constructed sparse autoencoder is trained based on the third training data to obtain the trained encoder. Obtain n sets of second working parameters from the n first samples; The n sets of second working parameters are respectively input into the trained encoder to obtain n sets of second target working parameters, and the n sets of second target working parameters correspond one-to-one with the n sets of second working parameters; Clustering is performed on the n groups of second working parameters to obtain a first clustering result, and clustering is performed on the n groups of second target working parameters to obtain a second clustering result; If the first clustering result matches the second clustering result, the trained encoder is output as a feature extraction model. If the first clustering result does not match the second clustering result, the sparse autoencoder is retrained. Acquire first training data, wherein the first training data includes first operating parameters of the tunneling equipment and geological types corresponding to the first operating parameters; The first training data is input into the pre-trained feature extraction model to obtain the second training data output by the feature extraction model. The second training data includes a first target working parameter and a geological type corresponding to the first target working parameter. The first target working parameter is a parameter obtained by the feature extraction model from the first working parameter. The second training data is input into a pre-built network model for training to obtain a geological type identification model.
2. The method according to claim 1, characterized in that, The second working parameters include m-dimensional data composed of m first sub-parameters, and the second target working parameters include s-dimensional data composed of s second sub-parameters, where m is greater than s and s is an integer greater than or equal to 1. The process of clustering the n sets of second working parameters to obtain a first clustering result, and clustering the n sets of second target working parameters to obtain a second clustering result, includes: The n sets of second working parameters are dimensionality reduced based on a preset algorithm to obtain n sets of k-dimensional first data. The n sets of second target working parameters are dimensionality reduced based on the preset algorithm to obtain n sets of k-dimensional second data, where k is a positive integer less than or equal to 3. Cluster the n groups of k-dimensional first data to obtain the first clustering result, and cluster the n groups of k-dimensional second data to obtain the second clustering result.
3. The method according to claim 1, characterized in that, The step of inputting the second training data into a pre-built network model for training to obtain a geological type identification model includes: The second training data is input into the pre-built network model for training to obtain the trained network model. The first working parameters are input into the feature extraction model to obtain the third target working parameters; The third target working parameters are input into the trained network model to obtain the first recognition result; The accuracy of the first identification result is determined based on the geological type corresponding to the first working parameter; If the accuracy is greater than a preset value, the trained network model will be output as the geological type identification model. If the accuracy is less than or equal to the preset value, the network model is retrained.
4. The method according to claim 3, characterized in that, Determining the accuracy of the first identification result based on the geological type corresponding to the first working parameter includes: Based on the first identification result and the geological type corresponding to the first working parameter, a confusion matrix is generated; The accuracy is determined based on the confusion matrix.
5. A method for identifying geological types, characterized in that, include: Obtain the first working parameters; The first working parameters are input into the feature extraction model to obtain the target working parameters; The target working parameters are input into the geological type identification model to obtain the target geological type; The geological type identification model is obtained by inputting the second training data into a pre-constructed network model for training. The geological type identification model is a model trained based on the model training method described in any one of claims 1 to 4. The second training data includes a first target working parameter and a geological type corresponding to the first target working parameter. The first target working parameter is a parameter obtained by the feature extraction model from the working parameters of the tunneling equipment.
6. A model training device, characterized in that, include: The second acquisition module is used to acquire the third training data, which includes the second operating parameters of the tunneling equipment. The third training data includes n first samples, where n is an integer greater than or equal to 1, and each first sample includes a set of the second operating parameters. The second training module includes: The first training submodule is used to train the pre-constructed sparse autoencoder based on the third training data to obtain the trained encoder. The first acquisition submodule is used to acquire n sets of second working parameters from the n first samples; The second acquisition submodule is used to input the n sets of second working parameters into the trained encoder to obtain n sets of second target working parameters, wherein the n sets of second target working parameters correspond one-to-one with the n sets of second working parameters. The clustering submodule is used to perform clustering processing on the n groups of second working parameters to obtain a first clustering result, and to perform clustering processing on the n groups of second target working parameters to obtain a second clustering result; The first output submodule is used to output the trained encoder as a feature extraction model when the first clustering result matches the second clustering result. The first training submodule is further configured to retrain the sparse autoencoder if the first clustering result does not match the second clustering result. The first acquisition module is used to acquire first training data, wherein the first training data includes first operating parameters of the tunneling equipment and geological types corresponding to the first operating parameters; The first feature extraction module is used to input the first training data into the pre-trained feature extraction model and obtain the second training data output by the feature extraction model. The second training data includes a first target working parameter and a geological type corresponding to the first target working parameter. The first target working parameter is a parameter obtained by the feature extraction model from the first working parameter. The first training module is used to input the second training data into a pre-built network model for training, thereby obtaining a geological type identification model.
7. A geological type identification device, characterized in that, include: The third acquisition module is used to acquire the first working parameters; The second feature extraction module is used to input the first working parameters into the feature extraction model to obtain the target working parameters; The identification module is used to input the target working parameters into the geological type identification model to obtain the target geological type; The geological type identification model is obtained by inputting the second training data into a pre-constructed network model for training. The geological type identification model is a model trained based on the model training method described in any one of claims 1 to 4. The second training data includes a first target working parameter and a geological type corresponding to the first target working parameter. The first target working parameter is a parameter obtained by the feature extraction model from the working parameters of the tunneling equipment.
8. An electronic device, characterized in that, include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the method as described in any one of claims 1 to 5.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the method as described in any one of claims 1 to 5.