Establishment of an intelligent selection auxiliary model and decision-making method for tunnel shield machines

Abstract

The present invention discloses a tunnel shield machine intelligent selection auxiliary model establishment and decision-making method, which belongs to the field of shield technology. The method uses an interval shield selection information map that integrates engineering geological longitudinal profiles, stratum physical information, shield selection information, and has undergone data expansion operations as a sample, and uses a real interval operation risk assessment map as a label to train a generative adversarial network, i.e., an image style transfer GAN model. Through a large amount of training, the model inputs the interval shield selection information map and can intelligently output the interval operation risk assessment map; through machine learning, the image style transfer GAN model internalizes the operation risk assessment information under various geological conditions and shield selection information, and displays the risk levels of different partitions within the range of the excavation face height through the interval operation risk assessment map, which is used as a reference for tunnel shield machine selection, thereby playing a role in assisting decision-making. The present invention solves the problem that the current selection is time-consuming and labor-intensive, and has low reliability by using expert experience.

Description

Technical Field

[0001] The present invention belongs to the technical field of shield machines, and more specifically, relates to an intelligent selection auxiliary model establishment and decision-making method for a tunnel shield machine. Background Art

[0002] Shield technology is of great significance in current tunnel excavation operations. When using shield operation method, the selection of shield in advance plays a decisive role in the safety and efficiency of the entire project.

[0003] The existing shield selection method is based on comprehensive considerations such as engineering geological and hydrological conditions, construction period, economy, environmental protection, and safety. The selection and configuration of the shield is a comprehensive technology involving geology, engineering, machinery, electrical and control. The main considerations include: 1) engineering geology, hydrological conditions and the size of the construction site. 2) requirements in the owner's bidding documents. 3) segment design size and block angle. 4) the advancement, adaptability and economy of the shield. 5) the reputation and performance of the shield machine manufacturer. 6) whether the shield machine can arrive at the site on schedule. Some of the factors are very complex, with high information dimensions and large amounts of information, such as geology and shield parameters; some factors are difficult to measure with data, such as the advancement and adaptability of the shield. Therefore, the current shield selection decision-making process mainly adopts the method of expert discussion and is determined by human experience. This is both time-consuming and labor-intensive, and the selection results have a certain degree of randomness.

[0004] Therefore, it is necessary to study an intelligent selection and decision-making auxiliary method for shield machines. Summary of the invention

[0005] In view of the above defects or improvement needs of the prior art, the present invention provides a method for establishing an intelligent selection auxiliary model and making decisions for a tunnel shield machine. Its purpose is to use an interval shield selection information map that integrates engineering geological longitudinal profiles, stratum physical information, and shield selection information as a sample, and a real interval operation risk assessment map as a label, to train a generative adversarial network, i.e., an image style transfer GAN model, so that the model can intelligently output an interval operation risk assessment map by inputting the interval shield selection information map, thereby assisting in the selection decision of the tunnel shield machine, thereby solving the technical problem that the prior art uses expert experience for manual selection, which is time-consuming, labor-intensive, and has low reliability.

[0006] To achieve the above object, according to one aspect of the present invention, the following technical solution is provided:

[0007] A method for establishing an intelligent selection auxiliary model for a tunnel shield machine, comprising:

[0008] The image style transfer GAN model is trained using the training set until the generator and the discriminator in the model reach a Nash equilibrium, and the gap between the interval operation risk assessment map generated by the generator and the actual interval operation risk assessment map is within a preset range, and the training is terminated;

[0009] Among them, the samples in the training set data are interval shield selection information maps, and the labels are real interval operation risk assessment maps; the interval shield selection information map is an image obtained after data expansion operation on the image obtained after integrating stratum physical information and shield selection information on the engineering geological longitudinal section map; the interval operation risk assessment map is an image obtained by generating partitions of different colors within the range of the excavation face height of the interval shield selection information map, and the different colors are used to mark the different risk levels of the shield selection information in the data sample under the geological conditions of each partition.

[0010] Preferably, the method for training the image style transfer GAN model is specifically:

[0011] (S1) taking a piece of data from the training set, inputting a true interval operation risk assessment graph in the data into the discriminator, and marking its returned value as true;

[0012] (S2) inputting the interval shield selection information diagram in the data into the generator, the generator generating an interval operation risk assessment diagram through image migration, inputting the generated interval operation risk assessment diagram into the discriminator, and marking its returned value as false;

[0013] (S3) feeding back the false value and the true value to the generator, and adjusting the generator according to the difference between the false value and the true value, so that the interval operation risk assessment map generated by the generator is closer to the real interval operation risk assessment map;

[0014] (S4) Repeat steps (S1)-(S3) until the generator and the discriminator reach a Nash equilibrium and the difference between the false and true values ​​is within a preset range, and then end the training.

[0015] Preferably, the method for obtaining the interval shield selection information diagram is specifically as follows:

[0016] (T1) The parameters of the stratum physical information are normalized and then represented using the image channel of the engineering geological longitudinal profile;

[0017] (T2) normalizing the attributes of the shield selection information, representing them with the grayscale size of a column of pixels, and splicing the column of pixels onto the engineering geological longitudinal profile obtained in step (T1);

[0018] (T3) Performing data expansion operations on the image obtained in step (T2) to obtain the interval shield selection information map.

[0019] Preferably, the image style transfer GAN model is a GAN model obtained by using a transformer framework to replace the CNN architecture in the image transfer pix2pixGAN.

[0020] Preferably, the information in the engineering geological longitudinal profile includes the mileage marks of the survey points within the interval, the elevation of each stratum and groundwater elevation passed by the survey points, and the top plate, bottom plate and rail surface positions of the tunnel.

[0021] Preferably, the parameters of the formation physical information include cohesion, internal friction angle, gravity, and static side pressure coefficient, and each parameter is represented by an image channel.

[0022] Preferably, the attributes of the shield selection information include shield diameter, shield tail gap, shield thrust, torque, main drive power, propulsion system power, rated grouting volume, pipe inner diameter, cutterhead opening rate, scraper ratio, and roller ratio.

[0023] Preferably, the data expansion operation is specifically: using differential enhancement and translation, cutout, color basic image operations to process the image obtained in step (T2).

[0024] According to another aspect of the present invention, the following technical solution is also provided:

[0025] A tunnel shield machine intelligent selection auxiliary decision method, comprising:

[0026] The interval shield machine selection information map is input into the generator of the image style transfer GAN model trained by the above method to generate an interval operation risk assessment map to assist in decision-making for tunnel shield machine selection.

[0027] Preferably, in the above-mentioned tunnel shield machine intelligent selection auxiliary decision-making method, the interval shield machine selection information diagram is obtained using the above-mentioned method.

[0028] According to another aspect of the present invention, the following technical solution is also provided:

[0029] A tunnel shield machine intelligent selection auxiliary model building system, comprising:

[0030] A training module, used to train the image style transfer GAN model using a training set until the generator and the discriminator in the model reach a Nash equilibrium, and the gap between the interval operation risk assessment map generated by the generator and the actual interval operation risk assessment map is within a preset range, and the training is terminated;

[0031] Among them, the samples in the training set data are interval shield selection information maps, and the labels are real interval operation risk assessment maps; the interval shield selection information map is an image obtained after data expansion operation on the image obtained after integrating stratum physical information and shield selection information on the engineering geological longitudinal section map; the interval operation risk assessment map is an image obtained by generating partitions of different colors within the range of the excavation face height of the interval shield selection information map, and the different colors are used to mark the different risk levels of the shield selection information in the data sample under the geological conditions of each partition.

[0032] According to another aspect of the present invention, the following technical solution is also provided:

[0033] An intelligent selection auxiliary decision-making system for a tunnel shield machine, comprising:

[0034] The decision-making module is used to input the interval shield machine selection information map into the generator of the image style transfer GAN model trained by the above method, and generate an interval operation risk assessment map to assist in the decision-making of tunnel shield machine selection.

[0035] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art:

[0036] 1. The method for establishing an intelligent selection auxiliary model for a tunnel shield machine provided by the present invention uses an interval shield selection information map that integrates engineering geological longitudinal profiles, stratum physical information, and shield selection information and has undergone data expansion operations as a sample, and uses a real interval operation risk assessment map as a label to train a generative adversarial network, namely, an image style transfer GAN model. Through a large amount of training, the model inputs the interval shield selection information map to intelligently output the interval operation risk assessment map; the real interval operation risk assessment map can be selected based on historical data or expert experience. Better data is paired with the interval shield selection information map containing engineering geological longitudinal profiles, stratum physical information, and shield selection information to form training data. Through machine learning, the image style transfer GAN model internalizes the operation risk assessment information under various geological conditions and shield selection information, and displays the risk levels of different partitions within the range of the excavation face height under the corresponding geological conditions and shield selection information through the interval operation risk assessment map, which is used as a reference for tunnel shield machine selection, thereby playing an auxiliary decision-making role. The image style transfer GAN model trained by the method of the present invention intelligently outputs the interval operation risk assessment map to assist decision-making. Compared with the prior art which adopts the expert discussion method, it greatly saves the manpower and time of shield machine selection, and the selection result is highly reliable. It is a highly intelligent shield machine selection auxiliary decision-making solution.

[0037] 2. The method for establishing a tunnel shield machine intelligent selection auxiliary model provided by the present invention uses the paired data of the interval shield selection information map and the actual interval operation risk assessment map as training data when training the image style transfer GAN model. The discriminator evaluates the actual interval operation risk assessment map and the interval operation risk assessment map generated by the generator, and feeds back the result to the generator. The generator is adjusted according to the feedback result so that the interval operation risk assessment map generated by it is closer to the actual interval operation risk assessment map; until the generator and the discriminator reach Nash equilibrium, and the gap between the interval operation risk assessment map generated by the generator and the actual interval operation risk assessment map is within a preset range, the training is terminated, thereby ensuring the accuracy of the interval operation risk assessment map intelligently output by the trained image style transfer GAN model and improving the decision accuracy.

[0038] 3. The method for establishing an intelligent selection auxiliary model for tunnel shield machines provided by the present invention integrates stratum physical information and shield selection information into the engineering geological longitudinal profile through image processing to form an interval shield selection information map, so that the interval shield selection information map to be input into the model contains geological information, stratum physical information and a preliminary shield selection plan related to the selection of tunnel shield machines, so that the operation risk assessment is in line with the actual engineering and can meet the information requirements of the selection, and the operation risk assessment map output by the model is authentic and reliable.

[0039] 3. The method for establishing an intelligent selection auxiliary model for a tunnel shield machine provided by the present invention uses a transformer framework to replace the CNN architecture in the image migration pix2pixGAN to obtain an image style transfer GAN model. The image style transfer GAN model is used to perform image migration on the basis of an interval shield selection information diagram. Different colored partitions are generated within the excavation surface on the diagram to mark the risk level of the shield selection information (i.e., the shield selection plan) under different partition geological conditions. The information display is concise and clear, with a high degree of intelligence.

[0040] 4. The method for establishing the intelligent selection auxiliary model of the tunnel shield machine provided by the present invention, the data expansion operation mainly includes using differential enhancement and three basic operators Translation, Cutout, and Color to transform the image. Translation is to make a slight offset, Cutout adds some pure white or pure black pixels to the image, and Color changes the contrast and saturation of the image; thereby improving the training effect of the image style transfer GAN model.

[0041] 5. The method for establishing the intelligent selection auxiliary model of the tunnel shield machine provided by the present invention includes the mileage marks of the survey points within the interval covered in the engineering geological longitudinal profile, the elevation of each stratum and groundwater elevation passed by the survey point, and the top plate, bottom plate, and track surface position information of the tunnel. The parameters of the stratum physical information include cohesion, internal friction angle, gravity, and static side pressure coefficient. The attributes of the shield selection information include shield diameter, shield tail gap, shield thrust, torque, main drive power, propulsion system power, rated grouting volume, pipeline inner diameter, cutterhead opening rate, scraper ratio, roller ratio, etc. The training data of the image style transfer GAN model fully fits the actual engineering and shield machine selection information requirements, and effectively meets the needs of intelligent decision-making for shield machine selection.

[0042] 6. The intelligent selection and decision-making assistance method for tunnel shield machines provided by the present invention uses a trained image style transfer GAN model to input an interval shield selection information diagram to output an interval operation risk assessment diagram for auxiliary decision-making. Compared with the prior art that uses an expert discussion method, it greatly saves manpower and time for shield machine selection, and the selection result is highly reliable. It is a highly intelligent shield machine selection and decision-making assistance solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 Schematic diagram of the process of training the image style transfer GAN model in a preferred embodiment of the present invention;

[0044] Figure 2 This is a schematic diagram of the actual use of the image style transfer GAN model in a preferred embodiment of the present invention;

[0045] Figure 3 It is a process for generating the interval operation risk assessment diagram in the preferred embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention 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 invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0047] The embodiment of the present invention first provides a method for establishing an auxiliary model for intelligent selection of a tunnel shield machine, wherein the auxiliary model refers to the image style transfer GAN model used in the present invention, which is a type of generative adversarial network. The method includes:

[0048] The image style transfer GAN model is trained using the training set until the generator and the discriminator in the model reach a Nash equilibrium, and the gap between the interval operation risk assessment map generated by the generator and the actual interval operation risk assessment map is within a preset range, and the training is terminated;

[0049] Among them, the samples in the training set data are interval shield selection information maps, and the labels are real interval operation risk assessment maps; the interval shield selection information map is an image obtained after data expansion operation on the image obtained after integrating stratum physical information and shield selection information on the engineering geological longitudinal section map; the interval operation risk assessment map is an image obtained by generating partitions of different colors within the range of the excavation face height of the interval shield selection information map, and the different colors are used to mark the different risk levels of the shield selection information in the data sample under the geological conditions of each partition.

[0050] In an embodiment of the present invention, the image style transfer GAN model (i.e., Tunnelrisk-GAN model) is a GAN model obtained by using a transformer framework to replace the CNN architecture in the image migration pix2pixGAN, thereby achieving the effect of image migration.

[0051] In the embodiment of the present invention, the training set data and the test set data are derived from the collected construction site top view and its actual engineering data, and the geological survey report file of the actual engineering. After all the data are processed, 20% are randomly selected as the test set, and the remaining 80% are used to generate the training set of the adversarial network. Among them, the engineering geological longitudinal profile itself includes the mileage mark of the survey point within a range, the elevation of each stratum passed by the survey point, the groundwater elevation, the top plate, bottom plate, and rail surface position of the tunnel;

[0052] The shield selection must also consider the physical information of the strata, including four attributes: cohesion, internal friction angle, gravity, and static side pressure coefficient. By sorting out the stratum data in the geological survey report, the above four attributes of the stratum physical information of different strata are placed in the four channels of Red, Green, Blue, and Alpha of the engineering geological longitudinal profile image, and the values ​​are normalized and recorded in each channel. Finally, on the interval shield selection information map, different strata are filled with different colors of different transparencies, thereby realizing the association between the strata and their physical properties.

[0053] For shield selection information, an engineering geological longitudinal profile image needs to correspond to a one-dimensional vector. The attributes in the vector include shield diameter, shield tail gap, shield thrust, torque, main drive power, propulsion system power, rated grouting volume, pipe inner diameter, cutter head opening rate, scraper ratio, and roller ratio, a total of 11 attributes. The 11 attributes of the shield selection information are normalized between (0, 256) and the information is represented by a column of pixel grayscale. They are spliced ​​on the back of the above-mentioned engineering geological longitudinal profile to form a complete interval shield selection information map.

[0054] Since the present invention uses the image style transfer GAN model based on the transformer framework to perform image processing, the interval shield selection information graph needs to use differential enhancement and three basic image operations of translation, cutout, and color for data expansion before entering the model. The data expansion operation is as follows:

[0055] Transformer-based architectures require a lot of data. Transformer itself is used for sentence conversion, and unless there is much larger external data for pre-training, it is not as effective as CNN in image recognition tasks. If you want to train a CNN-based GAN, the model involves almost no data enhancement. In contrast, data augmentation is crucial in transformer-based architectures. Data augmentation operations mainly include simply using differential enhancement and three basic operators Translation, Cutout, and Color to transform the image. Translation is to make a slight offset, Cutout adds some pure white or pure black pixels to the image, and Color is to change the contrast and saturation of the image.

[0056] The interval shield selection information diagram after data expansion operation can be used as sample data in the training set data.

[0057] The labels in the training set data, that is, the real interval operation risk assessment map, can be analyzed by experts based on the engineering geological longitudinal profile, stratum physical information, and shield selection information provided by the sample data, and the risk levels can be manually assigned to form an interval operation risk assessment map, so as to internalize the expert experience and knowledge into the image style transfer GAN model through machine learning.

[0058] like Figure 1 As shown, in the embodiment of the present invention, the specific method of training the image style transfer GAN model using the training set is as follows:

[0059] (S1) taking a piece of data from the training set, inputting the real interval operation risk assessment graph in the data into the discriminator, the discriminator discriminates its authenticity, and marking the value returned by the discriminator as true;

[0060] (S2) inputting the interval shield selection information diagram in the data into the generator, the generator generates an interval operation risk assessment diagram through image migration, and inputting the generated interval operation risk assessment diagram into the discriminator, the discriminator discriminates its authenticity, and marks the value returned by the discriminator as false;

[0061] (S3) feeding back the false value and the true value to the generator, and adjusting the generator according to the difference between the false value and the true value, so that the interval operation risk assessment map generated by the generator is closer to the real interval operation risk assessment map;

[0062] (S4) Repeat steps (S1)-(S3) until the generator and the discriminator reach a Nash equilibrium and the difference between the false and true values ​​is within a preset range, and then terminate the training. Save the trained image style transfer GAN model.

[0063] The generation of interval risk assessment images for shield selection is carried out with generative adversarial networks as the core. The core work is actually to translate the images using the pix2pixGAN model modified by transformer, so as to realize the migration of images from the interval shield selection information map to the interval operation risk assessment map.

[0064] In the above steps, step (S1) and step (S2) can be performed simultaneously.

[0065] The interval operation risk assessment map generated by the generator is based on the input interval shield selection information map. It uses red, orange, yellow and green colors to indicate the operation risk level at the mileage point within the range of the excavation face height. For example, high, medium, low and normal risks correspond to the colors red, orange, yellow and green respectively.

[0066] The interval operation risk assessment graph generated by the generator is expressed by a simple function:

[0067] B g =G(A),

[0068] A——Input interval shield selection information diagram

[0069] B g ——Generated interval operation risk assessment diagram

[0070] G — Transformer-built generator

[0071] In contrast, the real interval operation risk assessment diagram B r From the manually labeled training set B train middle:

[0072] B r ∈B train

[0073] For the discriminator D, its structure is also built by transformer, and its input is composed of N g With B r The discriminator of this model uses multi-scale cropping to segment the image and form small-sized sub-images. For details that need to be focused on, the model will split into smaller sub-images, which can save running memory and focus on local details based on the global information of the transformer.

[0074] The purpose of the generative adversarial network is to reach Nash equilibrium after repeated iterations. In such an iterative process, for the discriminator D, it is hoped that it can discriminate the generated interval operation risk assessment map as 0 and the real interval operation risk assessment map as 1, as shown below:

[0075]

[0076] In order to facilitate training, the log function is nested in its outer layer so that it all returns 0, that is:

[0077]

[0078] For the generator G, the interval operation risk assessment graph that is expected to be generated is identified as 0 by the discriminator D, which means that:

[0079] B g D(G(A))→0

[0080] In summary, the above optimization objectives can be written as an expression:

[0081]

[0082] Compared with traditional machine learning, which requires the definition of a specific loss function, the unsupervised adversarial learning loss L is defined by using a generative adversarial network. u =min G max D L avoids the problem of ambiguous output results caused by improper selection of loss functions.

[0083] like Figure 2 and Figure 3As shown in the figure, by using the trained image style transfer GAN model, the interval shield selection information map can be input, and the interval operation risk assessment map can be output to assist in shield selection decision-making. The interval shield selection information map input to the model comes from the engineering geological longitudinal profile, stratum physical information and shield selection information (corresponding to the proposed shield selection scheme and parameters to be evaluated), and the interval shield selection information map is obtained by the same method as the above training set sample data; the model outputs the interval operation risk assessment map that displays the risk level in different colors at the tunnel excavation depth, that is, the interval operation risk assessment map corresponding to different shield selection schemes and parameters, for users to select reference, thereby playing the role of shield selection decision-making assistance.

[0084] Finally, for evaluation, the trained image style transfer GAN model network is tested using the pre-prepared test data set B_test. ——Select appropriate evaluation parameters and indicators (Accuracy, Precision, recall, ROC curve) to evaluate the accuracy of the interval operation risk assessment output by the Tunnelrisk-GAN model.

[0085] The embodiment of the present invention further provides a tunnel shield machine intelligent selection auxiliary model establishment system, comprising:

[0086] A training module, used to train the image style transfer GAN model using a training set until the generator and the discriminator in the model reach a Nash equilibrium, and the gap between the interval operation risk assessment map generated by the generator and the actual interval operation risk assessment map is within a preset range, and the training is terminated;

[0087] Among them, the samples in the training set data are interval shield selection information maps, and the labels are real interval operation risk assessment maps; the interval shield selection information map is an image obtained after data expansion operation on the image obtained after integrating stratum physical information and shield selection information on the engineering geological longitudinal section map; the interval operation risk assessment map is an image obtained by generating partitions of different colors within the range of the excavation face height of the interval shield selection information map, and the different colors are used to mark the different risk levels of the shield selection information in the data sample under the geological conditions of each partition.

[0088] The embodiment of the present invention further provides a tunnel shield machine intelligent selection auxiliary decision-making system, comprising:

[0089] The decision-making module is used to input the interval shield machine selection information map into the generator of the image style transfer GAN model trained by the above method, and generate an interval operation risk assessment map to assist in the decision-making of tunnel shield machine selection.

[0090] The specific implementation of each module can refer to the description in the method embodiment, and the embodiment of the present invention will not be repeated.

[0091] The intelligent selection auxiliary decision-making method and system for subway tunnel shield machines provided by the embodiment of the present invention obtains a training set by collecting and preprocessing actual engineering data, and then uses the training set to generate adversarial training for the image style transfer GAN model, and finally uses the trained image style transfer GAN model for actual field deployment. After training, the method can migrate an interval risk assessment map on the interval geological longitudinal section map based on the input interval geology and shield selection information. The training data preprocessing process mainly includes data sorting, manual labeling and data expansion; the generative adversarial training process achieves the effect of generating image migration that meets the conditions through continuous optimization between the generator and the discriminator; in the actual selection decision, the trained generator can generate an interval risk assessment map to compare the performance of different shield selection schemes in the same interval. Starting from the stratum attributes, the interval geological longitudinal section and the pending selection scheme, the method uses the GAN image migration technology under the transformer framework to evaluate the risk of the interval shield, aiming to reflect the advantages and disadvantages of different selection schemes in the same interval to assist the shield selection decision. The application of the present invention can greatly reduce the human resources and time consumed in the selection of shield machines for tunnel construction, and ensure the rationality of scheme selection, thus providing strong technical support for the intelligent development of shield construction.

[0092] Matters not covered in the present invention are known technologies. It is easy for those skilled in the art to understand that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for establishing an intelligent selection auxiliary model for a tunnel shield machine, characterized in that: include: The image style transfer GAN model is trained using the training set until the generator and the discriminator in the model reach a Nash equilibrium, and the gap between the interval operation risk assessment map generated by the generator and the actual interval operation risk assessment map is within a preset range, and the training is terminated; Among them, the sample in the training set data is an interval shield selection information map, and the label is a real interval operation risk assessment map; the interval shield selection information map is an image obtained by performing data expansion operation on the image obtained by integrating stratum physical information and shield selection information on the engineering geological longitudinal section map; the interval operation risk assessment map is an image obtained by generating partitions of different colors within the range of the excavation face height of the interval shield selection information map, and the different colors are used to mark the different risk levels of the shield selection information in the sample under the geological conditions of each partition; The method for obtaining the interval shield selection information diagram is specifically as follows: (T1) The parameters of the stratum physical information are normalized and then represented using the image channel of the engineering geological longitudinal profile; (T2) normalizing the attributes of the shield selection information, representing them with the grayscale size of a column of pixels, and splicing the column of pixels onto the engineering geological longitudinal profile obtained in step (T1); (T3) performing data expansion operation on the image obtained in step (T2) to obtain the interval shield selection information map; The image style transfer GAN model is a GAN model obtained by using the transformer framework to replace the CNN architecture in the image transfer pix2pixGAN.

2. A method for establishing an intelligent selection auxiliary model for a tunnel shield machine according to claim 1, characterized in that: The method for training the image style transfer GAN model is specifically as follows: (S1) taking a piece of data from the training set, inputting a true interval operation risk assessment graph in the data into the discriminator, and marking its returned value as true; (S2) inputting the interval shield selection information diagram in the data into the generator, the generator generating an interval operation risk assessment diagram through image migration, inputting the generated interval operation risk assessment diagram into the discriminator, and marking its returned value as false; (S3) feeding back the false value and the true value to the generator, and adjusting the generator according to the difference between the false value and the true value, so that the interval operation risk assessment map generated by the generator is closer to the real interval operation risk assessment map; (S4) Repeat steps (S1)-(S3) until the generator and the discriminator reach a Nash equilibrium and the difference between the false and true values ​​is within a preset range, and then end the training.

3. A method for establishing an intelligent selection auxiliary model for a tunnel shield machine according to claim 1 or 2, characterized in that: The information in the engineering geological longitudinal profile includes the mileage marks of the survey points within the interval, the elevation of each stratum and groundwater elevation passed by the survey points, and the top plate, bottom plate and rail surface positions of the tunnel.

4. A method for establishing an intelligent selection auxiliary model for a tunnel shield machine as claimed in claim 1 or 2, characterized in that: The parameters of the formation physical information include cohesion, internal friction angle, gravity, and static side pressure coefficient, and each parameter is represented by an image channel.

5. A method for establishing an intelligent selection auxiliary model for a tunnel shield machine as claimed in claim 1 or 2, characterized in that: The attributes of the shield selection information include shield diameter, shield tail gap, shield thrust, torque, main drive power, propulsion system power, rated grouting volume, pipe inner diameter, cutterhead opening rate, scraper ratio, and roller ratio.

6. A method for establishing an intelligent selection auxiliary model for a tunnel shield machine according to claim 1, characterized in that: The data expansion operation is specifically: using differential enhancement and translation, cutout, color basic image operations to process the image obtained in step (T2).

7. A tunnel shield machine intelligent selection auxiliary decision method, characterized in that: include: The interval shield machine selection information graph is input into the generator of the image style transfer GAN model trained by any method of claims 1-6 to generate an interval operation risk assessment graph for auxiliary decision-making in tunnel shield machine selection.

8. The intelligent selection and decision-making auxiliary method for tunnel shield machine according to claim 7, characterized in that: The interval shield selection information diagram is obtained by the method described in any one of claims 1 and 3-6.

Images