Cross-engineering geological prediction model construction method and device, equipment and medium

Abstract

The invention provides a cross-project geological prediction model construction method and device, equipment and a medium, and the method comprises the steps: obtaining first tunneling data of a selected historical project and early-stage tunneling data in a tunneling early-stage plan, and constructing a first model based on the first tunneling data and the early-stage tunneling data; obtaining medium-term tunneling data in a tunneling medium-term plan, and constructing a second model based on the first tunneling data and the early-stage tunneling data and the medium-term tunneling data after the geological data is updated; acquiring later-stage tunneling data in the tunneling later-stage plan, and constructing a third model based on the first tunneling data, the early-stage tunneling data after the geological data is updated, and the medium-stage tunneling data and the later-stage tunneling data after the geological data is updated; a required geological prediction model is constructed based on the early-stage tunneling data after the geological data is updated, the middle-stage tunneling data after the geological data is updated and the later-stage tunneling data after the geological data is updated, so that the accurate prediction of cross-engineering geology is improved, and the adaptability to various tunneling projects is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of tunneling engineering, and in particular to a method, device, equipment and medium for constructing a cross-engineering geological prediction model. Background Art

[0002] Since the ground conditions, tunnel size, TBM (tunnel boring machine) type and excavation technology of each tunnel project are different, there will be a debugging and optimization phase to find the best excavation parameters when the TBM starts to excavate a new tunnel. This process can be time-consuming and costly. At this time, transfer learning can be used to apply historical engineering TBM data and experience to another project to improve the excavation efficiency, safety and cost-effectiveness of the TBM in the new project.

[0003] The main reason why transfer learning technology is difficult to apply in cross-project TBM projects is that the prediction of target objects is mainly data-driven, heavily dependent on the quantity and quality of data, and lacks the combination with the knowledge of geotechnical engineering experts; some specific geological knowledge or operational experience may be difficult to transfer through data-driven methods, especially when facing new unknown geological structures; in the early stages of construction of new projects, there may be insufficient data, resulting in poor parameter prediction effects of traditional machine learning methods; there are large differences between the source and target fields, and the significant differences in geological conditions may make it impossible to effectively migrate the model. Summary of the invention

[0004] In view of the problems existing in the prior art, the present invention provides a method, device, equipment and medium for constructing a cross-engineering geological prediction model.

[0005] The present invention provides a method for constructing a cross-engineering geological prediction model, comprising: Acquire a tunneling plan for the target project, the tunneling plan including an early tunneling plan, a mid-term tunneling plan and a late tunneling plan; wherein tunneling data can be obtained by executing the tunneling plan, and the tunneling data includes geological data; Acquire first excavation data of the selected historical project and early excavation data in the early excavation plan, and construct a first model based on the first excavation data and the early excavation data; Using the first model to update the geological data in the previous excavation data; Acquire mid-term excavation data in the mid-term excavation plan, and construct a second model based on the first excavation data, the early excavation data after updating the geological data, and the mid-term excavation data; Using the second model to update the geological data in the mid-term excavation data; Acquire late excavation data in the late excavation plan, and construct a third model based on the first excavation data, the early excavation data after updating the geological data, the mid-term excavation data after updating the geological data, and the late excavation data; Using the third model to update the geological data in the later excavation data; The required geological prediction model is constructed based on the early excavation data after the geological data is updated, the mid-term excavation data after the geological data is updated, and the late excavation data after the geological data is updated.

[0006] The present invention also provides a device for constructing a cross-engineering geological prediction model, comprising: An acquisition module is used to acquire a tunneling plan for a target project, wherein the tunneling plan includes an early tunneling plan, a mid-term tunneling plan, and a late tunneling plan; wherein tunneling data can be obtained by executing the tunneling plan, and the tunneling data includes geological data; A first construction module is used to obtain first excavation data of a selected historical project and early excavation data in the early excavation plan, and to construct a first model based on the first excavation data and the early excavation data; A first prediction module, used to update the geological data in the previous excavation data using the first model; A second construction module is used to obtain the mid-term excavation data in the mid-term excavation plan, and construct a second model based on the first excavation data, the early excavation data after updating the geological data, and the mid-term excavation data; A second prediction module, used for updating the geological data in the mid-term excavation data using the second model; A third construction module is used to obtain the later excavation data in the later excavation plan, and construct a third model based on the first excavation data, the early excavation data after the geological data is updated, the mid-term excavation data after the geological data is updated, and the later excavation data; A third prediction module, used to update the geological data in the later excavation data using the third model; The fourth construction module is used to construct the required geological prediction model based on the early excavation data after the geological data is updated, the mid-term excavation data after the geological data is updated, and the late excavation data after the geological data is updated.

[0007] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements any of the above-mentioned methods for constructing a cross-engineering geological prediction model.

[0008] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for constructing any of the above-mentioned cross-engineering geological prediction models is implemented.

[0009] The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned methods for constructing a cross-engineering geological prediction model.

[0010] The present invention provides a method, device, equipment and medium for constructing a cross-project geological prediction model. By rationally allocating and using the first excavation data of historical projects and the excavation data of pre-, mid- and late-stage plans for excavation, a geological prediction model suitable for cross-project use is constructed, thereby improving the accurate prediction of cross-project geology and improving the adaptability to various types of excavation projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0012] Figure 1 It is a flow chart of the method for constructing a cross-engineering geological prediction model provided by the present invention.

[0013] Figure 2 It is a specific flow chart of the method for constructing a cross-engineering geological prediction model provided by the present invention.

[0014] Figure 3 It is a distribution diagram of the kernel smoothing method provided by the present invention.

[0015] Figure 4 It is a schematic diagram of the construction process of the first model provided by the present invention.

[0016] Figure 5 It is a schematic diagram of the construction process of the second model provided by the present invention.

[0017] Figure 6 It is a schematic diagram of the construction process of the third model provided by the present invention.

[0018] Figure 7 It is a schematic diagram of the structure of a device for constructing a cross-engineering geological prediction model provided by the present invention.

[0019] Figure 8 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] Combine the following Figure 1-Figure 8 The invention describes the method, device, equipment and medium for constructing a cross-engineering geological prediction model.

[0022] Figure 1 A schematic diagram showing a process flow of a method for constructing a cross-engineering geological prediction model provided by the present invention is shown in FIG. Figure 1 , the method comprises the following steps: Step 11: obtaining a tunneling plan for the target project, the tunneling plan including an early tunneling plan, a mid-term tunneling plan and a late tunneling plan; wherein, executing the tunneling plan can obtain tunneling data, and the tunneling data includes geological data; Step 12: Acquire the first excavation data of the selected historical project and the early excavation data in the early excavation plan, and construct a first model based on the first excavation data and the early excavation data; Step 13: using the first model to update the geological data in the previous excavation data; Step 14, obtaining the mid-term excavation data in the mid-term excavation plan, and constructing a second model based on the first excavation data, the early excavation data after updating the geological data, and the mid-term excavation data; Step 15: using the second model to update the geological data in the mid-term excavation data; Step 16, obtaining the later excavation data in the later excavation plan, and constructing a third model based on the first excavation data, the early excavation data after updating the geological data, the mid-term excavation data after updating the geological data, and the later excavation data; Step 17: using the third model to update the geological data in the later excavation data; Step 18: construct the required geological prediction model based on the early excavation data after the geological data is updated, the mid-term excavation data after the geological data is updated, and the late excavation data after the geological data is updated.

[0023] Regarding steps 11 to 18, it should be noted that since the ground conditions, tunnel size, TBM (tunnel boring machine) type and excavation technology of each tunnel project are different, there will be a debugging and optimization phase to find the best excavation parameters when the TBM starts to excavate a new tunnel. This process can be time-consuming and costly. At this time, the TBM data and experience of the historical project can be applied to another project through transfer learning to improve the excavation efficiency, safety and cost-effectiveness of the TBM in the new project.

[0024] The main reason why transfer learning technology is difficult to apply in cross-project TBM projects is that the prediction of target objects is mainly data-driven, heavily dependent on the quantity and quality of data, and lacks the combination with the knowledge of geotechnical engineering experts; some specific geological knowledge or operational experience may be difficult to transfer through data-driven methods, especially when facing new unknown geological structures; in the early stages of construction of new projects, there may be insufficient data, resulting in poor parameter prediction effects of traditional machine learning methods.

[0025] It can be seen that the present invention needs to construct a cross-project geological prediction model to predict the geological data of the excavation project. Because it is a cross-project scenario, the excavation data of the historical project is more than the excavation data of the target project (i.e., the newly excavated project). Therefore, as the excavation progresses, the excavation data obtained will become more, and then the fusion of historical data and target data will be formed to build a prediction model suitable for the target project. To this end, in order to obtain more excavation data, it is necessary to determine an excavation plan, which includes an early excavation plan, a mid-term excavation plan, and a late excavation plan. The excavation plan here can be a part of the excavation project. It is mainly used to provide more excavation data to build a geological prediction model suitable for new excavation projects.

[0026] In the present invention, excavation is carried out in geological engineering according to the excavation plan, and excavation data will be obtained. The excavation data includes engineering data, geological indicators and geological data. The engineering data is the field construction data obtained by using TBM (tunnel boring machine) for construction in the excavation project. The geological indicators are indicators for judging geological conditions, such as rock torsion shear excavation index TPI and field penetration index FPI. The geological data are the geological conditions encountered by the excavation project during the excavation process, such as formation conditions, surrounding rock characteristics, etc.

[0027] For different geological data, there will be different features in the engineering data that affect the geological data. In order to facilitate the description of the technical content of the present invention, the engineering data mentioned above may refer to specific features that affect the target geological data.

[0028] For a certain geological data, the model is trained based on engineering data, geological indicators and geological data containing characteristics that affect the geological data, so that the model can become more accurate in predicting the results under the geological data.

[0029] In the present invention, in order to construct the required geological prediction model, it is necessary to reasonably use the excavation data of the historical project and the excavation data of the target project. At this time, different processing stages can be divided and the excavation data of the target project can be used sequentially. Figure 2 , the construction of the model is divided into three stages, mode 1, mode 2, and mode 3. The three modes use the corresponding excavation data in the early excavation plan, mid-term excavation plan, and late excavation plan in different ways.

[0030] based on Figure 2 In the present invention, the three modes can be summarized as follows: Acquire first excavation data of the selected historical project and early excavation data in an early excavation plan, and construct a first model based on the first excavation data and the early excavation data; The first model is used to update the geological data in the previous excavation data; Acquire mid-term excavation data within the mid-term excavation plan, and construct a second model based on the first excavation data, the early excavation data after updating the geological data, and the mid-term excavation data; The second model is used to update the geological data in the mid-term excavation data; Acquire the later excavation data in the later excavation plan, and construct a third model based on the first excavation data, the early excavation data after updating the geological data, the mid-term excavation data after updating the geological data, and the later excavation data; The third model is used to update the geological data in the later excavation data; The required geological prediction model is constructed based on the early excavation data after the geological data is updated, the mid-term excavation data after the geological data is updated, and the late excavation data after the geological data is updated.

[0031] The method for constructing a cross-project geological prediction model provided by the present invention constructs a geological prediction model suitable for cross-project use by rationally allocating and using the first excavation data of historical projects and the excavation data of the pre-, mid- and late-stage plans for excavation, thereby improving the accurate prediction of cross-project geology and improving the adaptability to various types of excavation projects.

[0032] In a further method of the above method, the excavation data also includes engineering data and geological indicators. Accordingly, an explanation of the processing process of obtaining the first excavation data of the selected historical engineering and the previous excavation data in the excavation preliminary plan, and constructing the first model based on the first excavation data and the previous excavation data is as follows: Acquire historical engineering data, geological data and geological indicators of the selected historical engineering, and construct an initial geological prediction model based on the historical engineering data, geological data and geological indicators of the historical engineering; Obtain preliminary engineering data and geological indicators for the preliminary excavation plan; Based on the geological indicators of the historical project and the geological indicators of the previous project data, determine the first preliminary geological data corresponding to the preliminary excavation plan, and obtain the conversion coefficient; Convert the preliminary engineering data based on the conversion coefficient to obtain converted preliminary engineering data; The converted preliminary engineering data are predicted using the initial geological prediction model to obtain the second preliminary geological data corresponding to the preliminary excavation plan; Based on the first preliminary geological data, the second preliminary geological data and the actual preliminary geological data, the initial geological prediction model is optimized to obtain a first geological prediction model; the first geological prediction model is used as the first model.

[0033] Correspondingly, the first model is used to update the geological data in the preliminary excavation data, including: using the first geological prediction model to predict the converted preliminary engineering data to obtain third preliminary geological data corresponding to the preliminary excavation plan.

[0034] In this regard, it is necessary to explain that the process of obtaining the characteristics that affect geological data is first explained. First, a large amount of historical TBM data is obtained, and the key influencing parameters of the target variable (referring to a type of geological data, such as surrounding rock category) are extracted through the "fusion of multiple feature screening methods", that is, the characteristics that affect the geological data.

[0035] The workflow of integrating multiple feature screening methods is as follows: Use different feature screening methods to screen historical TBM data, such as random forest, variance analysis, and Pearson correlation analysis. Each method will generate a feature importance list or score, as shown in Table 1: Table 1 Feature importance scores generated by different methods

[0036] Then, to facilitate comparison on the same scale, the feature importance scores generated by each method are standardized. Furthermore, the top 10 key influencing parameters are used to conduct preliminary training of the historical model. In the process, the structural parameters of the machine learning model are optimized by Bayesian optimization to obtain preliminary prediction results of the target variable. Then, the prediction results under different key influencing parameter combinations are ranked by the error indicator RMSE. Then, according to the ranking results, the scores corresponding to different feature screening methods are assigned. The scoring rules are shown in Table 2 below: Table 2 Rating rules

[0037] Note: The basic score is 2 points. No extra points will be given to the last place, 1 point will be given to the second to last place, and so on.

[0038] Finally, according to the ranking and weight scores of different feature screening methods, the scores of each feature of each method are multiplied to obtain the top 10 features with the highest scores under comprehensive consideration of multiple methods. These features are considered to have the greatest impact on the target variable.

[0039] In the present invention, based on the TBM rotational shear rock breaking mechanism, a knowledge conversion system that is independent of other factors such as TBM equipment model and excavation diameter is derived, and the obtained "theoretical" knowledge is integrated into the machine learning model for target variable prediction.

[0040] The derivation process of the knowledge conversion system is as follows: First, the rock torsion shear excavation index is calculated based on the existing expert experience model. TPI and field penetration index FPI .

[0041]

[0042] Wherein, T is the cutter head torque, in kN·m; F is the cutter head thrust, in kN; p is the penetration, in mm / r; v is the TBM excavation speed, in mm / min; and n is the cutter head rotation speed, in r / min.

[0043] In order to avoid the influence of the cutter head diameter, the single-knife torque and single-knife thrust are used instead of the cutter head torque and cutter head thrust. The new calculation formula is as follows:

[0044] Where N is the number of tools. In actual TBM projects, when solving TPI and FPI, the above formula is often converted to:

[0045] coefficient The value of is the fitting coefficient, which directly determines the accuracy of TPI and FPI. However, the current solution of the coefficient is greatly affected by different projects and different formation conditions, and there is no unified solution. Therefore, the data is classified according to the corresponding engineering information and geological conditions. In each category, the fitting coefficient is independently obtained, and the kernel smoothing method in non-parametric statistics is used for statistical analysis to obtain a reasonable kernel function and bandwidth of the kernel smoothing method, so that the fitting coefficient is expressed in the form of a kernel smoothing method distribution. a T The kernel smoothing method is distributed as Figure 3 As shown in the figure (II, III, IV, V represent the surrounding rock categories). When new engineering data arrives, the fitting coefficients are solved in the same way, and according to the obtained results, the kernel smoothing method distribution is approximated to which category it belongs. In this way, with a small number of samples, it is possible to obtain a more accurate TPI and FPI fitting coefficient to which surrounding rock category, thereby judging the formation conditions.

[0046] Kernel smoothing is a nonparametric statistical method used to estimate the value of a function at a data point when the data distribution is uncertain. It estimates the value of an unknown function by weighted averaging the observations around the data point, where the weight is given by a kernel function. This method does not depend on the specific distribution of the data, so it is suitable for processing data with unknown distribution. The advantages of kernel smoothing include: nonparametric, flexible, and easy to implement. Therefore, it is widely used in data analysis and statistical modeling, especially when dealing with complex data distributions and local trends.

[0047] Based on historical engineering data and new engineering data, the conversion relationship between TPI and FPI can be obtained:

[0048] The conversion ratio between TPI and FPI is also a dimensionless coefficient, and the ratio is also defined as:

[0049] Where, TPI T and FPI Fis the conversion coefficient between historical project data and new project data. In this way, if the conversion coefficients FPI and TPI are obtained, FPI and TPI can be converted between historical projects and new projects, which facilitates the subsequent adjustment of model parameters.

[0050] See also Figure 4 This is a flow chart of building the first model, and Figure 2 The stage flow chart of mode 1 is shown.

[0051] The historical engineering data, geological data and geological indicators of historical engineering were selected as the data set, and the data set was split in an 8:2 ratio, one part of which was used for training the machine learning model (training set) and the other part was used for model verification (verification set). This split ensures the independence of model training and verification, which helps to improve the generalization ability of the model.

[0052] In the process of model training and verification, the prediction results of the model are comprehensively evaluated using evaluation indicators such as confusion matrix, accuracy, precision, recall, and F value. At the same time, in order to optimize the structural parameters of the model, the Bayesian optimization and K-fold cross-validation methods are used to continuously adjust and optimize the model parameters, and finally obtain an initial geological prediction model with high accuracy and low error.

[0053] Based on the prior knowledge of the TBM rotational shear rock breaking mechanism and the preliminary engineering data in the preliminary excavation plan, the kernel smoothing method distribution results of the target project TPI and FPI are obtained (that is, the first preliminary geological data is obtained), and the precise TPI and FPI fitting coefficients under a small number of samples belong to which type of geological data. At the same time, the conversion coefficient is obtained TPI T and FPI F .

[0054] Based on the conversion factor obtained TPI T and FPI F , extract and transform the features of the previous engineering data to approximate the probability distribution of the historical engineering data. Use the trained initial geological prediction model to predict the target project and obtain the predicted surrounding rock category (i.e. the second previous geological data).

[0055] At the actual construction site, the surrounding rock categories approximated by the kernel smoothing method and predicted by the historical model are compared with the actual measured data of the engineering site (i.e., the first preliminary geological data, the second preliminary geological data and the actual preliminary geological data are compared).

[0056] If the three results are consistent, the initial geological prediction model will continue to be maintained; If the surrounding rock type approximated by the kernel smoothing method is consistent with the actual measurement at the engineering site, the historical model structure parameters are adjusted in the next stage according to the kernel smoothing method distribution results, and the prediction continues; If the surrounding rock type predicted by the historical model is consistent with the actual measurement at the project site, the historical model parameters will be kept unchanged in the next stage and prediction will continue; If the three results are inconsistent, such samples will be marked to remind the driver or engineer to perform a status check.

[0057] In a further method of the above method, the process of obtaining the mid-term excavation data within the mid-term excavation plan and constructing the second model based on the first excavation data, the early excavation data after updating the geological data, and the mid-term excavation data is explained as follows: Acquire the mid-term engineering data in the mid-term excavation plan, convert the mid-term engineering data based on the conversion coefficient, and obtain the converted mid-term engineering data; Based on the historical engineering data and geological data of the historical engineering, the geological indicators of the historical engineering, the converted preliminary engineering data, the geological indicators of the preliminary engineering data, and the third preliminary geological data, a second geological prediction model is constructed; The converted mid-term engineering data are predicted using the second geological prediction model to obtain the first mid-term geological data corresponding to the mid-term excavation plan; Based on the first mid-term geological data and the actual mid-term geological data, the second geological prediction model is optimized to obtain a third geological prediction model; the third geological prediction model is used as the second model; Correspondingly, the second model is used to update the geological data in the mid-term excavation data, including: using the third geological prediction model to predict the converted mid-term engineering data to obtain second mid-term geological data corresponding to the mid-term excavation plan.

[0058] In this regard, it should be noted that in the present invention, see Figure 5 is a flowchart for building the second model, and Figure 2 The stage flow chart of Mode 2 is shown.

[0059] Re-divide the training set and validation set, the historical engineering data and geological data of historical projects, the geological indicators of historical projects as the training set, the converted preliminary engineering data, the geological indicators of preliminary engineering data, the third preliminary geological data as the validation set, and the mid-term engineering data as the application set.

[0060] In the process of model training and verification, R e(Absolute error) is used to evaluate the prediction effect of a single feature sample. If the predicted geological data of a single sample is inconsistent with the actual geological data, the next step is to incrementally adjust the model structure parameters. Otherwise, the prediction result is directly output. The detailed process is as follows: For each sample, the relationship between its predicted geological data and actual geological data is compared. If the predicted geological data are inconsistent with the actual geological data, it is considered that the prediction performance of the model is poor and the model structure parameters need to be adjusted.

[0061] Incrementally adjust model structure parameters: When the prediction error exceeds the threshold, the following comprehensive strategy is adopted to adjust the model structure: a. Increase the complexity of the model, such as adding more layers or neurons.

[0062] b. Reduce the complexity of the model, such as removing certain layers in the neural network or reducing the number of neurons in the existing hidden layers.

[0063] c. Change the activation function or optimizer of the model, such as ReLU, Sigmoid, Tanh, Adam, RMSprop, SGD, etc.

[0064] d. Adjust regularization parameters, such as L1 regularization, L2 regularization, etc., to reduce overfitting or improve model generalization ability.

[0065] After adjusting the model structure parameters, use the new structure to retrain the model. Determine the prediction category again. If the error meets the requirements, directly output the prediction result; if the error does not meet the requirements, continue to dynamically adjust the structure parameters. When the number of iterations reaches the specified requirements, stop the model operation.

[0066] When the number of iterations reaches the specified requirement but the prediction accuracy still does not meet the requirement, risk marking is performed on the samples. This may be because the TBM has encountered potential adverse risk geology or equipment abnormality, which reminds the driver or engineer to conduct a status check.

[0067] In a further method of the above method, the later excavation data in the later excavation plan is mainly obtained, and the processing process of constructing the third model is explained as follows based on the first excavation data, the early excavation data after updating the geological data, the mid-term excavation data after updating the geological data, and the later excavation data: Acquire the mid-term engineering data in the late-stage excavation plan, convert the late-stage engineering data based on the conversion coefficient, and obtain the converted late-stage engineering data; Based on the historical engineering data and geological data of the historical engineering, the geological indicators of the historical engineering, the converted preliminary engineering data, the geological indicators of the preliminary engineering data, the third preliminary geological data, the converted mid-term engineering data, the geological indicators of the mid-term engineering data, and the second mid-term geological data, a fifth geological prediction model is constructed; The converted late-stage engineering data are predicted using the fourth geological prediction model to obtain first late-stage geological data corresponding to the late-stage excavation plan; Based on the first late geological data and the actual late geological data, the fourth geological prediction model is optimized to obtain a fifth geological prediction model; the fifth geological prediction model is used as the third model; Accordingly, the third model is used to update the geological data in the later excavation data, including: using the fifth geological prediction model to predict the converted later engineering data to obtain second later geological data corresponding to the later excavation plan.

[0068] See also Figure 6 It is a flowchart for building the third model. Figure 2 The stage flow chart of mode three is shown.

[0069] Re-divide the data set, use historical engineering data and geological data of historical projects for model training, convert the preliminary engineering data, and use the geological indicators of the preliminary engineering data and the third preliminary geological data for model training, convert the mid-term engineering data, and use the geological indicators of the mid-term engineering data and the second mid-term geological data for verification of the entire model.

[0070] Perform parallel modeling on the segmented data set to obtain the prediction matrix P 1_train and P 2_train In order to better integrate historical engineering data and target engineering data, the matrix P 1_train and P 2_train Perform weighted fusion to obtain the fused probability matrix P combine = w 1×P 1_train +w 2×P 2_train , then P combine The input is fed back into the fully connected neural network. In this process, the model’s structural parameters such as learning rate, objective function, and number of training rounds are specified. The cross entropy loss function is used as the objective function.

[0071]

[0072] Among them, Y represents the real geological data in the current prediction stage.

[0073] The following operations are performed in each training epoch: the cross entropy loss between the predicted substantive data and the actual substantive data is calculated, the model parameters are optimized using Adam, and back propagation and weight updates are performed by minimizing the loss function.

[0074] After training is complete, the best weights learned will be w 1 andw 2 and save the model structure for subsequent use or further analysis. At the same time, a model loading function interface is provided to reload the model and make predictions when needed.

[0075] The obtained prediction model is applied to the later stage of tunneling planning to provide the best prediction results. As the target engineering data continues to accumulate, the prediction probability matrix P will be dynamically updated in real time. combine , and retrain the model. This can continuously introduce new data features and knowledge to improve the prediction accuracy and adaptability of the model.

[0076] Finally, the required geological prediction model is constructed based on the early excavation data after the geological data is updated, the mid-term excavation data after the geological data is updated, and the late excavation data after the geological data is updated. The required geological prediction model is used to predict the geological data of the subsequent excavation plan to improve accuracy.

[0077] The following is a description of a device for constructing a cross-engineering geological prediction model provided by the present invention. The device for constructing a cross-engineering geological prediction model described below and the method for constructing a cross-engineering geological prediction model described above can refer to each other.

[0078] Figure 7 A schematic diagram of a process flow diagram of a device for constructing a cross-engineering geological prediction model provided by the present invention is shown in FIG. Figure 7 The device includes an acquisition module 71, a first construction module 72, a first prediction module 73, a second construction module 74, a second prediction module 75, a third construction module 76, a third prediction module 77 and a fourth construction module 78, wherein: The acquisition module 71 is used to acquire a tunneling plan for a target project, wherein the tunneling plan includes an early tunneling plan, a mid-term tunneling plan, and a late tunneling plan; wherein tunneling data can be obtained by executing the tunneling plan, and the tunneling data includes geological data; A first constructing module 72 is used to obtain first excavation data of the selected historical project and early excavation data in the early excavation plan, and construct a first model based on the first excavation data and the early excavation data; A first prediction module 73, used to update the geological data in the previous excavation data using a first model; The second construction module 74 is used to obtain the mid-term excavation data in the mid-term excavation plan, and construct a second model based on the first excavation data, the early excavation data after updating the geological data, and the mid-term excavation data; A second prediction module 75, for updating the geological data in the mid-term excavation data using the second model; The third construction module 76 is used to obtain the later excavation data in the later excavation plan, and construct a third model based on the first excavation data, the early excavation data after the geological data is updated, the mid-term excavation data after the geological data is updated, and the later excavation data; A third prediction module 77, for updating the geological data in the later excavation data using the third model; The fourth construction module 78 is used to construct the required geological prediction model based on the early excavation data after the geological data is updated, the mid-term excavation data after the geological data is updated, and the late excavation data after the geological data is updated.

[0079] Since the device of the present invention has the same principle as the above method, a more detailed explanation will not be repeated here.

[0080] It should be noted that, in the embodiments of the present invention, relevant functional modules may be implemented by a hardware processor.

[0081] The device for constructing a cross-project geological prediction model provided by the present invention constructs a geological prediction model suitable for cross-project use by rationally allocating and using the first excavation data of historical projects and the excavation data of the pre-, mid- and late-stage plans for excavation, thereby improving the accurate prediction of cross-project geology and the adaptability to various types of excavation projects.

[0082] Figure 8 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 8As shown, the electronic device may include: a processor 81 (processor), a communication interface 82 (Communications Interface), a memory 83 (memory) and a communication bus 84, wherein the processor 81, the communication interface 82, and the memory 83 communicate with each other via the communication bus 84. The processor 81 may call the logic instructions in the memory 83 to execute a method for constructing a cross-project geological prediction model, the method comprising: obtaining an excavation plan for the target project, the excavation plan comprising an early excavation plan, a mid-term excavation plan, and a late excavation plan; wherein the excavation data may be obtained by executing the excavation plan, and the excavation data comprises geological data; obtaining the first excavation data of the selected historical project and the early excavation data in the early excavation plan, and constructing a first model based on the first excavation data and the early excavation data; using the first model to update the geological data in the early excavation data; obtaining the mid-term excavation data in the mid-term excavation plan, Based on the first excavation data, the early excavation data after the geological data is updated, and the mid-term excavation data, a second model is constructed; the geological data in the mid-term excavation data is updated using the second model; the later excavation data in the later excavation plan is obtained, and a third model is constructed based on the first excavation data, the early excavation data after the geological data is updated, the mid-term excavation data after the geological data is updated, and the later excavation data; the geological data in the later excavation data is updated using the third model; the required geological prediction model is constructed based on the early excavation data after the geological data is updated, the mid-term excavation data after the geological data is updated, and the later excavation data after the geological data is updated.

[0083] In addition, the logic instructions in the above-mentioned memory 83 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0084] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the method for constructing a cross-engineering geological prediction model provided by the above methods, the method comprising: obtaining an excavation plan for a target project, the excavation plan including an early excavation plan, a mid-term excavation plan, and a late excavation plan; wherein, executing the excavation plan can obtain excavation data, and the excavation data includes geological data; obtaining first excavation data of a selected historical project and early excavation data in the early excavation plan, and constructing a first model based on the first excavation data and the early excavation data; using the first model to analyze the early excavation data of the target project; and The method comprises the following steps: updating the geological data in the early excavation data; obtaining the mid-term excavation data in the mid-term excavation plan, and constructing a second model based on the first excavation data, the early excavation data after the geological data is updated, and the mid-term excavation data; updating the geological data in the mid-term excavation data by using the second model; obtaining the late excavation data in the late excavation plan, and constructing a third model based on the first excavation data, the early excavation data after the geological data is updated, the mid-term excavation data after the geological data is updated, and the late excavation data; updating the geological data in the late excavation data by using the third model; constructing the required geological prediction model based on the early excavation data after the geological data is updated, the mid-term excavation data after the geological data is updated, and the late excavation data after the geological data is updated.

[0085] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the method for constructing a cross-project geological prediction model provided by the above-mentioned methods, the method comprising: obtaining an excavation plan for a target project, the excavation plan comprising an early excavation plan, an intermediate excavation plan and a late excavation plan; wherein, executing the excavation plan can obtain excavation data, and the excavation data comprises geological data; obtaining first excavation data of a selected historical project and early excavation data in the early excavation plan, and constructing a first model based on the first excavation data and the early excavation data; and using the first model to update the geological data in the early excavation data. ; Obtain mid-term excavation data within the mid-term excavation plan, and construct a second model based on the first excavation data, the early excavation data after the geological data is updated, and the mid-term excavation data; Use the second model to update the geological data in the mid-term excavation data; Obtain late excavation data within the late excavation plan, and construct a third model based on the first excavation data, the early excavation data after the geological data is updated, the mid-term excavation data after the geological data is updated, and the late excavation data; Use the third model to update the geological data in the late excavation data; Construct the required geological prediction model based on the early excavation data after the geological data is updated, the mid-term excavation data after the geological data is updated, and the late excavation data after the geological data is updated.

[0086] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0087] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0088] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing a cross-engineering geological prediction model, characterized in that: include: Acquire a tunneling plan for the target project, the tunneling plan including an early tunneling plan, a mid-term tunneling plan and a late tunneling plan; wherein tunneling data can be obtained by executing the tunneling plan, and the tunneling data includes geological data; Acquire first excavation data of the selected historical project and early excavation data in the early excavation plan, and construct a first model based on the first excavation data and the early excavation data; Using the first model to update the geological data in the previous excavation data; Acquire mid-term excavation data in the mid-term excavation plan, and construct a second model based on the first excavation data, the early excavation data after updating the geological data, and the mid-term excavation data; Using the second model to update the geological data in the mid-term excavation data; Acquire late excavation data in the late excavation plan, and construct a third model based on the first excavation data, the early excavation data after updating the geological data, the mid-term excavation data after updating the geological data, and the late excavation data; Using the third model to update the geological data in the later excavation data; The required geological prediction model is constructed based on the early excavation data after the geological data is updated, the mid-term excavation data after the geological data is updated, and the late excavation data after the geological data is updated.

2. The method for constructing a cross-project geological prediction model according to claim 1, characterized in that: The excavation data also includes engineering data and geological indicators. Accordingly, first excavation data of the selected historical engineering and early excavation data in the early excavation plan are obtained, and a first model is constructed based on the first excavation data and the early excavation data, including: Acquire historical engineering data, geological data and geological indicators of the selected historical engineering, and construct an initial geological prediction model based on the historical engineering data, geological data and geological indicators of the historical engineering; Acquiring preliminary engineering data and geological indicators for the preliminary excavation plan within the preliminary excavation plan; Based on the geological indicators of the historical project and the geological indicators of the previous project data, determining the first previous geological data corresponding to the previous excavation plan, and obtaining a conversion coefficient; Converting the preliminary engineering data based on the conversion coefficient to obtain converted preliminary engineering data; The converted preliminary engineering data are predicted using the initial geological prediction model to obtain second preliminary geological data corresponding to the preliminary excavation plan; Based on the first preliminary geological data, the second preliminary geological data and the actual preliminary geological data, the initial geological prediction model is optimized to obtain a first geological prediction model; the first geological prediction model is used as the first model; Correspondingly, the first model is used to update the geological data in the preliminary excavation data, including: using the first geological prediction model to predict the converted preliminary engineering data to obtain third preliminary geological data corresponding to the preliminary excavation plan.

3. The method for constructing a cross-project geological prediction model according to claim 2, characterized in that: Acquiring mid-term excavation data in the mid-term excavation plan, and constructing a second model based on the first excavation data, the early excavation data after updating the geological data, and the mid-term excavation data, including: Acquire the mid-term engineering data in the mid-term excavation plan, and convert the mid-term engineering data based on the conversion coefficient to obtain the converted mid-term engineering data; Based on the historical engineering data and geological data of the historical engineering, the geological indicators of the historical engineering, the converted preliminary engineering data, the geological indicators of the preliminary engineering data, and the third preliminary geological data, a second geological prediction model is constructed; The converted mid-term engineering data are predicted using a second geological prediction model to obtain first mid-term geological data corresponding to the mid-term excavation plan; Based on the first mid-term geological data and the actual mid-term geological data, the second geological prediction model is optimized to obtain a third geological prediction model; the third geological prediction model is used as the second model; Correspondingly, the second model is used to update the geological data in the mid-term excavation data, including: using the third geological prediction model to predict the converted mid-term engineering data to obtain second mid-term geological data corresponding to the mid-term excavation plan.

4. The method for constructing a cross-project geological prediction model according to claim 3, characterized in that: Acquiring the later excavation data in the later excavation plan, and constructing a third model based on the first excavation data, the early excavation data after updating the geological data, the mid-term excavation data after updating the geological data, and the later excavation data, including: Acquire the mid-term engineering data in the late-stage excavation plan, and convert the late-stage engineering data based on the conversion coefficient to obtain the converted late-stage engineering data; Based on the historical engineering data and geological data of the historical engineering, the geological indicators of the historical engineering, the converted preliminary engineering data, the geological indicators of the preliminary engineering data, the third preliminary geological data, the converted mid-term engineering data, the geological indicators of the mid-term engineering data, and the second mid-term geological data, a fifth geological prediction model is constructed; The converted late-stage engineering data are predicted using the fourth geological prediction model to obtain first late-stage geological data corresponding to the late-stage excavation plan; Based on the first late geological data and the actual late geological data, the fourth geological prediction model is optimized to obtain a fifth geological prediction model; the fifth geological prediction model is used as the third model; Correspondingly, the third model is used to update the geological data in the later excavation data, including: using the fifth geological prediction model to predict the converted later engineering data to obtain second later geological data corresponding to the later excavation plan.

5. The method for constructing a cross-project geological prediction model according to claim 4, characterized in that: Based on the historical engineering data and geological data of the historical engineering, the geological indicators of the historical engineering, the converted preliminary engineering data, the geological indicators of the preliminary engineering data, the third preliminary geological data, the converted mid-term engineering data, the geological indicators of the mid-term engineering data, the second mid-term geological data, the fifth geological prediction model is constructed, including: Based on the historical engineering data and geological data of the historical engineering and the geological indicators of the historical engineering, a first prediction matrix is ​​obtained; Based on the converted preliminary engineering data, the geological indicators of the preliminary engineering data, the third preliminary geological data, the converted mid-term engineering data, the geological indicators of the mid-term engineering data, and the second mid-term geological data, a second prediction matrix is ​​obtained; Based on the weight relationship, the first prediction matrix and the second prediction matrix are fused to obtain a fused probability matrix; Determine structural parameters of the model based on the fusion probability matrix, wherein the structural parameters include a learning rate, an objective function, and a training round; Accordingly, based on the first late geological data and the actual late geological data, the fourth geological prediction model is optimized to obtain a fifth geological prediction model, including: Based on the objective function, the cross entropy loss between the first late geological data and the actual late geological data is calculated, and back propagation and weight update are performed with the cross entropy loss to obtain the fifth geological prediction model.

6. The method for constructing a cross-project geological prediction model according to claim 3, characterized in that: Based on the first mid-term geological data and the actual mid-term geological data, the second geological prediction model is optimized to obtain a third geological prediction model, including: Based on the first mid-term geological data and the actual mid-term geological data, the prediction error is calculated. If the prediction error does not meet the threshold condition, one or more of the following strategies are adopted to adjust the structural parameters of the model; Increase or decrease the complexity of the model; Change the activation function of the model; Tune the regularization parameter.

7. The method for constructing a cross-project geological prediction model according to claim 2, characterized in that: Based on the first preliminary geological data, the second preliminary geological data and the actual preliminary geological data, the initial geological prediction model is optimized to obtain a first geological prediction model; the first geological prediction model is used as the first model, including: If the first preliminary geological data, the second preliminary geological data and the actual preliminary geological data satisfy the preset conditions, the initial geological prediction model is adopted; If the first preliminary geological data and the actual preliminary geological data meet the preset conditions, the initial geological prediction model is optimized based on the first preliminary geological data; If the second preliminary geological data and the actual preliminary geological data meet the preset conditions, the prediction process in the next stage will maintain the initial geological prediction model.

8. The method for constructing a cross-project geological prediction model according to claim 1, characterized in that: The excavation data also includes engineering data, and the method further includes a step of acquiring the engineering data, including: Obtain historical TBM data, use different feature screening methods to screen the historical TBM data, and obtain feature importance lists corresponding to different feature screening methods; Use the top n features in the feature importance list of each feature screening method to train the model and obtain the feature screening model; The feature screening model is used to predict historical TBM data and obtain the prediction results of geological data; Obtaining error indicators between the prediction results and actual geological data, and sorting them; Assign a score value to each sorting result; Select the feature importance list of the feature screening method corresponding to the top m scoring values, and select the top s features from the feature importance list of the feature screening method corresponding to the top m scoring values ​​as the required engineering data.

9. A device for constructing a cross-engineering geological prediction model, characterized in that: include: An acquisition module is used to acquire a tunneling plan for a target project, wherein the tunneling plan includes an early tunneling plan, a mid-term tunneling plan, and a late tunneling plan; wherein tunneling data can be obtained by executing the tunneling plan, and the tunneling data includes geological data; A first construction module is used to obtain first excavation data of a selected historical project and early excavation data in the early excavation plan, and to construct a first model based on the first excavation data and the early excavation data; A first prediction module, used to update the geological data in the previous excavation data using the first model; A second construction module is used to obtain the mid-term excavation data in the mid-term excavation plan, and construct a second model based on the first excavation data, the early excavation data after updating the geological data, and the mid-term excavation data; A second prediction module, used for updating the geological data in the mid-term excavation data using the second model; A third construction module is used to obtain the later excavation data in the later excavation plan, and construct a third model based on the first excavation data, the early excavation data after the geological data is updated, the mid-term excavation data after the geological data is updated, and the later excavation data; A third prediction module, used to update the geological data in the later excavation data using the third model; The fourth construction module is used to construct the required geological prediction model based on the early excavation data after the geological data is updated, the mid-term excavation data after the geological data is updated, and the late excavation data after the geological data is updated.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method for constructing a cross-engineering geological prediction model as described in any one of claims 1-8 is implemented.

Images