TBM operation parameter multi-objective optimization method based on mixed deep learning

Through a multi-objective optimization method based on hybrid deep learning, the mechanical parameter information of the TBM excavation process is used to automatically optimize the cutter wheel speed and total propulsion force, which solves the problem of insufficient scientific adjustment of TBM operating parameters, improves the excavation efficiency and reduces energy This has achieved more efficient tunnel construction.

CN118917181APending Publication Date: 2024-11-08SINOHYRDO ENG BUREAU 3 CO LTD +1
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202410949279.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-16
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The lack of scientific nature of TBM's operating parameters adjustments lead to low boring efficiency, high energy consumption and increased construction costs, and the lack of quantitative evaluation depends on subjective experience.

Method used

Using a multi-objective optimization method based on hybrid deep learning, a multi-objective prediction model considering spatiotemporal characteristics is constructed by obtaining mechanical parameter information of the TBM excavation process. The spatial and time series information is extracted using GCN and LSTM, and combined with the NSGA-II algorithm, it is automatically optimized. The cutting wheel speed and total propulsion force increase the excavation speed and reduce the specific energy of rock breaking.

Benefits of technology

It realizes more objective, fast and effective operation parameter optimization, improves TBM construction efficiency, reduces energy consumption costs, and avoids construction accidents.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118917181A_ABST
    Figure CN118917181A_ABST
Patent Text Reader

Abstract

The invention discloses a TBM operation parameter multi-objective optimization method based on hybrid deep learning, and the method comprises the steps: obtaining mechanical parameter information recorded in real time in a TBM tunneling process, carrying out the preprocessing of data to construct a database, converting the data into graph data, and dividing the graph data into a data set; on the basis of hybrid deep learning, taking mechanical parameter information as input, and taking current tunneling speed and rock breaking specific energy as output, constructing a multi-target prediction model considering spatio-temporal characteristic information; on the basis of the established multi-target prediction model considering the spatial-temporal characteristic information, suggested values of the cutterhead rotating speed and the total propulsive force of the TBM with the aims of improving the tunneling speed and reducing the rock breaking specific energy are given. According to the method, the TBM operation parameters can be intelligently optimized according to different construction conditions, and the method is of great significance to TBM long-distance construction under complex geology.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of TBM operating parameter optimization, and specifically relates to a TBM operating parameter multi-objective optimization method based on hybrid deep learning. Background Art

[0002] Tunnel Boring Machine (TBM) is a heavy-duty tunnel construction equipment with the advantages of high construction efficiency, good hole quality, small construction disturbance, safety and reliability, especially in the construction of deep buried long-distance tunnels. It is worth noting that the effective excavation of TBM depends on the matching degree of operating parameters and geological conditions. If the matching degree is low, it will lead to low TBM excavation efficiency and even cause problems such as machine jamming, abnormal tool wear, and increased construction costs. On-site staff mainly adjust operating parameters based on construction experience, which is subjective and lacks quantitative evaluation. Therefore, establishing a scientific TBM operating parameter adjustment method is of great significance to improving TBM excavation efficiency, reducing energy consumption costs, and avoiding construction accidents. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a multi-objective optimization method for TBM operating parameters based on hybrid deep learning in response to the deficiencies in the above-mentioned prior art. Only mechanical parameters need to be input, and partial response parameters are used to indirectly characterize the properties of the rock mass. Multi-objective optimization is performed using intelligent algorithms, which can more objectively, quickly and effectively search for the best result. Taking multiple indicators as optimization targets is more comprehensive than a single target, and realizes intelligent optimization of operating parameters while considering speed and energy, which is beneficial to the construction of TBM.

[0004] In order to solve the above technical problems, the first aspect of the present invention proposes a multi-objective optimization method for TBM operating parameters based on hybrid deep learning, which is characterized by comprising:

[0005] Obtain the mechanical parameter information recorded in real time during the TBM excavation process, pre-process the data to build a database, convert the data into graph data and divide the data set;

[0006] Based on hybrid deep learning, with mechanical parameter information as input and current tunneling speed and rock breaking specific energy as output, a multi-objective prediction model considering spatiotemporal feature information is constructed;

[0007] Based on the established multi-objective prediction model considering the temporal and spatial characteristic information, the recommended values ​​of the cutter head speed and total thrust of the TBM are given with the goal of increasing the excavation speed and reducing the rock breaking specific energy.

[0008] Furthermore, the mechanical parameter information includes the cutter disc speed, the cutter disc torque, the total thrust, the main thrust cylinder pressure and the auxiliary thrust cylinder pressure.

[0009] Furthermore, the isolation forest algorithm is used to quickly and accurately process database outliers.

[0010] Furthermore, the multi-target prediction model considering spatiotemporal feature information includes a GCN graph convolutional neural network layer, which converts tabular data into graph data to extract spatial information.

[0011] Furthermore, the multi-objective prediction model considering spatiotemporal characteristic information includes an LSTM long short-term memory network layer, which extracts time series information and realizes simultaneous prediction of excavation speed and rock-breaking specific energy.

[0012] Furthermore, during the excavation process, newly recorded data can be dynamically added to the training data set to update the parameters in the multi-objective prediction model that considers spatiotemporal feature information and enhance the model performance.

[0013] Furthermore, the process of providing recommended values ​​of the cutter head rotation speed and total thrust of the TBM with the goal of increasing the excavation speed and reducing the rock-breaking energy ratio based on the established multi-objective prediction model considering the spatiotemporal characteristic information is as follows: using the NSGA-II fast elite multi-objective genetic algorithm, based on actual engineering needs, to achieve intelligent optimization and decision-making of the excavation speed and rock-breaking energy ratio, and provide corresponding recommended values ​​of the cutter head rotation speed and total thrust.

[0014] Another aspect of the present invention provides a TBM operating parameter multi-objective optimization system based on hybrid deep learning, which is characterized by comprising:

[0015] The data collection and processing module is used to obtain the mechanical parameter information recorded in real time during the TBM excavation process, pre-process the data to build a database, convert the data into graph data and divide the data set;

[0016] The multi-objective prediction module is used to use the mechanical parameter information as input and realize the simultaneous prediction of the current excavation speed and rock breaking efficiency based on the hybrid deep learning algorithm;

[0017] The multi-objective optimization module is based on a multi-objective prediction model. It performs global intelligent optimization of operating parameters according to the actual project requirements for excavation speed and rock-breaking specific energy, and automatically gives recommended values ​​for cutterhead speed and total thrust.

[0018] On the other hand, the present invention also provides an electronic device, characterized in that the electronic device includes a processor and a memory, the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the multi-objective optimization method of TBM operating parameters based on hybrid deep learning.

[0019] On the other hand, the present invention also provides a computer-readable storage medium, characterized in that the storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set are loaded and executed by a processor to implement a multi-objective optimization method for TBM operating parameters based on hybrid deep learning.

[0020] Compared with the prior art, the present invention has the following advantages:

[0021] 1. The present invention does not use rock mass information as modeling input parameters, but only relies on mechanical parameters recorded in real time by TBM as input, in which response parameters are used to indirectly characterize rock mass information. This avoids on-site coring and indoor test measurement processes, effectively improving construction efficiency.

[0022] 2. The present invention develops a GCN-LSTM hybrid deep learning model, which effectively extracts the spatiotemporal information in the data and achieves multi-target prediction more accurately compared with existing machine learning methods such as random forest, support vector regression, and feedforward neural network.

[0023] 3. The present invention simultaneously considers two key performance indicators of TBM, namely, excavation speed and rock-breaking specific energy, and adopts NSGA-II fast elite multi-objective genetic algorithm to solve multi-objective optimization problems, thereby achieving the purpose of improving efficiency and reducing energy consumption, which has guiding significance for tunnel construction.

[0024] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a framework diagram of the multi-objective optimization method for TBM operating parameters based on hybrid deep learning in the present invention.

[0026] Figure 2 This is a data schematic diagram of an embodiment of the present invention.

[0027] Figure 3 Schematic diagram of a convolutional neural network according to an embodiment of the present invention.

[0028] Figure 4 Schematic diagram of a long short-term memory network according to an embodiment of the present invention.

[0029] Figure 5a Schematic diagram of the prediction result of the tunneling speed corresponding to the target according to an embodiment of the present invention.

[0030] Figure 5b It is a schematic diagram of the target prediction result corresponding to the rock breaking energy ratio according to an embodiment of the present invention.

[0031] Figure 6aSchematic diagram of optimization results of tunneling speed corresponding to target according to an embodiment of the present invention.

[0032] Figure 6b It is a schematic diagram of the target optimization result corresponding to the rock breaking energy ratio according to an embodiment of the present invention.

[0033] Figure 6c Schematic diagram of the target optimization result corresponding to the cutter head speed according to an embodiment of the present invention.

[0034] Figure 6d Schematic diagram of total propulsion force corresponding to target optimization results according to an embodiment of the present invention. DETAILED DESCRIPTION

[0035] like Figures 1 to 6d As shown in FIG. 1 , a multi-objective optimization method for TBM operation parameters based on hybrid deep learning of the present invention is as follows: Figure 1 As shown, the specific steps are as follows:

[0036] Step 1: Obtain the mechanical parameters of the TBM in real-time excavation process, perform preprocessing to build a database, convert the data into graph data and divide the data set. The specific steps are as follows:

[0037] Step 1.1, the acquired TBM real-time recorded data is divided into each individual excavation ring, the parameters include the cutter head speed, total thrust, cutter head torque, main thrust cylinder pressure and auxiliary thrust cylinder pressure, and the corresponding rock breaking specific energy and excavation speed are calculated at the same time to form an original database;

[0038] Step 1.2: Use the isolation forest algorithm to identify outliers in the original database, then use linear interpolation to replace outliers, and use the Min-Max method to normalize the data. The calculation formula is as follows;

[0039]

[0040] Among them, x and x* are the feature values ​​before and after normalization, and min(x) and max(x) are the minimum and maximum values ​​in the feature, respectively.

[0041] Step 1.3: Construct the following according to the relationship between the data: Figure 2 The graph data shown. In this embodiment, the nodes of the graph are input features, namely, the cutter head speed, total thrust, cutter head torque, main thrust cylinder pressure and auxiliary thrust cylinder pressure, which completes the construction of the feature matrix. The edges are defined by constructing the unit matrix as the adjacency matrix;

[0042] Step 1.4: Use the first 80% of the data of each excavation ring as a training set to complete the construction of the multi-objective prediction model, and use the remaining 20% ​​of the data as a test set to verify the performance of the model.

[0043] Based on hybrid deep learning, with mechanical parameter information as input and current tunneling speed and rock breaking specific energy as output, a multi-objective prediction model considering spatiotemporal feature information is constructed;

[0044] Based on the established multi-objective prediction model considering the temporal and spatial characteristic information, the recommended values ​​of the cutter head speed and total thrust of the TBM are given with the goal of increasing the excavation speed and reducing the rock breaking specific energy.

[0045] Step 2: Based on hybrid deep learning, with mechanical parameter information as input and current tunneling speed and rock breaking specific energy as output, a multi-objective prediction model considering spatiotemporal feature information is constructed. The specific steps are as follows:

[0046] Step 2.1, create Figure 3 The multi-layer graph convolution network shown accepts the dimension of the input features and the dimension of the output features. In this embodiment, three layers of graph convolution layers are constructed, each layer includes a graph convolution layer, a batch normalization BN layer and a ReLU activation function; each layer of graph convolution layer multiplies the adjacency matrix with the input feature matrix, and then processes it through a linear transformation layer;

[0047] Step 2.2: Build Figure 4 The LSTM layer shown is used to process the time series features extracted by the GCN layer; in this embodiment, two LSTM layers are constructed, and the number of hidden units in each layer is 64. The node features output by the graph convolution layer are input into the LSTM layer, the output of the last time step of the LSTM layer is extracted, and finally the final prediction result is obtained by building a fully connected layer;

[0048] Step 2.3, the constructed graph data is used as input, including feature matrix and adjacency matrix, and the excavation speed and rock breaking specific energy are used as output; in this embodiment, BatchSize is set to 20, Adam optimization method is used, and learning rate is 0.001; after each epoch, the evaluation model is verified and the model with the lowest verification loss is saved;

[0049] Step 2.4: In this example, the coefficient of determination (R 2 ) and root mean square error (RMSE) to verify the performance of the model on the test set. The calculation formula is as follows: 2 The closer the value is to 1, the closer the RMSE value is to zero, which means that the model prediction is closer to the actual value of the data set, that is, the more accurate the model prediction is.

[0050]

[0051] Among them, y i , and represent the actual value, predicted value and average value respectively, and n represents the total number of samples.

[0052] Step 2.5. In addition, a scatter plot is drawn to visualize the performance of the multi-objective prediction model on the test set. Figure 5a , 5b As shown in Figure 2, the model is effective in predicting the tunneling speed. 2 and RMSE are 0.8796 and 3.8445 respectively. 2 and RMSE are 0.9122 and 1.0978 respectively, among which R 2 Both are much larger than 0.70, meeting the actual needs of the project.

[0053] Step 3: Based on the established multi-objective prediction model that considers spatiotemporal feature information, the recommended values ​​of the cutter head speed and total thrust of the TBM are given to increase the tunneling speed and reduce the rock breaking specific energy. Combined with the NSGA-II algorithm, 100 sets of data in the test set are selected as an example. According to the actual engineering requirements, the cutter head speed and total thrust are globally optimized. The specific steps are as follows:

[0054] Step 3.1, use the cutter head speed and total propulsion force as decision variables, the excavation speed and rock breaking specific energy as target variables, and determine the constraints according to the engineering requirements. Use the trained GCN-LSTM model to predict the values ​​of PR and SE as the output of the objective function. In this embodiment, the initial population size is defined as 100, the crossover rate is set to 0.9, the mutation rate is set to 0.2, and the maximum number of iterations is set to 50;

[0055] Step 3.2, use the non-dominated sorting selection operation based on crowding distance to select individuals that are not dominated by other individuals or have a small crowding degree, and use simulated binary crossover and polynomial mutation to perform crossover and mutation operations to generate new solutions;

[0056] Step 3.3, the final optimization results are shown in Figures 6a to 6d As shown in the figure, it can be seen that the average excavation speed increased from 30.60mm / min to 43.43mm / min, an increase of 41.93%; the average rock breaking specific energy decreased from 5.12MJ / m3 to 3.74MJ / m3, a decrease of 27.95%. In addition, the corresponding optimization recommendations for the cutter head speed and total thrust are given. This shows that this method has a good effect on TBM performance optimization, significantly reduces energy loss, and contributes to the efficient construction of TBM.

[0057] Step 4: During the actual TBM excavation, incremental learning is used to improve the performance of the optimization method. The specific steps are as follows:

[0058] Step 4.1. New data are continuously collected during the TBM excavation process and added to the database every two to five tunnel rings. Steps 1-3 are repeated using the new database to retrain the hybrid model, automatically improve the optimization framework, and realize automatic updating of the entire method.

[0059] The present invention also provides a TBM operating parameter multi-objective optimization system based on hybrid deep learning, comprising:

[0060] The data collection and processing module is used to obtain the mechanical parameter information recorded in real time during the TBM excavation process, pre-process the data to build a database, convert the data into graph data and divide the data set;

[0061] The multi-objective prediction module is used to use the mechanical parameter information as input and realize the simultaneous prediction of the current excavation speed and rock breaking efficiency based on the hybrid deep learning algorithm;

[0062] The multi-objective optimization module is based on a multi-objective prediction model. It performs global intelligent optimization of operating parameters according to the actual project requirements for excavation speed and rock-breaking specific energy, and automatically gives recommended values ​​for cutterhead speed and total thrust.

[0063] In another embodiment provided by the present invention, a device is also provided, comprising a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the multi-objective optimization method of TBM operating parameters based on hybrid deep learning described in the embodiment of the present invention.

[0064] In another embodiment provided by the present invention, a computer-readable storage medium is also provided, wherein at least one instruction, at least one program, a code set or an instruction set is stored in the storage medium, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by a processor to implement the multi-objective optimization method of TBM operating parameters based on hybrid deep learning described in an embodiment of the present invention.

[0065] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes a plurality of computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes multiple available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state hard disk Solid State Disk (SSD)), etc.

[0066] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0067] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0068] The above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims

1. A multi-objective optimization method for TBM operating parameters based on hybrid deep learning, It is characterized in that include: Obtain the mechanical parameter information recorded in real time during the TBM excavation process, pre-process the data to build a database, convert the data into graph data and divide the data set; Based on hybrid deep learning, with mechanical parameter information as input and current tunneling speed and rock breaking specific energy as output, a multi-objective prediction model considering spatiotemporal feature information is constructed; Based on the established multi-objective prediction model considering the temporal and spatial characteristic information, the recommended values ​​of the cutter head speed and total thrust of the TBM are given with the goal of increasing the excavation speed and reducing the rock breaking specific energy.

2. A multi-objective optimization method for TBM operating parameters based on hybrid deep learning according to claim 1, Features: The mechanical parameter information includes the cutter head rotation speed, the cutter head torque, the total propulsion force, the main propulsion cylinder pressure and the auxiliary propulsion cylinder pressure.

3. A multi-objective optimization method for TBM operating parameters based on hybrid deep learning according to claim 1, Features: The isolation forest algorithm is used to quickly and accurately process database outliers.

4. A multi-objective optimization method for TBM operating parameters based on hybrid deep learning according to claim 1, Features: The multi-target prediction model considering spatiotemporal feature information includes a GCN graph convolutional neural network layer, which converts tabular data into graph data to extract spatial information.

5. A multi-objective optimization method for TBM operating parameters based on hybrid deep learning according to claim 1, Features: The multi-objective prediction model considering spatiotemporal feature information includes an LSTM long short-term memory network layer, which extracts time series information and realizes simultaneous prediction of excavation speed and rock breaking specific energy.

6. A multi-objective optimization method for TBM operating parameters based on hybrid deep learning according to claim 1, Features: During the excavation process, newly recorded data can be dynamically added to the training data set to update the parameters in the multi-objective prediction model that takes into account the temporal and spatial characteristic information and enhance the model performance.

7. A multi-objective optimization method for TBM operating parameters based on hybrid deep learning according to claim 1, Features: The process of providing recommended values ​​of the cutter head rotation speed and total thrust of the TBM with the goal of increasing the advancement speed and reducing the rock-breaking energy ratio based on the established multi-objective prediction model considering the spatiotemporal characteristic information is as follows: using the NSGA-II fast elite multi-objective genetic algorithm, based on actual engineering needs, to achieve intelligent optimization and decision-making of the advancement speed and rock-breaking energy ratio, and provide corresponding recommended values ​​of the cutter head rotation speed and total thrust.

8. A multi-objective optimization system for TBM operating parameters based on hybrid deep learning, Features: include: The data collection and processing module is used to obtain the mechanical parameter information recorded in real time during the TBM excavation process, pre-process the data to build a database, convert the data into graph data and divide the data set; The multi-objective prediction module is used to use the mechanical parameter information as input and realize the simultaneous prediction of the current excavation speed and rock breaking efficiency based on the hybrid deep learning algorithm; The multi-objective optimization module is based on a multi-objective prediction model. It performs global intelligent optimization of operating parameters according to the actual project requirements for excavation speed and rock-breaking specific energy, and automatically gives recommended values ​​for cutterhead speed and total thrust.

9. An electronic device, It is characterized in that The electronic device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the multi-objective optimization method of TBM operating parameters based on hybrid deep learning as described in any one of claims 1 to 7.

10. A computer-readable storage medium, It is characterized in that The storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set are loaded and executed by the processor to implement the multi-objective optimization method of TBM operating parameters based on hybrid deep learning as described in any one of claims 1 to 7.

Citation Information

Cited By

  • TBM operation parameter intelligent decision-making method based on stratum information dynamic identification

    CN119554041A

  • TBM tunneling control parameter intelligent optimization decision-making method based on deep transfer learning

    CN120968649A

  • TBM tunneling control parameter intelligent optimization decision method based on deep transfer learning

    CN120968649B