Shield tunneling parameter real-time multi-objective optimization decision-making method based on multi-mode NSGA-II and LSTM

Abstract

The invention discloses a real-time multi-objective optimization decision-making method for shield tunneling parameters based on multi-mode NSGA-II (Non-dominated Sorting Genetic Algorithm-II) and LSTM (Long Short Term Memory). Comprising the following steps: acquiring a plurality of data sets, training a constructed shield tunneling parameter prediction model and a tool abrasion loss model, acquiring a tunneling parameter time sequence to be measured and a geological parameter time sequence, and inputting the tunneling parameter time sequence and the geological parameter time sequence into the shield tunneling parameter prediction model to obtain a tunneling parameter prediction value, inputting the tunneling parameter prediction value into a tool wear loss model to obtain a wear loss prediction value of each tool; constructing and solving a tunneling parameter multi-objective optimization model to obtain an optimal solution set of tunneling parameters; and performing minimum similar distance processing according to the optimal solution set, the tunneling parameter prediction value and the wear loss prediction value of each cutter, and taking the to-be-set tunneling parameter corresponding to the minimum similar distance value as the finally-set tunneling parameter at the current moment. According to the method, self-adaption to the geological condition of the tunneling stratum is achieved, the comprehensive tunneling performance of the shield is improved, and the total tunneling time and the construction cost in engineering are saved.

Description

Technical Field

[0001] The present invention belongs to the technical field of shield intelligent tunneling, and specifically relates to a real-time multi-objective optimization decision method for shield tunneling parameters based on multimodal NSGA-II and LSTM. Background Art

[0002] In current tunneling projects, decisions about shield tunneling parameters are still made by the shield driver, relying on their driving experience. When the ground changes or complex geological conditions are encountered, the shield driver cannot detect them in a timely manner, resulting in a relatively slow adjustment of the tunneling parameter decision-making strategy. In actual projects, tunneling parameter decision-making strategies vary widely, and experienced shield drivers are generally able to determine optimal tunneling parameter values ​​and adjust them based on geological conditions. However, shield drivers vary in their experience, making it difficult to determine optimal tunneling operating parameters.

[0003] To optimize tunneling parameters and improve the overall tunneling performance of shield machines, a comprehensive tunneling performance evaluation system must be established. Shield tunneling energy consumption and tool wear rate each represent individual aspects of shield tunneling performance. The decision-making process for shield tunneling parameters in actual tunneling projects is a multi-objective optimization problem. In conventional geology, the goal is typically to maintain a low tool wear rate while maintaining low tunneling energy consumption. In challenging geology, safe tunneling is generally pursued, requiring lower tunneling energy consumption to minimize ground disturbance and ensure smooth tunneling. Currently, a widely accepted quantitative evaluation method for comprehensive tunneling performance has not yet been established. Therefore, it is necessary to establish a comprehensive shield tunneling performance evaluation index system to enable multi-objective optimization of tunneling parameters.

[0004] Currently, numerous researchers at home and abroad have conducted in-depth research on intelligent decision-making for shield tunneling parameters, including prediction-based and optimization-based decision-making. Predictive-based decision-making primarily leverages statistical or machine learning methods based on massive amounts of historical construction data to learn from the extensive experience of shield tunneling operators, specifically determining which tunneling parameter settings should be adopted under different geological conditions. Other researchers have used machine learning or deep learning methods to predict shield tunneling performance indicators. This research primarily aims to predict tunneling project costs and durations before construction commences, thereby enabling construction management. It is not suitable for real-time decision-making on tunneling parameters. The existing patent CN120124213A employs a supervised learning model to predict the objective function and employs subjective manual weighting to select the final parameter combination. This method still relies on manual weighting, and the resulting manually set weights make it difficult to adapt to complex and changing geological conditions. Summary of the Invention

[0005] In order to solve the problems existing in the background technology, the present invention provides a real-time multi-objective optimization decision-making method for shield tunneling parameters based on multimodal NSGA-II and LSTM, which solves the technical problems in the prior art such as the inevitable misoperation of shield drivers relying on experience to set shield operation parameter values, the limitation of manual subjective empowerment that is difficult to adapt to continuously changing geological conditions, the lack of timely and effective adjustment, and the lack of fine geological adaptability shield tunneling parameters, which lead to unsatisfactory construction results and low efficiency.

[0006] The technical solutions adopted in the present invention include: 1. A real-time multi-objective optimization decision-making method for shield tunneling parameters based on multimodal NSGA-II and LSTM: S1. Construct a shield tunneling dataset based on the continuous acquisition of tunneling parameter time series and geological parameter time series; obtain a tunneling parameter dataset by summarizing the tunneling parameters collected at different times; and construct a tool wear dataset based on the tunneling parameters at different times and the wear of each tool corresponding to each tunneling parameter.

[0007] S2. A shield tunneling parameter prediction model is constructed based on the training of the shield tunneling data set, and a tool wear model is constructed based on the training of the tool wear data set; the time series of the tunneling parameters to be measured and the time series of the geological parameters to be measured before the current moment are collected and combined and input into the trained shield tunneling parameter prediction model to obtain the tunneling parameter prediction value, and the tunneling parameter prediction value is input into the trained tool wear model to obtain the wear prediction value of each tool.

[0008] S3. Determine the objective function and constraints based on the current geological environment, the excavation parameters to be set, the wear of each tool, and the excavation parameter data set, and then construct a multi-objective optimization model for the excavation parameters. Use a multimodal non-dominated sorting genetic algorithm to solve the multi-objective optimization model for the excavation parameters and obtain the optimal solution set for the excavation parameters to be set under the current geological environment.

[0009] S4. Perform minimum similarity distance processing based on the obtained optimal solution set, the predicted values ​​of the excavation parameters and the predicted values ​​of the wear of each tool to obtain the minimum similarity distance value, and use the excavation parameter to be set corresponding to the minimum similarity distance value as the excavation parameter finally set at the current moment.

[0010] Furthermore, after obtaining the final excavation parameters set at the current moment, the shield machine is set to keep the excavation parameters constant for a preset period of time. After the preset period of time, the new final excavation parameters are obtained again, and then the shield machine is set again for a preset period of time according to the new final excavation parameters, thereby realizing continuous adjustment of the excavation parameters of the shield machine until the project is completed.

[0011] The step S1 is specifically as follows: S11. Collect continuous excavation parameters and geological parameters within a number of historical preset time periods.

[0012] S12. Combine the continuous tunneling parameter time series and geological parameter time series in the first half of the preset time period as input data, use the mean of the continuous tunneling parameter time series in the second half of the preset time period as the predicted label value, and combine the input data and the predicted label value into shield tunneling data.

[0013] S13. All shield tunneling data under the same geological type environment are aggregated to obtain a single shield tunneling dataset. Single shield tunneling datasets under all geological types are aggregated to obtain a shield tunneling dataset.

[0014] S14. Collect the excavation parameters at several different moments in history and the wear of each tool measured after each excavation parameter is set.

[0015] S15. Under the same geological type environment, all tunneling parameters are aggregated to obtain a single tunneling parameter data set; and the single tunneling parameter data sets under all geological types are aggregated to obtain a tunneling parameter data set.

[0016] S16. Using the excavation parameters as input data and the wear amounts of each tool corresponding to the excavation parameters as labels, the input data and the labels are combined into tool wear amount data at a single moment.

[0017] S17. The tool wear data at all times in the same geological environment are aggregated to obtain a single tool wear data set. The single tool wear data sets in all geological environments are aggregated to obtain a tool wear data set.

[0018] The step S2 is specifically as follows: S21. Construct a shield tunneling parameter prediction model, input the shield tunneling data set into the shield tunneling parameter prediction model for training, and obtain a trained shield tunneling parameter prediction model.

[0019] S22. Construct a tool wear model, input the tool wear data set into the tool wear model for training, and obtain a trained tool wear model.

[0020] S23. Under the current geological environment, a continuous time series of the tunneling parameters to be measured and a time series of the geological parameters to be measured within a preset time period before the current moment are collected and inputted into the trained shield tunneling parameter prediction model for processing to obtain a predicted value of the tunneling parameters.

[0021] S24. Input the predicted values ​​of the tunneling parameters into the trained tool wear model to obtain the predicted values ​​of the wear of each tool.

[0022] The shield tunneling parameter prediction model adopts a long short-term memory neural network.

[0023] The tool wear model adopts a multi-layer fully connected network.

[0024] The tunneling parameter time series includes the cutterhead rotation speed, propulsion speed, cutterhead torque and shield machine total propulsion force time series; the geological parameter time series includes the internal friction angle, cohesion, permeability coefficient, bearing capacity characteristic value, natural density, deformation modulus, tunnel burial depth and water level height time series.

[0025] The tunneling parameter prediction values ​​include a cutterhead rotation speed prediction value, a propulsion speed prediction value, a cutterhead torque prediction value and a shield machine total propulsion force prediction value.

[0026] The objective function and constraints of the multi-objective optimization model for tunneling parameters are set according to the following formula: Ming(x t )=[g1(x t ),g2(x t )] T g1(x t )=∑ i=1 N ((R i ×A i (x t )) / (ΔM×∑ i=1 N R i )), g2(x t )=4((n t ×T t ) / L(v t )+F t ) / (πD 2 ) L(v t )=v t , v t ≥10 -5 ;L(v t )=10 -5 , v t <10 -5 x t =[n t ,v t ,T t ,F t ] T , x t ∈X,X~R setC 4 Among them, Min means taking the minimum; t means the current time; g(x t ) represents the objective function of the multi-objective optimization model for tunneling parameters; g1(x t ) represents the comprehensive wear rate of the tool, g2(x t ) represents the energy consumption per unit volume of soil excavation; [ ] T represents transpose; x t Indicates the excavation parameters to be set; i indicates the index; R i Indicates the installation radius of the i-th tool; A i (x t ) represents the wear of the i-th tool corresponding to the excavation parameter to be set; ΔM represents the unit excavation mileage; N represents the total number of tools; n t Indicates the cutter head speed to be set; v t Indicates the propulsion speed to be set; T t Indicates the cutter head torque to be set; F t Indicates the total thrust of the shield machine to be set; D indicates the cutter head diameter of the shield machine; L(v t ) represents the propulsion speed v t Piecewise function; C is the index, indicating the Cth type of geological environment; R setC 4 represents a single excavation parameter dataset under the C-type geological environment in the excavation parameter dataset; X represents a dataset formed by selecting the excavation parameters of the excavation section from the single excavation parameter dataset; ~ represents included in; X~R setC 4 Indicates that the dataset X is contained in a single excavation parameter dataset R setC 4 .

[0027] The minimum similarity distance processing based on the obtained optimal solution set, the predicted values ​​of the excavation parameters and the predicted values ​​of the wear of each tool is specifically as follows: the comprehensive wear rate of the tool obtained according to the predicted values ​​of the wear of each tool, the energy consumption of excavating a unit volume of soil obtained according to the predicted values ​​of the excavation parameters, the comprehensive wear rate of the tool corresponding to each excavation parameter in the optimal solution set, and the energy consumption of excavating a unit volume of soil are subjected to minimum similarity distance processing.

[0028] The minimum similarity distance processing is set according to the following formula: Min ((g1(x tp )-g1(x t )) 2 +(g2(x tp )-g2(x t )) 2 ) 0.5 g1(x tp )=∑i=1 N ((R i ×A ip (x tp )) / (ΔM×∑ i=1 N R i )) g1(x t )=∑ i=1 N ((R i ×A i (x t )) / (ΔM×∑ i=1 N R i )) g2(x tp )=4((n tp ×T tp ) / L(v tp )+F tp ) / (πD 2 ),g2(x t )=4((n t ×T t ) / L(v t )+F t ) / (πD 2 ) L(v tp )=v tp ,v tp ≥10 -5 ;L(v tp )=10 -5 ,v tp <10 -5 L(v t )=v t ,v t ≥10 -5 ;L(v t )=10 -5 ,v t <10 -5 x tp =[n tp ,v tp ,T tp ,F tp ] T ,x t =[n t ,v t ,T t ,F t ] T ,x t ∈R paretoC4 Among them, x tp represents the predicted value of tunneling parameters; g1(x tp ) represents the comprehensive wear rate of the tool obtained based on the wear prediction value of each tool; g2(x tp ) represents the energy consumption per unit volume of soil excavated based on the predicted values ​​of excavation parameters; A ip (x tp ) represents the predicted value of wear of the i-th tool corresponding to the predicted value of the excavation parameter; n tp Indicates the predicted value of cutter head speed; v tp represents the predicted value of propulsion speed; T tp Indicates the predicted value of cutter head torque; F tp represents the predicted value of the total thrust of the shield machine; L(v tp ) represents the predicted value of propulsion speed v tp Piecewise function of R paretoC 4 represents the optimal solution set of the excavation parameters to be set under the current geological environment of type C; Min represents the minimum; t represents the current time; g1(x t ) represents the comprehensive wear rate of the tool, g2(x t ) represents the energy consumption per unit volume of soil excavation; [ ] T represents transpose; x t Indicates the excavation parameters to be set; i indicates the index; R i Indicates the installation radius of the i-th tool; A i (x t ) represents the wear of the i-th tool corresponding to the excavation parameter to be set; ΔM represents the unit excavation mileage; N represents the total number of tools; n t Indicates the cutter head speed to be set; v t Indicates the propulsion speed to be set; T t Indicates the cutter head torque to be set; F t Indicates the total thrust of the shield machine to be set; D indicates the cutter head diameter of the shield machine; L(v t ) represents the propulsion speed v t A piecewise function; C is the index, indicating the current C-type geological environment.

[0029] 2. A computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0030] The beneficial effects of the present invention are: The present invention adopts a multimodal NSGA-II optimization method and LSTM time series prediction technology to establish a real-time multi-objective optimization decision-making method for shield tunneling parameters based on multimodal NSGA-II and LSTM, and conducts engineering verification on an earth pressure shield in a certain city. The verification results show that the method proposed in the present invention can adapt to the geological conditions of the excavation stratum and significantly improve the comprehensive excavation performance of the shield, saving the total excavation time and construction cost of the project. It further solves the technical problems in the prior art such as the inevitable misoperation of the shield driver who relies on experience to set the shield operation parameter values, the limitation of manual subjective empowerment that is difficult to adapt to continuously changing geological conditions, the lack of timely and effective adjustment, and the lack of precise geological adaptability shield tunneling parameters, which lead to unsatisfactory construction results and low efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 Flowchart of the method of the present invention.

[0032] Figure 2 When the stratum is clayey silt or fine sand composite, a comparison chart of the verification results obtained by the method of the present invention is provided in the project.

[0033] Figure 3 When the stratum is fine sand, a comparison chart of the verification results obtained by using the method of the present invention in the project is shown. DETAILED DESCRIPTION

[0034] The present invention is described in more detail below with reference to the accompanying drawings and examples. However, the present invention is not limited thereto. A person skilled in the art may make various improvements and modifications without departing from the principles of the present invention, and such improvements and modifications are considered to be within the scope of protection of the present invention. Any matters not described in detail in this specification constitute prior art known to those skilled in the art.

[0035] like Figure 1 As shown, the real-time multi-objective optimization decision-making method for shield tunneling parameters of this embodiment includes the following steps: S1. Construct a shield tunneling dataset based on the continuous acquisition of tunneling parameter time series and geological parameter time series; obtain a tunneling parameter dataset by summarizing the tunneling parameters collected at different times; and construct a tool wear dataset based on the tunneling parameters at different times and the wear of each tool corresponding to each tunneling parameter.

[0036] S11. Collect continuous excavation parameters and geological parameters within a number of historical preset time periods.

[0037] S12. Combine the continuous tunneling parameter time series and geological parameter time series in the first half of the preset time period as input data, and use the temporal mean of the continuous tunneling parameter time series in the second half of the preset time period as the predicted label value. The input data and the predicted label value are combined into shield tunneling data.

[0038] The time series of tunneling parameters include the time series of cutterhead speed, propulsion speed, cutterhead torque and total propulsion force of the shield machine; the time series of geological parameters include the time series of internal friction angle, cohesion, permeability coefficient, bearing capacity characteristic value, natural density, deformation modulus, tunnel burial depth and water level height.

[0039] S13. All shield tunneling data under the same geological type environment are aggregated to obtain a single shield tunneling dataset. Single shield tunneling datasets under all geological types are aggregated to obtain a shield tunneling dataset.

[0040] The geological environment includes clayey silt, fine sand composite strata and fine sand strata.

[0041] S14, collecting the excavation parameters at several different moments in history and the wear of each tool measured by the sensor after each excavation parameter is set.

[0042] S15. Under the same geological environment, all tunneling parameters are aggregated to obtain a single tunneling parameter data set. The single tunneling parameter data sets under all geological environments are aggregated to obtain a tunneling parameter data set.

[0043] S16. Using the excavation parameters as input data and the wear amounts of each tool corresponding to the excavation parameters as labels, the input data and the labels are combined into tool wear amount data at a single moment.

[0044] S17. The tool wear data at all times in the same geological environment are aggregated to obtain a single tool wear data set. The single tool wear data sets in all geological environments are aggregated to obtain a tool wear data set.

[0045] S2. A shield tunneling parameter prediction model is constructed based on the training of the shield tunneling data set, and a tool wear model is constructed based on the training of the tool wear data set; the time series of the tunneling parameters to be measured and the time series of the geological parameters to be measured before the current moment are collected and combined and input into the trained shield tunneling parameter prediction model to obtain a constant tunneling parameter prediction value in the next time period, and the tunneling parameter prediction value is input into the trained tool wear model to obtain the wear prediction value of each tool.

[0046] S21. Construct a shield tunneling parameter prediction model, input the shield tunneling data set into the shield tunneling parameter prediction model for training, and obtain a trained shield tunneling parameter prediction model.

[0047] The shield tunneling parameter prediction model uses a long short-term memory (LSTM) neural network. Specifically, the LSTM layer has 32 units, uses the "glorot_uniform" initialization method, and optimizes the LSTM model using the Adam optimizer.

[0048] S22. Construct a tool wear model, input the tool wear data set into the tool wear model for training, and obtain a trained tool wear model.

[0049] The tool wear model uses a multi-layer fully connected network. In specific implementation, a three-layer fully connected network or a five-layer fully connected network can be used.

[0050] S23. Under the current geological environment, a continuous time series of the tunneling parameters to be measured and a time series of the geological parameters to be measured within a preset time period before the current moment are collected and inputted into the trained shield tunneling parameter prediction model for processing to obtain a predicted value of the tunneling parameters.

[0051] The predicted values ​​of tunneling parameters include the predicted value of cutterhead speed, the predicted value of propulsion speed, the predicted value of cutterhead torque and the predicted value of total propulsion force of the shield machine.

[0052] S24. Input the predicted values ​​of the tunneling parameters into the trained tool wear model to obtain the predicted values ​​of the wear of each tool.

[0053] S3. Determine the objective function and constraints based on the current geological environment, the tunneling parameters to be set at the current moment, the wear of each tool and the tunneling parameter data set at the current moment, and the set parameters of the shield machine, and then construct a multi-objective optimization model for the tunneling parameters. Use the multi-modal non-dominated sorting genetic algorithm (NSGA-II algorithm) to solve the multi-objective optimization model for the tunneling parameters and obtain the optimal solution set for the tunneling parameters to be set under the current Class C geological environment.

[0054] The objective function and constraints of the multi-objective optimization model for tunneling parameters are set according to the following formula: Ming(x t )=[g1(x t ),g2(x t )] T g1(x t )=∑ i=1 N ((R i ×A i (x t )) / (ΔM×∑ i=1 N R i )), g2(xt )=4((n t ×T t ) / L(v t )+F t ) / (πD 2 ) L(v t )=v t , v t ≥10 -5 ;L(v t )=10 -5 , v t <10 -5 x t =[n t ,v t ,T t ,F t ] T , x t ∈X,X~R setC 4 Among them, Min means taking the minimum; t means the current time; g(x t ) represents the objective function of the multi-objective optimization model for tunneling parameters; g1(x t ) represents the comprehensive wear rate of the tool at the current moment, g2(x t ) represents the energy consumption per unit volume of soil excavated at the current moment (excavation specific energy consumption); [ ] T represents transpose; x t Indicates the excavation parameters to be set at the current time t; i represents the index; R i A represents the installation radius of the i-th tool, which has been determined in advance; i (x t ) represents the wear of the i-th tool corresponding to the excavation parameters to be set at the current time t, and this value is measured by the tool wear sensor; ΔM represents the unit excavation mileage, and this value can be measured in advance by the sensor; N represents the total number of tools, and this value is determined in advance; n t Indicates the cutter head speed to be set at the current time t; v t Indicates the propulsion speed to be set at the current time t; T t Indicates the cutter head torque to be set at the current time t; F t represents the total thrust of the shield machine to be set at the current time t; D represents the cutter head diameter of the shield machine, which has been determined in advance; L(v t ) represents the propulsion speed v t Dimensionless piecewise function; C is the index, indicating the current C-type geological environment; R setC 4represents a single excavation parameter dataset under the C-type geological environment in the excavation parameter dataset; X represents a dataset formed by selecting the excavation parameters of the excavation section from the single excavation parameter dataset; ~ represents included in; X~R setC 4 Indicates that the dataset X is contained in a single excavation parameter dataset R setC 4 .

[0055] In the specific implementation, a single tunneling parameter data set R setC 4 The data set R includes the unstable excavation parameters such as the rising section, the falling section and the abnormal section, as well as the excavation parameters of the relatively stable excavation section. The method of the present invention mainly sets the excavation parameters of the excavation section, so the single excavation parameter data set R setC 4 The excavation parameters of the excavation section in the excavation section are of reference significance, and it is necessary to obtain the excavation parameters from the single excavation parameter data set R setC 4 The excavation parameters of the excavation section are selected to obtain the data set X.

[0056] In the specific implementation, the objective function of the multi-objective optimization model of tunneling parameters is an index function that integrates the tunneling performance, and its purpose is to minimize the comprehensive wear rate of the tool g1(x t ) and the energy consumption per unit volume of soil excavation g2(x t ) (Specific energy consumption during excavation).

[0057] In this embodiment, when solving the multi-objective optimization model of tunneling parameters using the multi-modal non-dominated sorting genetic algorithm (NSGA-II algorithm): The sample individuals are: data points (excavation parameters) in a data set X formed by selecting the excavation parameters of an excavation section from a single excavation parameter data set under the current geological type C environment.

[0058] The population is: a data set consisting of 100 sample individuals.

[0059] The fitness function is: g1(x t ) and g2(x t ).

[0060] The crossover and mutation processes are as follows: Crossover involves exchanging the weights of the corresponding tunneling parameters between two sample individuals; mutation involves swapping the different tunneling parameters within a sample. Both crossover and mutation aim to make it easier for the optimization algorithm to escape local optima within the multi-objective function space.

[0061] S4. Perform minimum similarity distance processing based on the obtained optimal solution set, the predicted values ​​of the excavation parameters and the predicted values ​​of the wear of each tool to obtain the minimum similarity distance value, and use the excavation parameter to be set corresponding to the minimum similarity distance value as the excavation parameter finally set at the current moment.

[0062] The minimum similarity distance processing is performed based on the obtained optimal solution set, the predicted values ​​of the excavation parameters and the predicted values ​​of the wear of each tool. Specifically, the minimum similarity distance processing is performed on the comprehensive wear rate of the tool obtained according to the predicted values ​​of the wear of each tool, the energy consumption of excavating a unit volume of soil obtained according to the predicted values ​​of the excavation parameters, the comprehensive wear rate of the tool corresponding to each excavation parameter in the optimal solution set, and the energy consumption of excavating a unit volume of soil.

[0063] The minimum similarity distance processing is set according to the following formula: Min ((g1(x tp )-g1(x t )) 2 +(g2(x tp )-g2(x t )) 2 ) 0.5 g1(x tp )=∑ i=1 N ((R i ×A ip (x tp )) / (ΔM×∑ i=1 N R i )), g1(x t )=∑ i=1 N ((R i ×A i (x t )) / (ΔM×∑ i=1 N R i )) g2(x tp )=4((n tp ×T tp ) / L(v tp )+F tp ) / (πD 2 ), g2(x t )=4((n t ×T t ) / L(v t )+F t ) / (πD 2 ) L(v tp)=v tp , v tp ≥10 -5 ;L(v tp )=10 -5 , v tp <10 -5 L(v t )=v t , v t ≥10 -5 ;L(v t )=10 -5 , v t <10 -5 x tp =[n tp ,v tp ,T tp ,F tp ] T , x t =[n t ,v t ,T t ,F t ] T , x t ∈R paretoC 4 Among them, x tp represents the predicted value of the tunneling parameters at the current time t; g1(x tp ) represents the comprehensive wear rate of the tool obtained based on the wear prediction value of each tool at the current time t; g2(x tp ) represents the energy consumption per unit volume of soil excavated according to the predicted value of the excavation parameters at the current time t; A ip (x tp ) represents the predicted value of wear of the i-th tool corresponding to the predicted value of the tunneling parameter at the current time t; n tp represents the predicted value of the cutter head speed at the current time t; v tp represents the predicted value of the propulsion speed at the current time t; T tp F represents the predicted value of the cutter head torque at the current time t; tp represents the predicted value of the total thrust of the shield machine at the current time t; L(v tp ) represents the predicted value of propulsion speed v tp The dimensionless piecewise function of R paretoC 4 represents the optimal solution set of tunneling parameters to be set under the C-type geological environment. Min represents the minimum; t represents the current time; g1(x t ) represents the comprehensive wear rate of the tool at the current moment, g2(x t) represents the energy consumption per unit volume of soil excavated at the current moment (excavation specific energy consumption); [ ] T represents transpose; x t Indicates the excavation parameters to be set at the current time t; i represents the index; R i Indicates the installation radius of the i-th tool; A i (x t ) represents the wear of the i-th tool corresponding to the excavation parameter to be set at the current time t, which is measured by the tool wear sensor; ΔM represents the unit excavation mileage, which can be measured in advance by the sensor; N represents the total number of tools; n t Indicates the cutter head speed to be set at the current time t; v t Indicates the propulsion speed to be set at the current time t; T t Indicates the cutter head torque to be set at the current time t; F t represents the total thrust of the shield machine to be set at the current time t; D represents the cutter head diameter of the shield machine; L(v t ) represents the propulsion speed v t A dimensionless piecewise function of ; C is the index, indicating the current C-type geological environment.

[0064] Furthermore, after obtaining the final excavation parameters set at the current moment, the shield machine is set to keep the excavation parameters constant for a preset period of time. After the preset period of time, the new final excavation parameters are obtained again, and then the shield machine is set again for a preset period of time according to the new final excavation parameters, thereby realizing continuous adjustment of the excavation parameters of the shield machine until the project is completed.

[0065] The reason for not obtaining the tunneling parameters at every moment and setting the shield machine to constant tunneling parameters for a period of time is to avoid unnecessary fluctuations of the shield machine due to adjusting the parameters at every moment, and the shield machine cannot withstand the adjustment of real-time parameters.

[0066] In this embodiment, the shield machine sets the tunneling parameters constantly for 60 seconds, and then re-acquires the tunneling parameters after 60 seconds and continues to set them for 60 seconds, and repeats this process until the project is completed.

[0067] This example verifies the effectiveness and feasibility of the present invention by comparing the set excavation parameters obtained by the method of the present invention with the excavation parameters set by the shield driver himself in actual excavation, in terms of the average tool comprehensive wear rate and average excavation specific energy consumption in the excavation section of the real site.

[0068] The method of the present invention was applied to the No. 998 earth pressure shield in a certain section of a certain city's rail transit line 6 from a certain station to a certain station. The engineering section mainly consists of clay silt, fine sand composite strata and fine sand strata. Figure 2As shown in the figure, in a clayey silt and fine sand composite stratum, the average excavation specific energy consumption of the 43 shield tunneling sections that the shield tunneling drivers selected the method of the present invention to set (the shield tunneling drivers selected the optimized decision value) was reduced by 2.34% and the average comprehensive tool wear rate was reduced by 8.49% compared with the remaining 40 excavation sections that the shield tunneling drivers did not select the method of the present invention to set (the shield tunneling drivers did not select the optimized decision value). Figure 3 As shown in Table 1, in the fine sand formation, the average excavation specific energy consumption of the 37 excavation sections in which the shield drivers chose the method of the present invention for setting (the shield drivers chose the optimized decision value) was reduced by 4.57%, and the average comprehensive tool wear rate was reduced by 13.00% compared with the remaining 41 excavation sections in which the shield drivers did not choose the method of the present invention for setting (the shield drivers did not choose the optimized decision value), as shown in Table 1.

[0069] Table 1 Improvement in tunneling performance in different strata: Geological type environment Average excavation specific energy consumption is reduced Average tool wear rate is reduced Clayey silt and fine sand composite stratum 2.34% 8.49% Fine sand formation 4.57% 13.00% The method of the present invention realizes self-adaptation to the geological conditions of the excavation stratum and significantly improves the comprehensive excavation performance of the shield, further saving the total excavation time and construction cost of the project.

[0070] It should be understood that the present embodiment is only used to illustrate the present invention and is not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms also fall within the scope limited by the appended claims of the application.

Claims

1. A real-time multi-objective optimization decision-making method for shield tunneling parameters based on multimodal NSGA-II and LSTM, characterized by: The following steps are involved: S1. Construct a shield tunneling dataset based on the continuous acquisition of tunneling parameter time series and geological parameter time series; obtain a tunneling parameter dataset by summarizing the tunneling parameters collected at different times; and construct a tool wear dataset based on the tunneling parameters at different times and the wear of each tool corresponding to each tunneling parameter; S2. A shield tunneling parameter prediction model and a tool wear model are trained and constructed based on the shield tunneling data set and the tool wear data set, respectively; a time series of the tunneling parameters to be measured and a time series of the geological parameters to be measured are collected and inputted into the trained shield tunneling parameter prediction model to obtain tunneling parameter prediction values, and the tunneling parameter prediction values ​​are inputted into the trained tool wear model to obtain wear prediction values ​​for each tool; S3. Determine the objective function and constraints based on the current geological environment, the excavation parameters to be set, the wear of each tool, and the excavation parameter data set, and then construct a multi-objective optimization model for the excavation parameters. Use a multimodal non-dominated sorting genetic algorithm to solve the multi-objective optimization model for the excavation parameters to obtain the optimal solution set for the excavation parameters to be set. S4. Perform minimum similarity distance processing based on the obtained optimal solution set, the predicted values ​​of the excavation parameters and the predicted values ​​of the wear of each tool to obtain the minimum similarity distance value, and use the excavation parameter to be set corresponding to the minimum similarity distance value as the excavation parameter finally set at the current moment.

2. The real-time multi-objective optimization decision-making method for shield tunneling parameters based on multimodal NSGA-II and LSTM according to claim 1 is characterized in that: The step S1 is specifically as follows: S11, collecting continuous excavation parameters and geological parameters within a number of historical preset time periods; S12. Combining the continuous tunneling parameter time series and geological parameter time series within the first half of the preset time period as input data, taking the mean of the continuous tunneling parameter time series within the second half of the preset time period as a predicted label value, and combining the input data and the predicted label value into shield tunneling data; S13. All shield tunneling data under the same geological environment are aggregated to obtain a single shield tunneling dataset. Single shield tunneling datasets under all geological environments are aggregated to obtain a shield tunneling dataset. S14, collecting the excavation parameters at several different moments in history and the wear of each tool measured after each excavation parameter is set; S15. Under the same geological type environment, all excavation parameters are aggregated to obtain a single excavation parameter data set; the single excavation parameter data sets under all geological type environments are aggregated to obtain an excavation parameter data set; S16. Using the excavation parameters as input data and the wear amounts of each tool corresponding to the excavation parameters as labels, the input data and the labels are combined to form tool wear data at a single moment; S17. The tool wear data at all times in the same geological environment are aggregated to obtain a single tool wear data set. The single tool wear data sets in all geological environments are aggregated to obtain a tool wear data set.

3. The real-time multi-objective optimization decision-making method for shield tunneling parameters based on multimodal NSGA-II and LSTM according to claim 1 is characterized in that: The step S2 is specifically as follows: S21, constructing a shield tunneling parameter prediction model, inputting the shield tunneling data set into the shield tunneling parameter prediction model for training, and obtaining a trained shield tunneling parameter prediction model; S22, constructing a tool wear model, inputting the tool wear data set into the tool wear model for training, and obtaining a trained tool wear model; S23, under the current geological environment, collecting a continuous time series of the tunneling parameter to be measured and a time series of the geological parameter to be measured within a preset time period before the current moment, combining them and inputting them into the trained shield tunneling parameter prediction model for processing, and predicting the tunneling parameter prediction value; S24. Input the predicted values ​​of the tunneling parameters into the trained tool wear model to obtain the predicted values ​​of the wear of each tool.

4. The method for real-time multi-objective optimization decision-making of shield tunneling parameters based on multimodal NSGA-II and LSTM according to claim 1 is characterized in that: The shield tunneling parameter prediction model adopts a long short-term memory neural network.

5. The real-time multi-objective optimization decision-making method for shield tunneling parameters based on multimodal NSGA-II and LSTM according to claim 1 is characterized by: The tool wear model adopts a multi-layer fully connected network.

6. The real-time multi-objective optimization decision-making method for shield tunneling parameters based on multimodal NSGA-II and LSTM according to claim 1 is characterized by: The tunneling parameter time series includes the cutterhead speed, propulsion speed, cutterhead torque and shield machine total propulsion force time series; the geological parameter time series includes the internal friction angle, cohesion, permeability coefficient, bearing capacity characteristic value, natural density, deformation modulus, tunnel burial depth and water level height time series; the tunneling parameter prediction value includes the cutterhead speed prediction value, propulsion speed prediction value, cutterhead torque prediction value and shield machine total propulsion force prediction value.

7. The real-time multi-objective optimization decision-making method for shield tunneling parameters based on multimodal NSGA-II and LSTM according to claim 6 is characterized in that: The objective function and constraints of the multi-objective optimization model for tunneling parameters are set according to the following formula: Min g(x t )=[g1(x t ),g2(x t )] T g1(x t )=∑ i=1 N ((R i ×A i (x t )) / (ΔM×∑ i=1 N R i )),g2(x t )=4((n t ×T t ) / L(v t )+F t ) / (πD 2 ) L(v t )=v t ,v t ≥10 -5 ;L(v t )=10 -5 ,v t <10 -5 x t =[n t ,v t ,T t ,F t ] T ,x t ∈X,X~R setC 4 Among them, g(x t ) represents the objective function of the multi-objective optimization model for tunneling parameters; g1(x t ) represents the comprehensive wear rate of the tool, g2(x t ) represents the energy consumption of excavating unit volume of soil; x t Indicates the excavation parameters to be set; i indicates the index; R i Indicates the installation radius of the i-th tool; A i (x t ) represents the wear of the i-th tool corresponding to the excavation parameter to be set; ΔM represents the unit excavation mileage; N represents the total number of tools; n t Indicates the cutter head speed to be set; v t Indicates the propulsion speed to be set; T t Indicates the cutter head torque to be set; F t Indicates the total thrust of the shield machine to be set; D indicates the cutter head diameter of the shield machine; L(v t ) represents the propulsion speed v t Piecewise function of; C is the index; R setC 4 represents a single excavation parameter dataset under the C-type geological environment in the excavation parameter dataset; X represents a dataset formed by selecting the excavation parameters of the excavation section from the single excavation parameter dataset; ~ represents included in.

8. The real-time multi-objective optimization decision-making method for shield tunneling parameters based on multimodal NSGA-II and LSTM according to claim 7 is characterized in that: The minimum similarity distance processing based on the obtained optimal solution set, the tunneling parameter prediction value and the wear prediction value of each tool is specifically as follows: The comprehensive tool wear rate obtained from the predicted values ​​of wear of each tool, the energy consumption per unit volume of soil excavated from the predicted values ​​of tunneling parameters, and the comprehensive tool wear rate and energy consumption per unit volume of soil excavated corresponding to each tunneling parameter in the optimal solution set are processed with the minimum similarity distance.

9. The real-time multi-objective optimization decision-making method for shield tunneling parameters based on multimodal NSGA-II and LSTM according to claim 8 is characterized in that: The minimum similarity distance processing is set according to the following formula: Min ((g1(x tp )-g1(x t )) 2 +(g2(x tp )-g2(x t )) 2 ) 0.5 g1(x tp )=∑ i=1 N ((R i ×A ip (x tp )) / (ΔM×∑ i=1 N R i )),g2(x t )=4((n t ×T t ) / L(v t )+F t ) / (πD 2 ) L(v tp )=v tp ,v tp ≥10 -5 ;L(v tp )=10 -5 ,v tpt <10 -5 x tp =[n tp ,v tp ,T tp ,F tp ] T ,x t ∈R paretoC 4 Among them, x tp represents the predicted value of tunneling parameters; g1(x tp ) represents the comprehensive wear rate of the tool obtained based on the wear prediction value of each tool; g2(x tp ) represents the energy consumption per unit volume of soil excavated based on the predicted values ​​of excavation parameters; A ip (x tp ) represents the predicted value of wear of the i-th tool corresponding to the predicted value of the excavation parameter; n tp Indicates the predicted value of cutter head speed; v tp represents the predicted value of propulsion speed; T tp Indicates the predicted value of cutter head torque; F tp represents the predicted value of the total thrust of the shield machine; L(v tp ) represents the predicted value of propulsion speed v tp Piecewise function of R paretoC 4 Represents the optimal solution set of tunneling parameters to be set under the C-type geological environment.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 9 are implemented.

Images