Method and equipment for optimizing deformation of shield under existing tunnel based on intelligent algorithm

Abstract

The invention belongs to the technical field of shield construction control, and particularly discloses an intelligent algorithm-based deformation optimization method and equipment for shield underneath passing of an existing tunnel. Comprising the steps of establishing an index system and constructing a sample data set; building a prediction model of the transverse and vertical deformation of the top of the tunnel lining and the transverse and vertical deformation rate of the top of the tunnel lining based on BO-CNN-three layer LSTM; and taking a nonlinear mapping relation function obtained by the prediction model as a target optimization fitness function, constructing a multi-target optimization model, and proposing a construction parameter setting suggestion range under the condition of minimum deformation. According to the method, multi-objective optimization of transverse and vertical deformation of the top of the tunnel lining, transverse and vertical deformation rates of the top of the tunnel lining and the like can be realized, and an optimization result can be used as an effective tool for regulating and controlling construction parameters of a shield adjacent tunnel and regulating and controlling deformation of an existing tunnel.

Description

Technical Field

[0001] The present invention belongs to the technical field of shield construction control, and more specifically, relates to a deformation optimization method and device for a shield machine passing through an existing tunnel based on an intelligent algorithm. Background Art

[0002] The multi-objective optimization and control of deformation of existing tunnels caused by shield tunneling underpass construction mainly focuses on the multi-objective optimization of deformation of existing tunnels and the optimization of construction parameters. There are generally several types of crossing existing tunnels: overpass, underpass and side pass, among which underpass is the most common. During the shield tunneling process, soil disturbance will be caused, and the displacement field of the stratum will be transmitted to the existing tunnel, breaking the original soil balance. The existing tunnel structure and stratum will undergo vertical displacement, horizontal displacement and other deformations. When the longitudinal and horizontal displacements reach the limit deformation degree, it will affect the use function and structural safety of the existing tunnel structure. Many researchers have conducted extensive research on the response of existing tunnels in underpass engineering and proposed a series of control measures to greatly reduce the risks of existing tunnels. At present, the research on deformation of existing tunnels caused by shield tunneling mainly focuses on theoretical formulas, model tests, on-site monitoring, numerical simulation, data-driven machine learning (ML) and other methods. The defects of existing research on deformation control of underpass existing tunnels are mainly manifested in: (1) Traditional theoretical research, model tests and numerical simulations consider limited factors, inevitably have simplifications, often ignore some influencing factors, and the results are not accurate enough. At the same time, a lot of computational work is time-consuming; (2) On-site monitoring methods cannot predict the actual situation in advance and need to process a large amount of data; (3) Existing ML methods mainly focus on the prediction of deformation and stress of existing tunnels, and research on multi-objective optimization control of deformation underpasses of existing tunnels is extremely rare. In addition, during the construction of underpass shield tunnels, the horizontal distance between the front end of the tunnel boring machine face and the central axis of the intersection of the new and old tunnels (HDFCAUT) and the health status of the existing tunnel have an important impact on the deformation of the existing tunnel and the safe excavation of the new tunnel. Existing studies rarely consider the control of the deformation of the existing tunnel and the optimization of the construction parameters of the new tunnel under the influence of the health of the existing tunnel, different excavation intervals, and different time periods. In other words, how to consider the control of the deformation of the existing tunnel by adjusting the shield parameters before, during, and after the shield tunnel is an issue that deserves continued attention.

[0003] In order to overcome the limitations of traditional methods, ML has been widely used in shield-related parameters and deformation prediction in recent years, such as NGBoost, CNN, LSTM, etc. Wu, X., et al. used BO-NGBoost to predict the deformation of existing tunnels. Ye, X.-W., et al used LSTM model to establish intelligent prediction relationship between new tunnel excavation parameters and existing tunnel deformation. Although ML algorithms can obtain better evaluation results than traditional prediction methods, they have some shortcomings. NGBoost cannot handle complex problems that need to consider time and space effects; CNN has a convolutional weight value sharing structure, which greatly reduces the number of parameters of the neural network while reducing the complexity of the neural network model, giving it a strong feature self-learning ability to prevent overfitting. However, CNN ignores the long-term dependency between sequence data at the time level. Although LSTM can be used to solve time series problems, it is not effective for data with spatial features, and the single layer will have gradient vanishing or explosion problems as the output sequence grows. Other measures need to be taken, such as using multi-layer LSTM that can solve complex time series problems to make up for it. Summary of the invention

[0004] In view of the above defects or improvement needs of the prior art, the present invention provides a multi-objective optimization and control method and equipment for deformation of existing tunnels caused by shield tunneling close to underpassing construction based on BO-CNN-three layerLSTM-NSGA-Ⅲ, which is used to solve the multi-objective optimization and control of deformation of existing tunnels caused by shield tunneling close to underpassing construction and the optimization of construction parameters. The monitored sample data is simulated and trained using the BO-CNN-three layer LSTM model to provide a high-precision prediction function for NSGA-Ⅲ optimization. Then, NSGA-Ⅲ is used to perform multi-objective optimization to obtain the optimal shield segment parameter value, thereby achieving the purpose of multi-objective optimization and control of deformation of existing tunnels caused by shield tunneling close to underpassing construction.

[0005] To achieve the above-mentioned purpose, according to one aspect of the present invention, a deformation optimization method and device for shield tunneling under an existing tunnel based on CNN-three layer LSTM optimized by Bayesian (BO) combined with non-dominated sorting genetic algorithm (NSGA)-Ⅲ with elite strategy are proposed, which mainly includes the following steps: step one, determining the sensitive factors causing tunnel deformation at different approaching sections of the shield approaching tunnel construction stage, establishing an indicator system, collecting actual monitoring data during the shield tunneling process, and constructing a sample data set; step two, based on the sample data set, establishing and testing a prediction model for the lateral and vertical deformation of the top of the tunnel lining and the lateral and vertical deformation rate of the top of the tunnel lining based on BO-CNN-three layer LSTM; step three, using the nonlinear mapping relationship function obtained by the prediction model as the target optimization fitness function, constructing a multi-objective optimization model based on BO-CNN-threelayer LSTM-NSGA-Ⅲ, and proposing a recommended range for setting construction parameters under the condition of minimum deformation. The present invention combines BO-CNN-three layer LSTM with NSGA-Ⅲ. Through high-precision prediction of the lateral and vertical deformation of the tunnel lining top and the lateral and vertical deformation rate of the tunnel lining top, multi-objective optimization of the lateral and vertical deformation of the tunnel lining top and the lateral and vertical deformation rate of the tunnel lining top can be achieved, so that the optimization result is more in line with actual engineering needs. It can be used as an effective tool for shield tunneling proximity tunnel construction parameter control and existing tunnel deformation control.

[0006] Specifically, according to one aspect of the present invention, a deformation optimization method for a shield machine passing through an existing tunnel based on an intelligent algorithm is proposed, comprising the following steps:

[0007] Step 1: Determine the sensitive factors that cause tunnel deformation at different approaching sections during the shield tunneling construction phase, establish an indicator system, collect actual monitoring data during the shield tunneling process, and construct a sample data set;

[0008] Step 2: Establish a prediction model of the lateral and vertical deformation of the top of the tunnel lining and the lateral and vertical deformation rate of the top of the tunnel lining based on BO-CNN-three layer LSTM, and use the sample data set to train the prediction model to obtain the lateral and vertical deformation of the top of the tunnel lining and the nonlinear mapping relationship function of the lateral and vertical deformation of the top of the tunnel lining;

[0009] Step three, taking the nonlinear mapping relationship function as the target optimization fitness function, a multi-objective optimization model based on BO-CNN-three layer LSTM-NSGA-Ⅲ is constructed to propose the optimal shield segment parameter setting recommended values.

[0010] As further preferred, in step 1, the sensitive factors include soil bin pressure, cutter head speed, cutter head torque, propulsion speed, total thrust, screw machine speed, screw machine torque, grouting pressure, and horizontal distance from the working face to the intersection axis of the new and old tunnels;

[0011] The actual monitoring data include the lateral and vertical deformation of the top of the existing tunnel lining, and the lateral and vertical deformation rates of the top of the tunnel lining.

[0012] As a further preferred embodiment, in step 2, a grid search method is used to optimize the key parameters of the BO-CNN-three layer LSTM, and then a 5-fold cross validation method is used to select the optimal parameter combination, and a BO-CNN-three layer LSTM prediction model is established based on the optimal parameter combination;

[0013] Among them, BO is used to optimize the four hyperparameters of the model: learning_rate, minibatch_frac, col_sample, and n_estimators.

[0014] As a further preferred embodiment, step 2 further comprises:

[0015] The reliability of the prediction model was evaluated using three performance indicators: root mean square error, mean absolute error and mean absolute percentage error.

[0016] As further preferred, step three comprises the following steps:

[0017] (31) Determine a fitness function, and use the nonlinear mapping relationship function as the target optimization fitness function:

[0018] min1=min(BO-CNN-three layer LSTM regression (X i ))

[0019] min2=min(BO-CNN-three layer LSTM regression (X i ))

[0020] min3=min(BO-CNN-three layer LSTM regression (X i ))

[0021] ming4=min(BO-CNN-three layer LSTM regression (X i ))

[0022] Where, X iIt represents the soil bin pressure, cutter head speed, cutter head torque, propulsion speed, total thrust, screw machine speed, screw machine torque, grouting pressure, and the horizontal distance from the working face to the intersection axis of the new and old tunnels. ming1 is the target function of the horizontal displacement of the arch, ming2 is the target function of the vertical displacement of the arch, ming3 is the target function of the deformation rate of the vertical displacement of the arch, and ming4 is the target function of the deformation rate of the horizontal displacement of the arch.

[0023] (32) Determine the scope of constraints and set constraints on the values ​​of target decision parameters based on engineering requirements and existing project data:

[0024]

[0025] in, and Respectively represent the lower and upper limits of the input parameter value

[0026] (33) The NSGA-Ⅲ algorithm is used to realize the multi-objective optimization control of the deformation of the existing tunnel caused by the shield tunneling underpass and the optimization of construction parameters.

[0027] As a further preferred method, the NSGA-III algorithm is used to realize the multi-objective optimization control of the deformation of the existing tunnel caused by the shield tunneling underpass and the optimization of the construction parameters. The specific steps include:

[0028] (331): Set external solution set Used to store the algorithm and all the Pareto non-dominated solutions currently obtained, and calculate each weight vector λ j Generate the distance matrix d from the Euclidean distance of other vectors, select T weight vectors with the smallest Euclidean distance, and for each i=1,…,N, let B(i)={i1,i2,…,i T},λ 1 ,…,λ T is i The nearest T weight vectors, randomly initialize the population {x 1 ,x 2 ,…,x N}, x j is the solution to the jth subproblem. j , j = 1, 2, ..., N, calculate its objective function value FV j =F(x j ). According to MOP calculation, the initial ideal point z * =(z1,z2,…,z m ),satisfy j∈{1,2,…,N P};

[0029] (332) Randomly select two entities x from the set B(i) k and x l , using a specific recombination operator to make the parent individual x k and x l Perform gene crossover and mutation to generate new sub-individuals x new , use repair or heuristic methods to convert x new Repair to x re , for the new child individual x re Perform function evaluation and calculate F(x re ) function value, update the ideal point z according to the following rules * :

[0030]

[0031] Update the neighborhood solution, If g is satisfied tch (x re λ l ,z)≤g tch (x l |λ l ,z), then replace the individuals in the neighborhood, let x l =x re , FV l =F(x re ). Update EP and determine whether all individuals are related to x re The dominance relationship of EP is to remove all x re Dominate the vector, move in all the vectors not affected by x re Dominant vector;

[0032] (333) Repeat (332) until the number of iterations is reached, and output the vertical deformation, etc. as the Pareto solution set.

[0033] As a further preferred method, a fuzzy decision method is used to select an optimal solution from the Pareto solution set, as follows:

[0034] Each solution is scored based on its distance to the positive ideal solution and the negative ideal solution. The scoring formula for each solution is as follows:

[0035]

[0036] In the formula, S i is the score corresponding to the ith Pareto solution, f i - is the distance from the ith Pareto solution to the negative ideal solution, f i + is the distance from the ith Pareto solution to the positive ideal solution, f i - and fi + The calculation formula is as follows:

[0037]

[0038] In the formula, m is the number of optimization objectives, x ij is the ith Pareto solution under the jth objective, and is the input value corresponding to the negative ideal solution and the positive ideal solution under the jth objective, and the calculation formula is as follows:

[0039]

[0040] Calculate the percentage improvement for each Pareto solution, where the improvement for each objective is:

[0041]

[0042] The overall improvements for each solution are:

[0043]

[0044] In the formula, Imp ij is the improvement of the jth objective of each i-th Pareto solution, Org j,ave is the original mean value of the jth target, Opt ij is the optimal value of the jth objective of each i-th Pareto solution, Impoverall i is the overall improvement of each i Pareto solution.

[0045] According to another aspect of the present invention, a deformation optimization system for a shield machine passing through an existing tunnel based on an intelligent algorithm is provided, comprising: a first main control module, used to determine sensitive factors causing tunnel deformation at different approaching sections of the shield machine during the construction stage of approaching the tunnel, establish an indicator system, collect actual monitoring data during the shield machine excavation process, and construct a sample data set; a second main control module, used to establish a prediction model for the lateral and vertical deformation of the top of the tunnel lining and the lateral and vertical deformation rate of the top of the tunnel lining based on BO-CNN-three layer LSTM, and use the sample data set to train the prediction model to obtain the lateral and vertical deformation of the top of the tunnel lining and the nonlinear mapping relationship function of the lateral and vertical deformation of the top of the tunnel lining; a third main control module, used to use the nonlinear mapping relationship function as the target optimization fitness function, and construct a multi-objective optimization model based on BO-CNN-three layer LSTM-NSGA-Ⅲ to propose the optimal recommended values ​​for the setting of shield segment parameters.

[0046] According to another aspect of the present invention, there is also provided an electronic device, comprising:

[0047] At least one processor, at least one memory and a communication interface; wherein,

[0048] The processor, memory and communication interface communicate with each other;

[0049] The memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute a deformation optimization method for a shield machine passing through an existing tunnel based on an intelligent algorithm according to any of the above embodiments or a combination of multiple embodiments.

[0050] According to another aspect of the present invention, a non-transitory computer-readable storage medium is also provided, wherein the non-transitory computer-readable storage medium stores computer instructions, wherein the computer instructions enable the computer to execute a deformation optimization method for a shield machine passing through an existing tunnel based on an intelligent algorithm according to the above-mentioned embodiment or a combination of multiple embodiments.

[0051] In general, the above technical solution conceived by the present invention has the following technical advantages compared with the prior art:

[0052] 1. The present invention utilizes BO-CNN-three layer LSTM to combine decision trees and non-parametric statistics, and synthesizes regression or classification functions based on discrete or continuous data sets, which effectively avoids overfitting and has strong generalization. It realizes high-precision prediction of deformation of existing tunnels, and realizes multi-objective optimization and control of deformation of existing tunnels caused by shield tunneling underpass construction and optimization of construction parameters.

[0053] 2. The present invention uses NSGA-III with a fast non-dominated sorting operator, an individual crowding distance operator and an elite strategy selection operator to establish a multi-objective optimization model, which can quickly find the optimal solution and perform selective sorting, greatly improving the efficiency and accuracy of optimization.

[0054] 3. The present invention replaces the traditional mathematical model with the BO-CNN-three layer LSTM prediction model, uses it as the fitness function of target optimization, and combines it with NSGA-Ⅲ, which well reflects the complex nonlinear mapping relationship between shield construction parameters and existing tunnel deformation, and provides a basis for realizing the multi-objective optimization control of existing tunnel deformation caused by shield close-through construction and precise optimization of construction parameters.

[0055] 4. The present invention takes the existing lateral and vertical deformations of the top of the tunnel lining, and the lateral and vertical deformation rates of the top of the tunnel lining as optimization targets at the same time, and performs multi-target optimization, which can comprehensively consider the contradictory relationships among multiple targets and make the optimized scheme more in line with the actual needs of the project. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 Flow chart of the method of the present invention.

[0057] Figure 2 A green and intelligent design method for shield tunnels based on BO-CNN-three layer LSTM-NSGA-Ⅲ and a schematic diagram of the equipment structure are provided in an embodiment of the present invention.

[0058] Figure 3 A schematic diagram of the physical structure of an electronic device provided by an embodiment of the present invention.

[0059] Figure 4 This is a diagram of the hyperparameter optimization results of predicting the vertical displacement of the tunnel lining crown by using BO-CNN-three layer LSTM in an embodiment of the present invention.

[0060] Figure 5 This is a diagram of the hyperparameter optimization results of predicting the horizontal displacement of the tunnel lining crown by using BO-CNN-three layer LSTM in an embodiment of the present invention.

[0061] Figure 6 This is a diagram of the hyperparameter optimization results of predicting the vertical displacement deformation rate of the tunnel lining crown by using BO-CNN-three layer LSTM in an embodiment of the present invention.

[0062] Figure 7 This is a diagram showing the hyperparameter optimization result of predicting the horizontal displacement deformation rate of the tunnel lining vault by using BO-CNN-three layer LSTM in an embodiment of the present invention;

[0063] Figure 8 is a training curve for predicting the vertical displacement of the tunnel vault by using BO-CNN-three layer LSTM in an embodiment of the present invention;

[0064] Fig. 9 The BO-CNN-three layer LSTM in the embodiment of the present invention predicts the horizontal displacement training curve of the tunnel vault;

[0065] Fig.10 The BO-CNN-three layer LSTM in the embodiment of the present invention predicts the vertical displacement deformation rate training curve of the tunnel vault;

[0066] Fig.11 The training curve of BO-CNN-three layer LSTM predicting the horizontal displacement deformation rate of the tunnel vault in the embodiment of the present invention;

[0067] Fig.121 is a graph showing the prediction result of the vertical displacement of the tunnel vault predicted by BO-CNN-three layer LSTM in an embodiment of the present invention;

[0068] Fig.13 This is a graph showing the prediction result of the horizontal displacement of the tunnel vault using BO-CNN-three layer LSTM in an embodiment of the present invention;

[0069] Fig.14 1 is a graph showing the prediction result of the vertical displacement deformation rate of the tunnel vault predicted by BO-CNN-three layer LSTM in an embodiment of the present invention;

[0070] Fig.15 This is a graph showing the prediction result of the vertical displacement deformation rate of the tunnel vault predicted by BO-CNN-three layer LSTM in an embodiment of the present invention;

[0071] Fig.16 It is the Pareto solution of the four objectives in each section in the embodiment of the present invention. DETAILED DESCRIPTION

[0072] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0073] In order to achieve the above-mentioned purpose, the present invention provides a multi-objective optimization and control method for deformation of existing tunnels caused by shield tunneling close-to-underpassing construction based on BO-CNN-three layer LSTM-NSGA-Ⅲ, which is used to solve the multi-objective optimization and control of deformation of existing tunnels caused by shield tunneling close-to-underpassing construction and the optimization of construction parameters. The monitored sample data is simulated and trained using the BO-CNN-three layerLSTM model to provide a high-precision prediction function for NSGA-Ⅲ optimization. Then, NSGA-Ⅲ is used to perform multi-objective optimization to obtain the optimal shield segment parameter value, thereby achieving the purpose of multi-objective optimization and control of deformation of existing tunnels caused by shield tunneling close-to-underpassing construction. The method comprises the following steps:

[0074] (1) Construction of the index system of influencing factors

[0075] Through a combination of a large number of engineering practices and literature research, the existing tunnel deformation evaluation index system includes two parts: influencing factors and evaluation indicators. Sensitive factors mainly involve three categories: segment parameters, shield construction parameters and geological parameters, including soil bin pressure, cutterhead speed, cutterhead torque, propulsion speed, total thrust, screw machine speed, screw machine torque, grouting pressure, and the horizontal distance from the working face to the intersection axis of the new and old tunnels. The existing tunnel deformation and cost evaluation indicators are the lateral and vertical deformation of the tunnel lining top of the existing tunnel, and the lateral and vertical deformation rate of the tunnel lining top.

[0076] (2) Establishment and evaluation of BO-CNN-three layer LSTM prediction model

[0077] S1: Data processing and sample division

[0078] By collecting actual monitoring data, a total of 221 sets of typical sample data were obtained. In order to better train and verify the model, 80% of the sample data were randomly selected as the training set, and the remaining 20% ​​were used to test the model. The sample data was normalized between [-1, 1] to eliminate the impact of different sample feature sizes on prediction accuracy and efficiency. The calculation formula for normalization is as follows:

[0079]

[0080] Among them, x nor represents the standard value after normalization, and The default values ​​are 1 and 0, respectively, indicating the maximum and minimum values ​​after normalization. x represents the sample value, and x max and x min are the maximum and minimum sample values ​​respectively.

[0081] S2: Model parameter optimization

[0082] In order to prevent the CNN-three layer LSTM model from having under-learning problems and ensure the good generalization ability of the model to obtain better prediction results, it is necessary to optimize the model's hyperparameters and select the optimal parameter combination with the highest prediction accuracy. BO is used to optimize the main hyperparameters of the CNN-three layer LSTM model (see Table 1 for their meanings). A training model is established based on the Python platform, MSE is used as the loss function, and the number of training iterations is set to 200 generations to find the hyperparameter combination with the best training performance.

[0083] Table 1 Meaning and range of hyperparameters of BO-CNN-three layer LSTM model

[0084]

[0085] This paper uses BO for hyperparameter selection, takes the MSE of K-fold cross validation as the selection basis, finds the parameters corresponding to the model with the highest accuracy, and thus determines the final optimal parameters.

[0086] S3: Prediction model establishment and evaluation

[0087] The sample training set is used to train the BO-CNN-three layer LSTM prediction model for the lateral and vertical deformation of the top of the tunnel lining, and the lateral and vertical deformation rate of the top of the tunnel lining. Then the test set is used to verify the generalization performance and prediction accuracy of the BO-CNN-three layer LSTM prediction model for the lateral and vertical deformation of the top of the tunnel lining, and the lateral and vertical deformation rate of the top of the tunnel lining.

[0088] In order to effectively evaluate the prediction performance of the prediction model, three performance indicators, namely root mean square error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE), are used to evaluate the reliability of the model. The calculation formulas of the three performance indicators are as follows:

[0089]

[0090] In the formula, n is the number of data samples, and are the detection value and the prediction value respectively.

[0091] (3) NSGA-III multi-objective optimization

[0092] S1: Determine the objective function and constraints

[0093] Using the BO-CNN-three layer LSTM regression prediction algorithm for existing tunnel vault settlement, vault bottom horizontal displacement, vault bottom vertical displacement and settlement / d to replace the traditional mathematical function as the objective function in the multi-objective genetic algorithm can well solve the problem that the specific function expression cannot be given due to the complex nonlinear relationship between the input variables and the output targets. The objective functions of the lateral and vertical deformation of the top of the existing tunnel lining and the lateral and vertical deformation rate of the top of the tunnel lining based on BO-CNN-three layer LSTM can be expressed as:

[0094]

[0095] S2: Determine the constraints

[0096] In order to make the generated shield construction parameter combination more reasonable and feasible, it is also necessary to set a limit range for each decision variable according to the actual situation of the project and relevant specifications to form the constraint conditions of the variables. The general form of the constraint conditions is as follows:

[0097]

[0098] in, and They represent the lower and upper limits of the target parameter values ​​respectively.

[0099] S3: NSGA-Ⅲ multi-objective optimization

[0100] The NSGA-Ⅲ algorithm is used to realize the multi-objective optimization control of the deformation of the existing tunnel caused by the shield tunneling underpass and the optimization of construction parameters. The specific implementation steps are as follows:

[0101] Step 1: Initialization. Set the external solution set (elite population) Used to store all Pareto non-dominated solutions obtained by the algorithm. Calculate each weight vector λ j Generate the distance matrix d by the Euclidean distance with other vectors, and select T weight vectors with the smallest Euclidean distance. For each i=1,…,N, let B(i)={i1,i2,…, i T},λ 1 ,…,λ T is i The nearest T weight vectors. Randomly initialize the population {x 1 ,x 2 ,…,x N}, x j is the solution to the jth subproblem. j , j = 1, 2, ..., N, calculate its objective function value FV j =F(x j ). According to MOP calculation, the initial ideal point z * =(z1,z2,…,z m ),satisfy j∈{1,2,…,N P}.

[0102] Step 2: Individual update. Randomly select two individuals x from the set B(i) k and x l , using a specific recombination operator to make the parent individual x k and x l Perform gene crossover and mutation to generate new sub-individuals x new . Use repair or heuristic methods to convert x new Repair to x re For the new child individual x re Perform function evaluation and calculate F(x re ) function value. Update the ideal point z according to the following rules * :

[0103]

[0104] Update the neighborhood solution, If g is satisfied tch (x re λ l ,z)≤g tch (x l |λ l ,z), then replace the individuals in the neighborhood, let x l =x re , FV l =F(x re ). Update EP and determine whether all individuals are related to x re The dominance relationship of EP is to remove all x re Dominate the vector, move in all the vectors not affected by x re Dominant vector.

[0105] Step 3: Repeat step 2 until the number of iterations is reached, and output the vertical deformation as the Pareto solution set.

[0106] (4) TOPSIS method to obtain the optimal solution

[0107] The Pareto optimal solution obtained by the NSGA-Ⅲ algorithm is a set of solutions that meet the requirements of multi-objective optimization, corresponding to multiple objective parameter decision-making schemes. In actual tunnel engineering, only one decision-making scheme is needed to guide the multi-objective optimization and control of existing tunnel deformation caused by shield tunneling underpass construction and the optimization of construction parameters. Therefore, a certain method is needed to select an optimal solution from the Pareto solution set. The fuzzy decision method is used to select the optimal decision plan to avoid the influence of human subjective preferences on the results during the selection process, so as to obtain more scientific and objective results. The implementation process is as follows:

[0108] The Pareto optimal solution obtained by the NSGA-Ⅲ algorithm is a set of solutions that meet the Pareto optimal state decision variables. However, in actual tunnel construction, it is often necessary to guide the tunneling of the shield machine on site based on only one decision plan. To this end, the TOPSIS method is used to select the optimal solution from the Pareto solution set. The core is to score each solution based on its distance to the positive ideal solution and the negative ideal solution. The scoring calculation formula for each solution is as follows:

[0109]

[0110] In the formula, S i is the score corresponding to the ith Pareto solution, f i - is the distance from the ith Pareto solution to the negative ideal solution, f i +is the distance from the ith Pareto solution to the positive ideal solution. i - and f i + The calculation formulas are shown in (8) and (9):

[0111]

[0112] In the formula, m is the number of optimization objectives, x ij is the ith Pareto solution under the jth objective, and are the input values ​​corresponding to the negative ideal solution and the positive ideal solution under the jth objective, and the calculation formulas are shown in (10) and (11):

[0113]

[0114] In order to better present the degree of improvement of the optimization scheme, the improvement percentage of each Pareto solution is calculated. The improvement of each objective is calculated by formula (12), and the overall improvement of each scheme is calculated by formula (13):

[0115]

[0116] In the formula, Imp ij is the improvement of the jth objective of each i-th Pareto solution, Org j,ave is the original mean value of the jth target, Opt ij is the optimal value of the jth objective of each i-th Pareto solution, Impoverall i is the overall improvement of each i Pareto solution. For the existing tunnel deformation selected in this study, the larger the deformation, the more dangerous it is. Therefore, in TOPSIS, a very small index can be used to evaluate the target.

[0117] (5) Evaluation of BO-CNN-three layer LSTM-NSGA-Ⅲ multi-objective optimization results

[0118] In order to verify the effectiveness of the BO-CNN-three layer LSTM-NSGA-Ⅲ optimization model, the mean values ​​of shield machine parameters obtained by monitoring are compared with the optimized mean values, and the optimization ratios of various parameters are calculated to prove the feasibility of the provided method. The optimized data are compared with the construction monitoring data. It is proved that in actual engineering, the use of this method for multi-objective optimization is effective and feasible.

[0119] More specifically, in one embodiment of the present invention,

[0120] The multi-objective optimization control and construction parameter optimization method and equipment for existing tunnel deformation caused by shield tunneling close-to-underpass construction based on BO-CNN-three layer LSTM-NSGA-Ⅲ mainly include the following steps:

[0121] (1) Construction of influencing factor index system and sample data collection

[0122] Through a combination of a large number of engineering practices and literature research, 9 influencing factors including soil bin pressure, cutterhead speed, cutterhead torque, propulsion speed, total thrust, screw machine speed, screw machine torque, grouting pressure, and HDFCAUT were selected as input variables. The lateral and vertical deformation of the top of the existing tunnel lining in a certain section of a certain line of Wuhan Metro and the lateral and vertical deformation rate of the top of the tunnel lining were used as output variables. 1237 sets of field monitoring data were selected as sample data sets, some of which are shown in Table 2. 990 sets were randomly selected as training sets, and the remaining 247 sets were used as test sets.

[0123] Table 2 Input and output parameter data information

[0124]

[0125]

[0126] (2) BO-CNN-three layer LSTM parameter optimization

[0127] In order to make the BO-CNN-three layer LSTM regression model obtain better prediction results, this paper uses BO to optimize the four hyperparameters of the model: learning_rate, minibatch_frac, col_sample, and n_estimators, and combines the 5-fold cross-validation method to verify the accuracy of the model. Before optimizing the parameters, the initial ranges of the four parameters are given in Table 3. The hyperparameter search results for the lateral and vertical deformation of the top of the tunnel lining and the lateral and vertical deformation rate of the top of the tunnel lining are shown in Table 3. Figure 4 , Figure 5 , Figure 6 , Figure 7 shown.

[0128] Table 3 Meaning and range of hyperparameters of BO-CNN-three layer LSTM model

[0129]

[0130] Depend on Figure 4 , Figure 5 , Figure 6 , Figure 7It can be seen that the combination values ​​that minimize the prediction model errors of the lateral and vertical deformations of the tunnel lining top and the lateral and vertical deformation rates of the tunnel lining top are shown in Table 4 below.

[0131] Table 4 BO-CNN-three layer LSTM hyperparameter range setting and optimization results

[0132]

[0133]

[0134] (3) Establishment and evaluation of BO-CNN-three layer LSTM prediction model

[0135] Based on the BO-CNN-three layer LSTM parameter optimization results, the training set was used for learning simulation, and the prediction model training of the lateral and vertical deformation of the top of the tunnel lining and the lateral and vertical deformation rate of the top of the tunnel lining of the BO-CNN-three layer LSTM was established, respectively. Figure 8-11 , and then use the test set to test the prediction model of the training set. The prediction results are as follows: Figure 12-15 shown.

[0136] from Figure 8 and Fig.12 It can be seen that the BO-CNN-three layer LSTM model can better predict the vertical displacement of the tunnel vault. Figure 8 The results of BO-CNN-three layer LSTM prediction for tunnel roadbed settlement training set. The results show that the BO-CNN-three layer LSTM model has fully learned the rules between the input variables and output indicators of the training sample set. The results of predicting roadbed settlement through the training set are basically consistent with the actual observed values, with a mean square error of 0.011 and a goodness of fit of 0.970. Fig.12 Comparing the prediction results of the test set, it can be seen that the predicted value of the vault settlement in the test set is very consistent with the actual observed value, with a mean square error of 0.015 and a goodness of fit of 0.966. The results obtained from the training set and the test set both show that the BO-CNN-three layer LSTM prediction model has high accuracy and good generalization performance, and its fitted nonlinear prediction function has high accuracy.

[0137] Similarly, from Fig. 9 and Fig.13 , Fig.10 and Fig.14 , Fig.11 and Fig.15It can be seen that BO-CNN-three layerLSTM also has good accuracy in predicting the horizontal displacement, vertical and horizontal displacement deformation rate of the tunnel vault.

[0138] (4) Establishing the objective function

[0139] The BO-CNN-three layer LSTM regression function is introduced as the objective function of NSGA-III.

[0140] The objective function of vertical displacement of the vault based on BO-CNN-three layer LSTM ming1:

[0141] ming1(BO-CNN-three layer LSTM(χ1,χ2,χ3,χ4,χ5,χ6,χ7,χ8,χ9))

[0142] The objective function of vault horizontal displacement based on BO-CNN-three layer LSTM ming2:

[0143] ming2(BO-CNN-three layer LSTM(χ1,χ2,χ3,χ4,χ5,χ6,χ7,χ8,χ9))

[0144] The objective function of the vertical displacement deformation rate of the vault based on BO-CNN-three layer LSTM ming3:

[0145] ming3(BO-CNN-three layer LSTM(χ1,χ2,χ3,χ4,χ5,χ6,χ7,χ8,χ9))

[0146] The objective function of the horizontal displacement deformation rate of the vault based on BO-CNN-three layer LSTM ming4:

[0147] ming4(BO-CNN-three layer LSTM(χ1,χ2,χ3,χ4,χ5,χ6,χ7,χ8,χ9))

[0148] Among them, χ1, χ2, χ3, χ4, χ5, χ6, χ7, χ8, and χ9 represent the soil bin pressure, cutterhead speed, cutterhead torque, propulsion speed, total thrust, screw machine speed, screw machine torque, grouting pressure, and HDFCAUT (the horizontal distance from the working face to the intersection axis of the new and old tunnels), respectively.

[0149] (5) Determine the scope of constraints

[0150] Set the constraint range of each influencing factor parameter according to the actual engineering situation. When using the NSGA-Ⅲ algorithm to find the optimal solution for shield construction parameters, the decision range of the initial population must be set first to ensure that the initial population has practical significance. In order to avoid the large difference between the optimized shield machine parameter results and the actual engineering parameters of the shield machine, resulting in a large parameter adjustment span and possible safety hazards, this patent sets the range of the initial decision variables based on the fluctuation range of each input parameter in the selected sample as the main reference, see Table 5 below. In addition, the input parameter HDFCAUT (x9) has a range of [-2D, 2D].

[0151] Table 5 Reasonable range of shield parameters

[0152]

[0153]

[0154] (6) Multi-objective optimization and decision-making based on NSGA-Ⅲ

[0155] Before performing NSGA-Ⅲ multi-objective optimization, we first determine the number of genetic algorithm targets, population size, crossover and mutation operator values, and optimization stop criteria. Considering that appropriate population size and iteration times can promote the convergence of multi-objective optimization, this paper takes the crossover operator of NSGA-Ⅲ algorithm as 0.7, the mutation operator as 0.01, the population size as 200, and the maximum evolutionary generation and stop generation as 100. After the parameter setting is completed, the NSGA-Ⅲ algorithm is run, and 692 pairs of Pareto frontier solutions are obtained. The Pareto frontier solution diagram is shown in Fig.16 .

[0156] Further, based on the content of the above-mentioned device embodiment, as an optional embodiment, a deformation optimization system for shield underpass under existing tunnels based on intelligent algorithm provided in the embodiment of the present invention is used to implement and execute the method of any of the above-mentioned embodiments or a combination of multiple embodiments, including: a first main control module, used to determine the sensitive factors causing tunnel deformation in different approach sections of the shield approaching the tunnel construction stage, establish an indicator system, collect actual monitoring data during the shield excavation process, and construct a sample data set; a second main control module, used to establish a prediction model for the lateral and vertical deformation of the top of the tunnel lining and the lateral and vertical deformation rate of the top of the tunnel lining based on BO-CNN-three layer LSTM, and use the sample data set to train the prediction model to obtain the lateral and vertical deformation of the top of the tunnel lining and the nonlinear mapping relationship function of the lateral and vertical deformation of the top of the tunnel lining; a third main control module, used to use the nonlinear mapping relationship function as the target optimization fitness function, and construct a multi-objective optimization model based on BO-CNN-three layer LSTM-NSGA-Ⅲ to propose the optimal shield segment parameter setting recommended value.

[0157] The method of the embodiment of the present invention is implemented by relying on electronic devices, so it is necessary to introduce the relevant electronic devices. Based on this purpose, the embodiment of the present invention provides an electronic device, such as Figure 3 As shown, the electronic device includes: at least one processor, a communication interface, at least one memory, and a communication bus, wherein at least one processor, the communication interface, and at least one memory communicate with each other through the communication bus. At least one processor can call the logic instructions in at least one memory to execute all or part of the steps of the method provided by the aforementioned various method embodiments.

[0158] In addition, the logic instructions in the at least one memory mentioned above 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 this understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several 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 method embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0159] 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.

[0160] 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, it 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.

[0161] The flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the systems, methods and computer program products according to multiple embodiments of the present invention. Based on this understanding, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and sometimes in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or the flowchart, and the combination of the boxes in the block diagram and / or the flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0162] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A deformation optimization method for shield tunneling under an existing tunnel based on an intelligent algorithm, characterized in that: The following steps are involved: Step 1: Determine the sensitive factors that cause tunnel deformation at different approaching sections during the shield tunneling construction phase, establish an indicator system, collect actual monitoring data during the shield tunneling process, and construct a sample data set; Step 2: Establish a prediction model of the lateral and vertical deformation of the top of the tunnel lining and the lateral and vertical deformation rate of the top of the tunnel lining based on BO-CNN-three layer LSTM, and use the sample data set to train the prediction model to obtain the lateral and vertical deformation of the top of the tunnel lining and the nonlinear mapping relationship function of the lateral and vertical deformation of the top of the tunnel lining; Step three, taking the nonlinear mapping relationship function as the target optimization fitness function, a multi-objective optimization model based on BO-CNN-three layer LSTM-NSGA-Ⅲ is constructed to propose the optimal shield segment parameter setting recommended values.

2. According to the method of optimizing deformation of shield tunneling under existing tunnel based on intelligent algorithm in claim 1, it is characterized in that: In step 1, the sensitive factors include soil bin pressure, cutter head speed, cutter head torque, propulsion speed, total thrust, screw machine speed, screw machine torque, grouting pressure, and horizontal distance from the working face to the intersection axis of the new and old tunnels; The actual monitoring data include the lateral and vertical deformation of the top of the existing tunnel lining, and the lateral and vertical deformation rates of the top of the tunnel lining.

3. According to the method of optimizing deformation of shield tunneling under existing tunnel based on intelligent algorithm in claim 1, it is characterized in that: In step 2, the grid search method is used to optimize the key parameters of BO-CNN-three layer LSTM, and then the 5-fold cross-validation method is used to select the optimal parameter combination, and the BO-CNN-three layerLSTM prediction model is established based on the optimal parameter combination; Among them, BO is used to optimize the four hyperparameters of the model: learning_rate, minibatch_frac, col_sample, and n_estimators.

4. According to the method of optimizing deformation of shield tunneling under existing tunnel based on intelligent algorithm in claim 1, it is characterized in that: Step 2 also includes: The reliability of the prediction model was evaluated using three performance indicators: root mean square error, mean absolute error and mean absolute percentage error.

5. According to the method of optimizing deformation of shield tunneling under existing tunnel based on intelligent algorithm in claim 1, it is characterized in that: Step three includes the following steps: (31) Determine a fitness function, and use the nonlinear mapping relationship function as the target optimization fitness function: ming1=min(BO-CNN-three layer LSTM regression (X i )) ming2=min(BO-CNN-three layer LSTM regression (X i )) ming3=min(BO-CNN-three layer LSTM regression (X i )) ming4=min(BO-CNN-three layer LSTM regression (X i )) Where, X i It represents the soil bin pressure, cutter head speed, cutter head torque, propulsion speed, total thrust, screw machine speed, screw machine torque, grouting pressure, and the horizontal distance from the working face to the intersection axis of the new and old tunnels. ming1 is the target function of the horizontal displacement of the arch, ming2 is the target function of the vertical displacement of the arch, ming3 is the target function of the deformation rate of the vertical displacement of the arch, and ming4 is the target function of the deformation rate of the horizontal displacement of the arch. (32) Determine the scope of constraints and set constraints on the values ​​of target decision parameters based on engineering requirements and existing project data: in, and Respectively represent the lower and upper limits of the input parameter value (33) The NSGA-Ⅲ algorithm is used to realize the multi-objective optimization control of the deformation of the existing tunnel caused by the shield tunneling underpass and the optimization of construction parameters.

6. The method for optimizing deformation of a shield machine passing through an existing tunnel based on an intelligent algorithm according to claim 5 is characterized in that: The NSGA-Ⅲ algorithm is used to realize the multi-objective optimization control of existing tunnel deformation and construction parameter optimization caused by shield tunneling underpass construction. The specific steps include: (331): Set external solution set Used to store the algorithm and all the Pareto non-dominated solutions currently obtained, and calculate each weight vector λ j Generate the distance matrix d from the Euclidean distance of other vectors, select T weight vectors with the smallest Euclidean distance, and for each i=1,…,N, let B(i)={i1,i2,…,i T },λ 1 ,…,λ T is i The nearest T weight vectors, randomly initialize the population {x 1 ,x 2 ,…,x N }, x j is the solution to the jth subproblem. j , j = 1, 2, ..., N, calculate its objective function value FV j =F(x j ), according to MOP calculation, the initial ideal point z * =(z1,z2,…,z m ),satisfy j∈{1,2,…,N P }; (332) Randomly select two entities x from the set B(i) k and x l , using a specific recombination operator to make the parent individual x k and x l Perform gene crossover and mutation to generate new sub-individuals x new , use repair or heuristic methods to convert x new Repair to x re , for the new child individual x re Perform function evaluation and calculate F(x re ) function value, update the ideal point z according to the following rules * : Update the neighborhood solution, If g is satisfied tch (x re λ l ,z)≤g tch (x l |λ l ,z), then replace the individuals in the neighborhood, let x l =x re , FV l =F(x re ), update EP, and judge all individuals and x re The dominance relationship of EP is to remove all x re Dominate the vector, move in all the vectors not affected by x re Dominant vector; (333) Repeat (332) until the number of iterations is reached, and output the vertical deformation, etc. as the Pareto solution set.

7. The method for optimizing deformation of a shield machine passing through an existing tunnel based on an intelligent algorithm according to claim 6 is characterized in that: The fuzzy decision method is used to select an optimal solution from the Pareto solution set, as follows: Each solution is scored based on its distance to the positive ideal solution and the negative ideal solution. The scoring formula for each solution is as follows: In the formula, S i is the score corresponding to the ith Pareto solution, f i - is the distance from the ith Pareto solution to the negative ideal solution, f i + is the distance from the fth Pareto solution to the positive ideal solution, f i - and f i + The calculation formula is as follows: In the formula, m is the number of optimization objectives, x ij is the ith Pareto solution under the jth objective, and is the input value corresponding to the negative ideal solution and the positive ideal solution under the jth objective, and the calculation formula is as follows: Calculate the percentage improvement of each Pareto solution, where the improvement of each objective is: The overall improvements for each solution are: In the formula, Imp ij is the improvement of the jth objective of each i-th Pareto solution, Org j,ave is the original mean value of the jth target, Opt ij is the optimal value of the jth objective of each i-th Pareto solution, Impoverall i is the overall improvement of each i Pareto solution.

8. A deformation optimization system for shield tunneling under existing tunnels based on intelligent algorithm, characterized in that: include: The first main control module is used to determine the sensitive factors that cause tunnel deformation at different approaching sections during the shield tunneling construction phase, establish an indicator system, collect actual monitoring data during the shield tunneling process, and construct a sample data set; The second main control module is used to establish a prediction model of the lateral and vertical deformation of the top of the tunnel lining and the lateral and vertical deformation rate of the top of the tunnel lining based on BO-CNN-three layer LSTM, and train the prediction model using the sample data set to obtain the lateral and vertical deformation of the top of the tunnel lining and the nonlinear mapping relationship function of the lateral and vertical deformation of the top of the tunnel lining; The third main control module is used to use the nonlinear mapping relationship function as the target optimization fitness function, and to construct a multi-objective optimization model based on BO-CNN-three layer LSTM-NSGA-Ⅲ to propose the optimal shield segment parameter setting recommended values.

9. An electronic device, characterized in that: include: At least one processor, at least one memory and a communication interface; wherein, The processor, memory and communication interface communicate with each other; The memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the deformation optimization method for shield tunneling under an existing tunnel based on an intelligent algorithm as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, characterized in that: The non-transitory computer-readable storage medium stores computer instructions, and the computer instructions enable the computer to execute the deformation optimization method for shield tunneling under an existing tunnel based on an intelligent algorithm as described in any one of claims 1 to 7.

Images