Shield construction control method for fractured rock stratum close to building

Abstract

The invention discloses a shield construction control method for a fractured rock stratum close to a building. The shield construction control method comprises the steps that S1, monitoring data of a plurality of shield areas in the shield construction process are obtained; s2, constructing a first orthogonal test table; s3, obtaining a first training sample; s4, obtaining an optimized first regression prediction model; s5, performing inversion by adopting an intelligent optimization algorithm to obtain specific stratum parameter values; s6, constructing a second orthogonal test table; s7, obtaining a second training sample; s8, obtaining an optimized second regression prediction model; and S9, performing construction parameter optimization based on the optimized second regression prediction model by adopting an intelligent optimization algorithm and taking the construction cost for optimizing the construction parameters as an optimization target to obtain a corresponding construction parameter value when the construction cost is the lowest, and performing shield construction. According to the method, specific stratum parameter values are obtained through inversion, and the problem that accurate stratum parameters of tunnel surrounding rocks are difficult to accurately obtain is solved; and finally obtaining a construction parameter value corresponding to the lowest construction cost through the optimized second regression prediction model, and carrying out shield construction based on the construction parameter value, so that a theoretical basis is provided for shield method construction, and the construction efficiency is improved on the premise that the project progress and the construction cost are not influenced during construction. The influence on the ground surface and the building is reduced to the greatest extent.

Description

Technical Field

[0001] The invention relates to the technical field of shield construction control, and in particular to a shield construction control method for a fragmented rock layer adjacent to a building. Background Art

[0002] During the shield construction process, the shield machine will break the original soil pressure balance and disturb the surrounding rock in a certain area. Especially for the broken rock layer, due to the poor rock stability of the broken rock layer, problems such as stratum deformation, ground settlement, building settlement and tilt are likely to occur during the construction process; among them, ground settlement is an important factor leading to collapse and uneven settlement of adjacent buildings, but there are many factors affecting ground settlement. For example, when the shield method is used for construction, construction parameters such as soil bin pressure, face construction spacing, wall grouting range, grouting materials, excavation rate, etc. will have a significant impact on foundation settlement, surface settlement, etc. Therefore, it is of practical safety significance to study soil settlement and building deformation under different construction conditions of shield tunnels. However, in practice, it is difficult to accurately obtain the accurate stratum parameters of the tunnel surrounding rock, which makes it difficult to measure more accurate construction parameters. At the same time, the cost is not fully considered when selecting construction parameters, which may lead to the problem of excessively high construction costs. Summary of the invention

[0003] The present invention provides a shield construction control method for a fractured rock layer near a building, so as to overcome the technical problem of how to reasonably select construction parameters during shield construction and minimize the impact on the ground and buildings without affecting the project progress and cost.

[0004] In order to achieve the above object, the technical solution of the present invention is:

[0005] A shield construction control method for a fractured rock layer near a building, the specific steps comprising:

[0006] S1: Obtain monitoring data of several shield areas during shield construction;

[0007] S2: Determine the range of formation parameters in the shield area according to the geological survey report, so as to construct the first orthogonal test table characterizing the formation mechanical parameters;

[0008] S3: constructing a three-dimensional numerical model simulating a shield tunnel passing through a building, and performing a numerical simulation of the shield tunnel construction process based on the three-dimensional numerical model and a first orthogonal test table to obtain a first training sample;

[0009] S4: establishing a first regression prediction model and training it based on the first training sample, optimizing the parameters of the first regression prediction model during the training process to obtain an optimized first regression prediction model;

[0010] S5: using an intelligent optimization algorithm, and obtaining specific formation parameter values ​​based on the optimized first regression model and monitoring data inversion;

[0011] S6: Select several construction parameters in the shield construction process to construct a second orthogonal test table representing the values ​​of the construction parameters;

[0012] S7: performing numerical simulation of the shield construction process based on the specific formation parameter values, the three-dimensional numerical model and the second orthogonal test table to obtain a second training sample;

[0013] S8: establishing a second regression prediction model and training it based on the second training sample, optimizing the parameters of the second regression prediction model during the training process, and obtaining an optimized second regression prediction model;

[0014] S9: Adopting intelligent optimization algorithm and taking the construction cost of optimizing construction parameters as the optimization target, optimizing construction parameters is performed based on the optimized second regression prediction model, and based on the set deformation control value of the shield area, the corresponding construction parameter value when the construction cost is the lowest is obtained, and shield construction is performed.

[0015] Further, in S4, a first regression prediction model is established and trained based on the first training sample, and during the training process, parameters of the first regression prediction model are optimized to obtain the optimized first regression prediction model:

[0016] S41: establishing a least square vector machine model, i.e., the first regression prediction model, based on the support vector machine (SVM) standard algorithm;

[0017] S42: Training the first regression prediction model based on the first training sample, optimizing the model parameters in the first regression prediction model based on the cuckoo intelligent optimization algorithm during the training process, so as to obtain an optimized first regression prediction model.

[0018] Furthermore, in S41, the process of establishing the least squares vector machine model based on the support vector machine SVM standard algorithm is as follows:

[0019] The calculation expression of the support vector machine SVM standard algorithm is set as:

[0020]

[0021] In the formula, ω=[ω 1 ,ω 2 ,...,ω n ] T is the weight coefficient vector, is the mapping function, b is the offset; n is the data dimension;

[0022] The inequality constraint optimization problem in the standard support vector machine (SVM) algorithm is transformed into an equality constraint optimization problem through the risk minimization principle. The transformed equality constraint formula is:

[0023]

[0024] The Lagrange equation is used to transform the equality constraint problem of formula (2) into an unconstrained optimization problem. The formula is:

[0025]

[0026] Solving the partial derivatives of formula (3) yields:

[0027]

[0028] Finally, the unconstrained optimization problem is transformed into the Lagrangian operator α k The problem of solving the linear equations with the bias b is expressed as:

[0029]

[0030] In the formula, y=[y 1 ,y 2 ,...,y N ] is the output data of the training sample: L is an n-dimensional array, L = [l 1 ,l 2 ,...,l N ],α=[α 1 ,α 2 ,...,α N ]; k,l=1,2,...,N, I is a unit vector; γ is a regularization parameter representing the degree of interval fitting;

[0031] By solving formula (5), we can get α k and b, the least squares vector machine model, i.e. the first regression prediction model, is obtained, which is expressed as:

[0032]

[0033] In the formula, K(x,x k ) uses a Gaussian kernel function, and K(x,x k )=exp{-xx k 2 / σ 2},σ 2 represents the bandwidth of the support vector machine kernel function; x k is the input sample, and x k ∈R m .

[0034] Further, in S42, the first regression prediction model is trained based on the first training sample, and the process of optimizing the model parameters in the first regression prediction model based on the cuckoo intelligent optimization algorithm during the training process is:

[0035] S421: Determine the regularization parameter γ and square bandwidth σ corresponding to the bird's nest in the cuckoo intelligent optimization algorithm 2 The value range of

[0036] S422: Randomly generate the initial position of the bird's nest Represent the initial parameters γ and σ of the support vector machine respectively. 2 , i is the sequence number of the bird's nest, i=1,2,...,S, S is the total number of bird's nests;

[0037] S423: Calculate the degree of fit corresponding to each bird's nest position and find the current optimal bird's nest position And the optimal fit F, the optimal fit F is the smallest fit value, the optimal nest position is the nest position corresponding to the optimal fitness F, and the fitness F is expressed as:

[0038]

[0039] In the formula, y r is the deformation value in the first training sample, y r ' is the deformation value predicted by the first regression prediction model; q is the number of samples;

[0040] S424: Keep the best bird's nest position of the previous generation And update the location of the bird's nest to achieve optimized exploration of the search space. The update formula is:

[0041]

[0042] In the formula, k i (t+1) represents the location of the bird's nest at the t+1th iteration, k i (t) represents the position of the bird's nest at the tth iteration; α is the step scale factor, and α>0; ⊕ is point-to-point multiplication; L(λ) represents the random search path, i is the nest number, and S is the total number of nests;

[0043] S425: Find the optimal nest location after update And calculate the corresponding fitness F'. If the fitness F' corresponding to the updated optimal nest position is less than the fitness F corresponding to the previous optimal nest position, then replace the previous optimal nest position with the updated optimal nest position, otherwise retain the previous optimal nest position.

[0044] S426: Determine whether the current number of iterations reaches the maximum number of iterations. If so, output γ, σ 2 The value of is assigned to the first regression prediction model, thereby obtaining the optimized first regression prediction model; otherwise, executing S427;

[0045] S427: Update the position of the bird's nest again according to the rule of S424, and then operate according to the rules of S425-S426.

[0046] Furthermore, in S5, the process of using the intelligent optimization algorithm and inverting the optimized first regression model and the monitoring data to obtain specific formation parameter values ​​is as follows:

[0047] S51: Set the initial parameters of the cuckoo optimization algorithm and randomly generate the initial position of the bird's nest are the initial positions of M bird nests randomly generated within a given range based on various formation parameters, and n is the number of formation parameters;

[0048] S52: Calculate the fitness value corresponding to each bird's nest and find the current optimal bird's nest location And the optimal fit F", calculated as:

[0049]

[0050] In the formula, y l is the monitoring data, y l ' is the deformation value predicted by the optimized first regression model; M is the number of monitoring points;

[0051] S53: retains the best bird's nest position of the previous generation And update the location of the bird's nest according to the rule of S424;

[0052] S54: Find the updated optimal bird's nest position and calculate the corresponding fitness. If the fitness corresponding to the updated optimal bird's nest position is less than the fitness F' corresponding to the previous optimal bird's nest position, replace the previous optimal bird's nest position with the updated optimal bird's nest position; otherwise, retain the previous optimal bird's nest position.

[0053] S55: Determine whether the current iteration number reaches the maximum iteration number, if so, obtain the specific formation parameter value, otherwise, execute S56;

[0054] S56: Update the position of the bird's nest again according to the rule of S53, and then operate according to the rules of S54-S55.

[0055] Furthermore, in S9, an intelligent optimization algorithm is used and the construction cost of the optimized construction parameters is taken as the optimization target. The construction parameters are optimized based on the optimized second regression prediction model. The process of obtaining the corresponding construction parameter value when the construction cost is the lowest is as follows:

[0056] S91: Set the initial parameters of the cuckoo optimization algorithm and randomly generate the initial position k of the bird's nest i ” (0) , are the initial positions of P bird nests randomly generated within a given range based on each construction parameter, and v is the number of construction parameters;

[0057] S92: Input the construction parameters into the optimized second regression prediction model to calculate the response deformation, and express the cost generated by the construction according to the construction parameters as follows:

[0058]

[0059] Where L' is the length of the interval tunnel, is the soil bin pressure, k 1 is the basic fee per meter, m 1 is the pressure adjustment factor, is the construction spacing of the tunnel face, k 2 is the adjustment cost of shield spacing, k 3 The cost per cubic meter of grouting, is the grouting range behind the wall, m 3 is the arc length of the grouting layer;

[0060] At the same time, the lowest construction cost is used as the fitness function to find the current optimal bird's nest location and the optimal fit. The fitness function is expressed as:

[0061]

[0062] In the formula, is the construction parameter value; They represent the calculated ground settlement, left-line segment settlement, right-line segment settlement and building displacement respectively; 1max ,y 2max ,y 3max ,y 4max They represent the control value of ground settlement, the control value of left-line segment settlement, the control value of right-line segment settlement and the control value of building displacement respectively;

[0063] S93: retain the optimal bird's nest position of the previous generation, and update the position of the bird's nest according to the rule of S424;

[0064] S94: Find the updated optimal bird's nest position and calculate the corresponding fitness. If the fitness corresponding to the updated optimal bird's nest position is less than the fitness corresponding to the previous optimal bird's nest position, replace the previous optimal bird's nest position with the updated optimal bird's nest position; otherwise, retain the previous optimal bird's nest position.

[0065] S95: Determine whether the current number of iterations reaches the maximum number of iterations. If so, obtain the construction parameter value corresponding to the lowest construction cost. Otherwise, execute S96.

[0066] S96: Update the position of the bird's nest again according to the rules of S93, and then operate according to the rules of S94-S95.

[0067] Beneficial effects: The present invention constructs an optimized first regression prediction model and monitoring data of the shield area to invert specific formation parameter values, thereby constructing a second training sample, overcoming the problem of difficulty in accurately obtaining accurate formation parameters of tunnel surrounding rock; the second regression prediction model is trained and optimized through the second training sample to obtain an optimized second regression prediction model; an intelligent optimization algorithm is used and the construction cost of the optimized construction parameters is taken as the optimization target, and construction parameters are optimized based on the optimized second regression prediction model, and finally the construction parameter values ​​corresponding to the lowest construction cost are obtained, and shield construction is performed based on the construction parameter values, thereby providing a theoretical basis for shield construction and providing reliable design parameters for actual construction, forming a stable, safe and reliable shield construction plan, and providing a basis for engineering practice; ensuring that during construction, through the reasonable selection of shield construction parameters, the impact on the surface and buildings is minimized without affecting the progress and cost of the project. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0069] Figure 1 It is a flow chart of the shield construction control method of the fragmented rock layer near the building in the present invention;

[0070] Figure 2 A model diagram for performing numerical calculations in an embodiment of the present invention;

[0071] Figure 3 It is a survey line arrangement diagram in an embodiment of the present invention;

[0072] Figure 4 The following is a layout diagram of some sensors in an embodiment of the present invention;

[0073] Figure 5 A graph showing changes in settlement values ​​for some surface settlement measurement lines in an embodiment of the present invention;

[0074] Figure 6 It is a settlement curve diagram of some axial segments in the embodiments of the present invention;

[0075] Figure 7 The graphs are simulated and measured surface settlement curves of the embodiments of the present invention. DETAILED DESCRIPTION

[0076] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0077] This embodiment provides a shield construction control method for a broken rock layer near a building. Figure 1 As shown, the specific steps include:

[0078] S1: Obtain monitoring data of several shield areas during shield construction;

[0079] Specifically, before shield construction is carried out on the fractured rock layer near the building, sensors are deployed at the surface settlement monitoring points in the shield area, the left-line segment settlement monitoring points, the right-line segment settlement monitoring points and the building displacement monitoring points, and the surface settlement data, the left-line segment settlement data, the right-line segment settlement data and the building displacement data are obtained;

[0080] S2: Determine the range of formation parameters in the shield area according to the geological survey report, so as to construct the first orthogonal test table characterizing the formation mechanical parameters;

[0081] Specifically, a stratum to be subjected to inversion parameter operation is selected according to the geological survey report, and a value range of the elastic modulus and a value range of the Poisson's ratio of the stratum are obtained; the value range of the elastic modulus and the value range of the Poisson's ratio of the stratum are divided, and parameter levels are set, thereby constructing a first orthogonal test table;

[0082] Specifically, in this embodiment, the orthogonal test table is determined according to the number of factors and the number of divided levels, wherein the number of factors is the number of types of learning sample data, and the levels are usually evenly divided into 2 to 5 levels according to the parameter value range. In this embodiment, the number of factors is first determined, and a suitable orthogonal scheme is selected according to the orthogonal design table corresponding to the number of factors, and the parameters are divided according to the number of levels corresponding to the scheme, thereby constructing an orthogonal test table.

[0083] S3: Based on FlAC 3D The numerical model constructs a three-dimensional numerical model simulating the shield tunneling under the building, and performs a numerical simulation of the shield tunneling process based on the three-dimensional numerical model and the first orthogonal test table to obtain a first training sample;

[0084] Specifically, in this embodiment, a range analysis is performed on the calculation results obtained based on the first orthogonal test table to obtain the order of influence of the range values ​​of different formation parameter combinations on the surface settlement, left-line segment settlement, right-line segment settlement and building displacement changes; based on the range analysis, parameters that meet the set sensitivity are selected according to actual needs to form the first training sample.

[0085] S4: establishing a first regression prediction model and training it based on the first training sample, optimizing the parameters of the first regression prediction model during the training process to obtain an optimized first regression prediction model;

[0086] Specifically, the first training samples include different schemes and the corresponding calculated ground settlement, left line segment settlement, right line segment settlement and building displacement;

[0087] In a specific embodiment, in S4, a first regression prediction model is established and trained based on a first training sample, and parameters of the first regression prediction model are optimized during the training process to obtain the optimized first regression prediction model.

[0088] S41: establishing a least square vector machine model, i.e., the first regression prediction model, based on the support vector machine (SVM) standard algorithm;

[0089] In a specific embodiment, in S41, the process of establishing a least squares vector machine model based on the support vector machine SVM standard algorithm is:

[0090] The calculation expression of the support vector machine SVM standard algorithm is set as:

[0091]

[0092] In the formula, ω=[ω 1 ,ω 2 ,...,ω n ] T is the weight coefficient vector, is the mapping function, b is the offset; n is the data dimension;

[0093] The inequality constraint optimization problem in the standard support vector machine (SVM) algorithm is transformed into an equality constraint optimization problem through the risk minimization principle. The transformed equality constraint formula is:

[0094]

[0095] The Lagrange equation is used to transform the equality constraint problem of formula (2) into an unconstrained optimization problem. The formula is:

[0096]

[0097] Solving the partial derivatives of formula (3) yields:

[0098]

[0099] Finally, the unconstrained optimization problem is transformed into the Lagrangian operator α k The problem of solving the linear equations with the bias b is expressed as:

[0100]

[0101] In the formula, y=[y 1 ,y 2 ,...,y N ] is the output data of the training sample: L is an n-dimensional array, L = [l 1 ,l 2 ,...,l N ],α=[α 1 ,α 2 ,...,α N ]; k,l=1,2,...,N, I is a unit vector; γ is a regularization parameter representing the degree of interval fitting;

[0102] By solving formula (5), we can get α k and b, the least squares vector machine model, i.e. the first regression prediction model, is obtained, which is expressed as:

[0103]

[0104] In the formula, K(x,x k ) uses a Gaussian kernel function, and K(x,x k )=exp{-xx k 2 / σ 2},σ 2 represents the bandwidth of the support vector machine kernel function, which determines the final fitting accuracy.k is the input sample, and x k ∈R m .

[0105] S42: Training the first regression prediction model based on the first training sample, optimizing the model parameters in the first regression prediction model based on the cuckoo intelligent optimization algorithm during the training process, so as to obtain an optimized first regression prediction model.

[0106] In a specific embodiment, in S42, the first regression prediction model is trained based on the first training sample, and the process of optimizing the model parameters in the first regression prediction model based on the cuckoo intelligent optimization algorithm during the training process is:

[0107] S421: Determine the regularization parameter γ and square bandwidth σ corresponding to the bird's nest in the cuckoo intelligent optimization algorithm 2 The value range of

[0108] S422: Randomly generate the initial position of the bird's nest Represent the initial parameters γ and σ of the support vector machine respectively. 2 , i is the sequence number of the bird's nest, i=1,2,...,S, S is the total number of bird's nests;

[0109] S423: Calculate the degree of fit corresponding to each bird's nest position and find the current optimal bird's nest position And the optimal fit F, the optimal fit F is the smallest fit value, the optimal nest position is the nest position corresponding to the optimal fitness F, and the fitness F is expressed as:

[0110]

[0111] In the formula, y r is the deformation value in the first training sample, y r ' is the deformation value predicted by the first regression prediction model; q is the number of samples;

[0112] S424: Keep the best bird's nest position of the previous generation And update the location of the bird's nest to achieve optimized exploration of the search space. The update formula is:

[0113]

[0114] In the formula, k i (t+1) represents the location of the bird's nest at the t+1th iteration, k i (t)represents the position of the bird's nest at the tth iteration; α is the step size proportional factor, which is used to control the scaling of the step size, and α>0; ⊕ is the point-to-point multiplication, which is used to calculate the specific method of position update; L(λ) represents the random search path, i is the nest number, and S is the total number of nests;

[0115] S425: Find the optimal nest location after update And calculate the corresponding fitness F'. If the fitness F' corresponding to the updated optimal nest position is less than the fitness F corresponding to the previous optimal nest position, then replace the previous optimal nest position with the updated optimal nest position, otherwise retain the previous optimal nest position.

[0116] S426: Determine whether the current number of iterations reaches the maximum number of iterations. If so, output γ, σ 2 The value of is assigned to the first regression prediction model, thereby obtaining the optimized first regression prediction model; otherwise, executing S427;

[0117] S427: Update the position of the bird's nest again according to the rule of S424, and then operate according to the rules of S425-S426.

[0118] Specifically, after the above steps, the optimized first regression prediction model representing the nonlinear relationship between the formation parameters and the deformation parameters can be obtained, which is expressed as:

[0119] Y=LSSVM(X) (9)

[0120] X=[x 1 ,x 2 ,...,x n ] T ,Y=[y 1 ,y 2 ,...,y m ] T

[0121] In the formula, x 1 ,x 2 ,...,x n To represent the formation parameter value, y 1 ,y 2 ,...,y m is the corresponding deformation value, n is the number of formation parameters, and m is the number of deformation parameters.

[0122] S5: using an intelligent optimization algorithm, and obtaining specific formation parameter values ​​based on the optimized first regression model and monitoring data inversion;

[0123] Specifically, in this embodiment, specific elastic modulus and Poisson's ratio values ​​are obtained based on the optimized first regression model and monitoring data inversion.

[0124] In a specific embodiment, in S5, the process of using an intelligent optimization algorithm and inverting the optimized first regression model and monitoring data to obtain specific formation parameter values ​​is as follows:

[0125] S51: Set the initial parameters of the cuckoo optimization algorithm and randomly generate the initial position of the bird's nest are the initial positions of M bird nests randomly generated within a given range based on various formation parameters, and n is the number of formation parameters. During the inversion process, each bird nest position corresponds to an initial solution of the formation parameters.

[0126] S52: Calculate the fitness value corresponding to each bird's nest and find the current optimal bird's nest position k b (t) And the optimal fit F", calculated as:

[0127]

[0128] In the formula, y l is the monitoring data, y l ' is the deformation value predicted by the optimized first regression model; M is the number of monitoring points;

[0129] S53: retains the best bird's nest position of the previous generation And update the location of the bird's nest according to the rule of S424;

[0130] S54: Find the updated optimal bird's nest position and calculate the corresponding fitness. If the fitness corresponding to the updated optimal bird's nest position is less than the fitness F' corresponding to the previous optimal bird's nest position, replace the previous optimal bird's nest position with the updated optimal bird's nest position; otherwise, retain the previous optimal bird's nest position.

[0131] S55: Determine whether the current iteration number reaches the maximum iteration number, if so, obtain the specific formation parameter value, otherwise, execute S56;

[0132] S56: Update the position of the bird's nest again according to the rule of S53, and then operate according to the rules of S54-S55.

[0133] Specifically, in this embodiment, the nonlinear relationship between the formation parameters and the response deformation is described according to the optimized first regression prediction model established as above, the monitoring data is input into the optimized first regression model to calculate the response deformation value, and the calculated response deformation value and the monitoring data are substituted into the fitness function to obtain the optimal formation parameter value.

[0134] S6: Select several construction parameters that have a great influence on the shield construction process, so as to construct a second orthogonal test table representing the values ​​of the construction parameters;

[0135] Specifically, in this embodiment, the influence of various construction parameters on soil deformation is analyzed, and the construction parameters with greater influence, such as soil bin pressure, face construction spacing and behind-the-wall grouting range, are found, so as to construct the second orthogonal test table, thereby providing a theoretical basis for the construction of the shield method and providing reliable design parameters for actual construction.

[0136] S7: By FlAC 3D The software performs numerical simulation of the shield construction process based on the specific formation parameter values, the three-dimensional numerical model and the second orthogonal test table to obtain a second training sample;

[0137] Specifically, based on the specific elastic modulus and Poisson's ratio values ​​and the three-dimensional numerical model, numerical calculations are performed on different schemes in the second orthogonal test table to form a second training sample. The second training sample includes different schemes and the calculated corresponding surface settlement, left-line segment settlement, right-line segment settlement and building displacement.

[0138] S8: establishing a second regression prediction model and training it based on the second training sample, optimizing the parameters of the second regression prediction model during the training process, and obtaining an optimized second regression prediction model;

[0139] Specifically, in this embodiment, a second regression prediction model is established and trained based on the second training sample. During the training process, the parameters of the second regression prediction model are optimized. The process of obtaining the optimized second regression prediction model is the same as the rules of S41 and S42, which will not be repeated here. At the same time, when the cuckoo optimization algorithm is used to optimize the parameters in the second regression prediction model, the fitness The calculation formula is:

[0140]

[0141] In the formula, is the deformation value in the second training sample; is the predicted deformation value;

[0142] The optimized second regression prediction model is obtained through optimization, which is used to characterize the nonlinear relationship between construction parameters and response deformation, and is expressed as:

[0143]

[0144] In the formula, is the construction parameter value, is the corresponding deformation value of the measuring point, v is the number of construction parameters, and w is the number of corresponding deformation parameters.

[0145] S9: Adopting intelligent optimization algorithm and taking the construction cost of optimizing construction parameters as the optimization target, optimizing construction parameters is performed based on the optimized second regression prediction model, and based on the set deformation control value of the shield area, the corresponding construction parameter value when the construction cost is the lowest is obtained, and shield construction is performed.

[0146] In a specific embodiment, in S9, an intelligent optimization algorithm is used and the construction cost of the optimized construction parameters is optimized as the optimization target. The construction parameters are optimized based on the optimized second regression prediction model. The process of obtaining the construction parameter value corresponding to the lowest construction cost is as follows:

[0147] S91: Set the initial parameters of the cuckoo optimization algorithm and randomly generate the initial position of the bird's nest The initial positions of P bird nests are randomly generated within a given range based on various construction parameters, and v is the number of construction parameters; each bird nest position corresponds to an initial solution of the construction parameters.

[0148] S92: Input the construction parameters into the optimized second regression prediction model to calculate the response deformation, and express the cost generated by the construction according to the construction parameters as follows:

[0149]

[0150] Where L' is the length of the interval tunnel, is the soil bin pressure, k 1 is the basic fee per meter, m 1 is the pressure adjustment factor, is the construction spacing of the tunnel face, k 2 is the adjustment cost of shield spacing, k 3 The cost per cubic meter of grouting, is the grouting range behind the wall, m 3 is the arc length of the grouting layer;

[0151] At the same time, the lowest construction cost is used as the fitness function to find the current optimal bird's nest location and the optimal fit. The fitness function is expressed as:

[0152]

[0153] In the formula, is the construction parameter value; They represent the calculated ground settlement, left-line segment settlement, right-line segment settlement and building displacement respectively; 1max ,y 2max ,y 3max ,y 4max They represent the control value of ground settlement, the control value of left-line segment settlement, the control value of right-line segment settlement and the control value of building displacement respectively;

[0154] S93: retain the optimal bird's nest position of the previous generation, and update the position of the bird's nest according to the rule of S424;

[0155] S94: Find the updated optimal bird's nest position and calculate the corresponding fitness. If the fitness corresponding to the updated optimal bird's nest position is less than the fitness corresponding to the previous optimal bird's nest position, replace the previous optimal bird's nest position with the updated optimal bird's nest position; otherwise, retain the previous optimal bird's nest position.

[0156] S95: Determine whether the current number of iterations reaches the maximum number of iterations. If so, obtain the construction parameter value corresponding to the lowest construction cost. Otherwise, execute S96.

[0157] S96: Update the position of the bird's nest again according to the rules of S93, and then operate according to the rules of S94-S95.

[0158] Specifically, in order to determine the specific formation parameter values ​​suitable for numerical simulation, improve the calculation accuracy, and overcome the problem of difficulty in accurately obtaining the accurate mechanical parameters of the tunnel surrounding rock, this embodiment obtains the formation parameters of the shield tunnel surrounding rock through the inversion method, performs numerical simulation again, obtains the second training sample, and uses the second training sample as the learning sample, and optimizes and trains the LSSVM (Least-Squares Support Vector Machine), i.e., the second regression prediction model, in combination with the cuckoo optimization algorithm, to obtain the optimized second regression prediction model to describe the nonlinear relationship between the construction parameters and the engineering structure response deformation. Further, the cuckoo optimization algorithm is used again and the construction cost is considered to achieve the optimization of the construction parameters, and the corresponding construction parameter values ​​when the construction cost is the lowest are obtained.

[0159] Specifically, in this embodiment, shield construction is performed by determining the soil bin pressure, the construction spacing of the tunnel face and the grouting range behind the wall, thereby providing a theoretical basis for the construction of the shield method and providing reliable design parameters for actual construction.

[0160] Specifically, the present invention designs a shield construction control method for fractured rock strata adjacent to buildings based on support vector machine, and improves the accuracy of calculation results through machine learning algorithms, avoiding errors caused by previous manual subjective evaluation or numerical simulation calculations, and providing an accurate and safe evaluation method for determining shield construction parameters.

[0161] In this embodiment, a subway section in Nanchang that passes under a nearby building is used as a project support to further illustrate the design principle and specific workflow of this embodiment;

[0162] First, according to the position of the shield, sensors are deployed at the surface settlement monitoring points in the shield area, the settlement monitoring points of the left-line segment, the settlement monitoring points of the right-line segment, and the displacement monitoring points of the buildings. The relative positions of the buildings and the arrangement of the monitoring points are as follows: Figure 3 and Figure 4 As shown in Table 1; According to the geological survey results and specifications of the engineering broken rock mass, the distribution ranges of the elastic modulus of the miscellaneous fill, the Poisson's ratio of the miscellaneous fill, the elastic modulus of the silty clay, the Poisson's ratio of the silty clay, the elastic modulus of the medium sand and the Poisson's ratio of the medium sand in this section are obtained, as shown in Table 1:

[0163] Table 1:

[0164] parameter Minimum Maximum Elastic modulus of miscellaneous fill / GPa 0.01 0.05 Poisson's ratio of miscellaneous fill / μ 0.22 0.26 Elastic modulus of silty clay / GPa 0.02 0.06 Poisson's ratio of silty clay / μ 0.28 0.32 Medium sand elastic modulus / GPa 0.20 1.00 Poisson's ratio of medium sand / μ 0.25 0.29

[0165] According to the survey results, the first orthogonal experimental table was constructed, and a five-level design was performed on the six parameters of elastic modulus of miscellaneous fill, Poisson's ratio of miscellaneous fill, elastic modulus of silty clay, Poisson's ratio of silty clay, elastic modulus of medium sand and Poisson's ratio of medium sand, forming a total of 25 groups of orthogonal samples. The parameter orthogonal experimental table is shown in Table 2.

[0166] Table 2:

[0167]

[0168]

[0169] Based on the engineering example, the soil around the tunnel with an area of ​​4.5 times the diameter is retained, the bottom of the model is fixedly constrained, and the boundaries on both sides are constrained in the normal direction. The tunnel section is established according to the tunnel design drawings, and the surrounding rock and soil are established according to the on-site survey drawings. Finally, a three-dimensional numerical model simulating the shield construction process is established according to the situation of the shield passing under the building. The positions and sizes of each structure are as follows: Figure 2 The 3D model is divided into units. In order to statistically analyze the deformation law, the model arranges multiple measurement lines. The measurement points are consistent with the on-site monitoring points. The position and coordinate information of each point are shown in Figure 4 And as shown in Table 3.

[0170] Table 3:

[0171]

[0172]

[0173] Based on the three-dimensional numerical model, numerical calculations are performed on different schemes in the first orthogonal test table, and the surface settlement, left-line segment settlement, right-line segment settlement, and building displacement values ​​generated during the construction simulation are recorded. The first orthogonal test table and the calculation results of the numerical model are used as the first training sample, the first 20 groups of data are used as learning samples, and the last 5 groups of data are used as prediction test samples. Specific algorithm parameters are set, and the cuckoo algorithm is used to select the optimal parameters of the least squares vector machine model. The maximum number of iterations of the cuckoo algorithm is set to 2000, and the probability of the bird's nest being discovered P is set to 0. a =0.25, the number of bird nests is 50, the upper and lower limits of the regularization parameter γ of the interval fitting degree of the support vector machine are [0,10]; set the bandwidth σ of the support vector machine kernel function 2 The upper and lower limits are [0,1]. The parameters of the first regression prediction model obtained by the cuckoo algorithm have the best values ​​of γ = 1.01, σ 2 = 0.17. Set γ = 1, σ 2 =0.1 is substituted into the least square vector machine model to obtain the optimized first regression prediction model. The deformation data of four measuring points (one surface settlement measuring point, one left line segment settlement measuring point, one right line segment settlement measuring point and one building displacement measuring point) selected from the prediction test samples of the last five groups are compared with the results of the first regression prediction model as shown in Table 4. From the data in the table below, it can be seen that the maximum prediction error of the deformation data is 3.2%, which is relatively small and can be basically ignored, indicating that the first regression prediction model established by the present invention has high reliability and accuracy.

[0174] Table 4:

[0175]

[0176] The on-site shield construction monitoring results are Figures 5 to 7 The deformation values ​​of each measuring point shown in are substituted into the optimized first regression prediction model to determine the formation parameters of the construction site. The output elastic modulus of miscellaneous fill is 0.02GPa, Poisson's ratio of miscellaneous fill is 0.24, elastic modulus of silty clay is 0.03GPa, Poisson's ratio of silty clay is 0.28, elastic modulus of medium sand is 0.62GPa, and Poisson's ratio of medium sand is 0.26. The actual construction on site is guided by the obtained formation parameters. Compared with the previous geological survey data, the formation parameters obtained by this prediction model are more accurate, indicating that the formation parameters obtained based on this method have good reliability.

[0177] By using numerical simulation, the effects of different construction parameters on the deformation of the strata and buildings are compared. The construction parameters that have a greater impact on the deformation of the soil and buildings during the shield construction process are selected, namely different soil bin pressures, different face construction spacings and different back-wall grouting ranges. The construction parameters are divided into three levels, as shown in Table 5. The soil pressure at the site is about 0.40MPa, so it will increase by 37.5%

[0178] (0.55MPa) and 75% (0.70MPa) as the comparison conditions; the construction spacing is set to 0 ring behind the right hole, 8 rings behind the right hole, and 16 rings behind the right hole; the equivalent grouting thickness of the grouting ring is divided into 30cm, 40cm, and 50cm. According to the number of parameters and the number of levels, a second orthogonal scheme with three parameters and three levels can be formed, as shown in Table 6.

[0179] Table 5:

[0180] Parameter Type Soil pressure Face construction spacing Behind-the-wall grouting range Level 1 0.70MPa Right hole 16 rings behind 50cm Level 2 0.55MPa Right hole 8 rings behind 40cm Level 3 0.40MPa Right hole 0 rings behind 30cm

[0181] Table 6:

[0182]

[0183] By FlAC 3D Software, using the numerical simulation method of shield construction process, based on the formation elastic modulus and Poisson's ratio obtained previously, the three-dimensional numerical model is substituted to perform numerical calculations on the shield construction process for different schemes in the second orthogonal test table, and the surface settlement, left-line segment settlement, right-line segment settlement and building displacement generated during the simulation process are obtained to obtain the second training sample for determining the construction parameters. The first 6 groups of data in the second training sample are used as learning samples, and the last 2 groups of data are used as prediction test samples. The specific algorithm parameters are set, and the cuckoo algorithm is used to select the optimal parameters of the second regression prediction model. The maximum number of iterations of the cuckoo algorithm is set to 2000, and the probability of the bird's nest being discovered P is set to 0. a =0.25, the number of bird nests is 50, the upper and lower limits of the regularization parameter γ of the interval fitting degree of the support vector machine are set to [0,10]; the bandwidth σ of the kernel function of the support vector machine is set 2 The upper and lower limits are [0,1]. The parameters of the second regression prediction model obtained by the cuckoo algorithm have the best values ​​of γ=1.23, σ 2 =0.15. Set γ=1.23, σ 2 =0.15 was substituted into the second regression prediction model to establish the optimized second regression prediction model. The deformation data of the predicted test samples of the last two groups were compared with the results of the second regression prediction model. It was found that the maximum prediction error of the deformation data was 4.56%, and the prediction error was relatively small, indicating that the established optimized second regression prediction model has higher accuracy.

[0184] The construction parameters considered for shield construction optimization are soil bin pressure, face construction spacing, and wall back grouting range. In this embodiment, the length of the interval tunnel is taken as 30m before and after the building influence range plus the length of the building along the shield direction of 53.41m, that is, L'=83.41m; k 1 =200 yuan / m, m 1 =1.2, k 2 = 1500 yuan, k 3＝ 150 yuan / cubic meter, m 3 = 10m. The control values ​​of ground settlement, left line segment settlement, right line segment settlement and building displacement are set as 25mm, 10mm, 10mm and 5mm respectively according to the specification. The second regression prediction model after optimization is established as above, and the construction parameters (Bird's Nest position) are used to calculate the response deformation, and the objective function is used As a fitness function to measure the quality of the bird's nest, the output soil bin pressure is 0.54MPa, the construction spacing of the face is 6 rings behind the right hole, and the grouting range behind the wall is 34cm. The construction cost and deformation caused by the output construction parameters are compared with the original plan before construction, as shown in Table 7. The obtained construction parameters are used to guide the actual construction on site, which ensures the safety of construction and reduces the construction cost compared with the original plan. The construction parameters through the prediction model are more economical and reliable.

[0185] Table 7:

[0186]

[0187] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A shield construction control method for a fractured rock layer near a building, characterized in that: The specific steps include: S1: Obtain monitoring data of several shield areas during shield construction; S2: Determine the range of formation parameters in the shield area according to the geological survey report, so as to construct the first orthogonal test table characterizing the formation mechanical parameters; S3: constructing a three-dimensional numerical model simulating a shield tunnel passing through a building, and performing a numerical simulation of the shield tunnel construction process based on the three-dimensional numerical model and a first orthogonal test table to obtain a first training sample; S4: establishing a first regression prediction model and training it based on the first training sample, optimizing the parameters of the first regression prediction model during the training process to obtain an optimized first regression prediction model; S5: using an intelligent optimization algorithm, and obtaining specific formation parameter values ​​based on the optimized first regression model and monitoring data inversion; S6: Select several construction parameters in the shield construction process to construct a second orthogonal test table representing the values ​​of the construction parameters; S7: performing numerical simulation of the shield construction process based on the specific formation parameter values, the three-dimensional numerical model and the second orthogonal test table to obtain a second training sample; S8: establishing a second regression prediction model and training it based on the second training sample, optimizing the parameters of the second regression prediction model during the training process, and obtaining an optimized second regression prediction model; S9: Adopting intelligent optimization algorithm and taking the construction cost of optimizing construction parameters as the optimization target, optimizing construction parameters is performed based on the optimized second regression prediction model, and based on the set deformation control value of the shield area, the corresponding construction parameter value when the construction cost is the lowest is obtained, and shield construction is performed.

2. The shield construction control method for fractured rock strata adjacent to buildings according to claim 1 is characterized in that: In S4, a first regression prediction model is established and trained based on the first training sample. During the training process, the parameters of the first regression prediction model are optimized. The process of obtaining the optimized first regression prediction model is as follows: S41: establishing a least square vector machine model, i.e., the first regression prediction model, based on the support vector machine (SVM) standard algorithm; S42: Training the first regression prediction model based on the first training sample, and optimizing the model parameters in the first regression prediction model based on the cuckoo intelligent optimization algorithm during the training process, so as to obtain an optimized first regression prediction model.

3. The shield construction control method for fractured rock strata adjacent to buildings according to claim 2 is characterized in that: In S41, the process of establishing the least squares vector machine model based on the support vector machine SVM standard algorithm is: The calculation expression of the support vector machine SVM standard algorithm is set as: Where ω=[ω1,ω2,...,ω n ] T is the weight coefficient vector, is the mapping function, b is the offset; n is the data dimension; The inequality constraint optimization problem in the standard support vector machine (SVM) algorithm is transformed into an equality constraint optimization problem through the risk minimization principle. The transformed equality constraint formula is: The Lagrange equation is used to transform the equality constraint problem of formula (2) into an unconstrained optimization problem. The formula is: Solving the partial derivatives of formula (3) yields: Finally, the unconstrained optimization problem is transformed into the Lagrangian operator α k The problem of solving the linear equations with the bias b is expressed as: In the formula, y=[y1,y2,...,y N ] is the output data of the training sample: L is an n-dimensional array, L = [l1,l2,...,l N ],α=[α1,α2,...,α N ]; k,l=1,2,...,N, I is a unit vector; γ is a regularization parameter representing the degree of interval fitting; By solving formula (5), we can get α k and b, the least squares vector machine model, i.e. the first regression prediction model, is obtained, which is expressed as: In the formula, K(x,x k ) uses a Gaussian kernel function, and K(x,x k )=exp{-xx k 2 / σ 2 },σ 2 represents the bandwidth of the support vector machine kernel function; x k is the input sample, and x k ∈R m .

4. The shield construction control method for fractured rock strata adjacent to buildings according to claim 3 is characterized in that: In S42, the first regression prediction model is trained based on the first training sample, and the process of optimizing the model parameters in the first regression prediction model based on the cuckoo intelligent optimization algorithm during the training process is as follows: S421: Determine the regularization parameter γ and square bandwidth σ corresponding to the bird's nest in the cuckoo intelligent optimization algorithm 2 The value range of S422: Randomly generate the initial position of the bird's nest Represent the initial parameters γ and σ of the support vector machine respectively. 2 , i is the sequence number of the bird's nest, i=1,2,...,S, S is the total number of bird's nests; S423: Calculate the degree of fit corresponding to each bird's nest position and find the current optimal bird's nest position And the optimal fit F, the optimal fit F is the smallest fit value, the optimal nest position is the nest position corresponding to the optimal fitness F, and the fitness F is expressed as: In the formula, y r is the deformation value in the first training sample, y r ' is the deformation value predicted by the first regression prediction model; q is the number of samples; S424: Keep the best bird's nest position of the previous generation And update the location of the bird's nest to achieve optimized exploration of the search space. The update formula is: In the formula, k i (t+1) represents the location of the bird's nest at the t+1th iteration, k i (t) represents the position of the bird's nest at the tth iteration; α is the step scale factor, and α>0; ⊕ is point-to-point multiplication; L(λ) represents the random search path, i is the nest number, and S is the total number of nests; S425: Find the optimal nest location after update And calculate the corresponding fitness F'. If the fitness F' corresponding to the updated optimal nest position is less than the fitness F corresponding to the previous optimal nest position, then replace the previous optimal nest position with the updated optimal nest position, otherwise retain the previous optimal nest position. S426: Determine whether the current number of iterations reaches the maximum number of iterations. If so, output γ, σ 2 The value of is assigned to the first regression prediction model, thereby obtaining the optimized first regression prediction model; otherwise, executing S427; S427: Update the position of the bird's nest again according to the rule of S424, and then operate according to the rules of S425-S426.

5. The shield construction control method for fractured rock strata adjacent to buildings according to claim 4 is characterized in that: In S5, the process of using the intelligent optimization algorithm and inverting the optimized first regression model and monitoring data to obtain specific formation parameter values ​​is as follows: S51: Set the initial parameters of the cuckoo optimization algorithm and randomly generate the initial position of the bird's nest are the initial positions of M bird nests randomly generated within a given range based on various formation parameters, and n is the number of formation parameters; S52: Calculate the fitness value corresponding to each bird's nest and find the current optimal bird's nest location And the optimal fit F", calculated as: In the formula, y l is the monitoring data, y l ' is the deformation value predicted by the optimized first regression model; M is the number of monitoring points; S53: retains the best bird's nest position of the previous generation And update the location of the bird's nest according to the rule of S424; S54: Find the updated optimal bird's nest position and calculate the corresponding fitness. If the fitness corresponding to the updated optimal bird's nest position is less than the fitness F' corresponding to the previous optimal bird's nest position, replace the previous optimal bird's nest position with the updated optimal bird's nest position; otherwise, retain the previous optimal bird's nest position. S55: Determine whether the current iteration number reaches the maximum iteration number, if so, obtain the specific formation parameter value, otherwise, execute S56; S56: Update the position of the bird's nest again according to the rule of S53, and then operate according to the rules of S54-S55.

6. The shield construction control method for fractured rock strata adjacent to buildings according to claim 5 is characterized in that: In S9, the intelligent optimization algorithm is used and the construction cost of the optimized construction parameters is taken as the optimization target. The construction parameters are optimized based on the optimized second regression prediction model. The process of obtaining the corresponding construction parameter value when the construction cost is the lowest is as follows: S91: Set the initial parameters of the cuckoo optimization algorithm and randomly generate the initial position k of the bird's nest i ” (0) , are the initial positions of P bird nests randomly generated within a given range based on each construction parameter, and v is the number of construction parameters; S92: Input the construction parameters into the optimized second regression prediction model to calculate the response deformation, and express the cost generated by the construction according to the construction parameters as follows: Where L' is the length of the interval tunnel, is the soil bin pressure, k1 is the basic cost per meter, m1 is the pressure adjustment coefficient, is the construction spacing of the tunnel face, k2 is the adjustment cost of the shield spacing, k3 is the cost of grouting per cubic meter, is the grouting range behind the wall, m3 is the arc length of the grouting layer; At the same time, the lowest construction cost is used as the fitness function to find the current optimal bird's nest location and the optimal fit. The fitness function is expressed as: In the formula, x1,x2,...,x v is the construction parameter value; y1, y2, y3, y4 represent the calculated surface settlement, left line segment settlement, right line segment settlement and building displacement respectively; y 1max ,y 2max ,y 3max ,y 4max They represent the control value of ground settlement, the control value of left-line segment settlement, the control value of right-line segment settlement and the control value of building displacement respectively; S93: retain the optimal bird's nest position of the previous generation, and update the position of the bird's nest according to the rule of S424; S94: Find the updated optimal bird's nest position and calculate the corresponding fitness. If the fitness corresponding to the updated optimal bird's nest position is less than the fitness corresponding to the previous optimal bird's nest position, replace the previous optimal bird's nest position with the updated optimal bird's nest position; otherwise, retain the previous optimal bird's nest position. S95: Determine whether the current number of iterations reaches the maximum number of iterations. If so, obtain the construction parameter value corresponding to the lowest construction cost. Otherwise, execute S96. S96: Update the position of the bird's nest again according to the rules of S93, and then operate according to the rules of S94-S95.

Images