Shield tunnel irregular load inversion method based on neural network surrogate model

Abstract

The present application relates to a kind of shield tunnel irregular load inversion method based on neural network agent model, belong to shield tunnel load inversion field.The method, discard the assumption of load distribution mode in traditional lining load inversion method, inversion result can reveal the irregular load distribution mode of on-site tunnel, provide technical support for the service safety of tunnel service;Inversion is based on the deformation of lining whole circumference obtained by three-dimensional laser scanning technology, significantly increase the amount of field data, with the increase of n, the dimension of deformation vector can be increased sufficiently, avoid the defect that traditional inversion method easily leads to the non-uniqueness of inversion result.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of shield tunnel load inversion, and particularly relates to a shield tunnel irregular load inversion method based on a neural network proxy model. BACKGROUND

[0002] The size and distribution of the load of a shield tunnel are important factors affecting the structural service life safety of the tunnel. To obtain the size and distribution of the load of a shield tunnel in service, a traditional method needs to arrange water and soil pressure sensors behind the lining. Although the water and soil pressure sensors can directly obtain the load acting on the tunnel lining, the arrangement is complex, the survival rate of the sensors is low, and the sensors are limited by cost and are difficult to be applied to practical engineering on a large scale.

[0003] Compared with directly obtaining the lining load, it is lower in cost and more mature in technology to obtain the internal force and deformation of the tunnel lining. Based on the detected or monitored internal force and deformation of the lining, the size and distribution of the load acting on the lining can be inverted in combination with a shield tunnel mechanical model.

[0004] Most existing lining load inversion methods are based on a load distribution mode assumed in the design of a tunnel, and the load size is obtained by globally optimizing an intelligent algorithm in combination with a mechanical forward model. However, this method depends on the load distribution mode in the design period and has a large gap with the actual load distribution mode of the project. If the assumed load distribution mode is abandoned for inversion, there may be a problem of non-unique inversion load. At this time, various regularization techniques or prior distributions need to be introduced, which is subjective. Therefore, there is no inversion method for the lining load that is low in cost, high in feasibility in the engineering site, and has a solid theoretical basis.

[0005] Since the load distribution mode of an actual tunnel project cannot be known, the assumed load distribution mode must be abandoned. The problem of non-unique inversion load caused thereby can be avoided by increasing the measured data on site. The three-dimensional laser scanning technology can obtain the deformation of the shield tunnel in the whole circumference by scanning the tunnel contour line, greatly enriching the measured data of the tunnel deformation. The deformation of the tunnel in the whole circumference obtained by the three-dimensional laser scanning can improve the inversion accuracy of the lining load of the tunnel. SUMMARY

[0006] The purpose of the present application is to overcome the problems existing in the prior art and provide a shield tunnel irregular load inversion method based on a neural network proxy model.

[0007] To achieve the above purpose, the technical scheme of the present application is as follows: a shield tunnel irregular load inversion method based on a neural network proxy model, comprising:

[0008] Step S1, according to the geometric size, joint structure and material mechanics parameters of the actual tunnel engineering, a refined shield tunnel lining three-dimensional finite element model is established;

[0009] Step S2, the lining is divided into n equal parts according to the circumference, forming n load inversion target points, forming an N-dimensional load inversion vector P=(P1, P2, …, P n ); m load vectors P are randomly generated based on the Monte Carlo method;

[0010] Step S3, the load determined in step S2 is applied to the lining, and the finite element model established in step S1 is input, the lining deformation corresponding to each load vector P is calculated; the lining deformation is divided into 2n equal parts according to the circumference, the lining deformation of each equal part is obtained, and m 2n-dimensional deformation vectors D are formed;

[0011] Step S4, a neural network surrogate model is selected, the deformation vector D is taken as the model input, the load vector P is taken as the model output, and the m samples obtained in step S3 are used to train and verify the neural network surrogate model;

[0012] Step S5, the three-dimensional laser scanning technology is used to obtain the full circumference deformation of the actual tunnel; according to the division method in step S3, the lining deformation vector d of the engineering site is determined;

[0013] Step S6, using the neural network surrogate model trained in step S4, based on the field lining deformation vector, the load vector p acting on the lining is calculated.

[0014] In an embodiment of the present application, in step S1, the characteristics of the tunnel lining three-dimensional finite element numerical model are: a geometric model is established according to the diameter, lining thickness and lining block condition of the actual tunnel engineering; the joints of different block segments are connected by bolts; the segments are connected by friction contact; the constitutive model of the concrete lining is a three-fold line elastic-plastic model, and the mechanical parameters are selected according to the concrete grade; the joint bolt is a double-fold line elastic-plastic model, and the mechanical parameters are selected according to the bolt type.

[0015] In an embodiment of the present application, in step S2, the load between the inversion target points is obtained by using a cubic spline interpolation method.

[0016] In an embodiment of the present application, in step S2, the load between the inversion target points is obtained as follows:

[0017] For the i-th load inversion target point, the polar coordinate is θ i , and the load P i acting on it forms a data point (θ i , P i ); for adjacent data points, a cubic polynomial y=ax 3+ bx 2 + cx + d are interpolated; according to continuity requirement, each piece of cubic function needs to satisfy:

[0018]

[0019]

[0020]

[0021]

[0022] …

[0023]

[0024]

[0025] Get 2n equations; then according to the smoothness requirement, the first derivative of the piecewise function is continuous at the connection point to get n equations:

[0026]

[0027]

[0028] …

[0029]

[0030]

[0031] Again according to the second derivative is continuous to get n equations:

[0032] 6a1θ2+2b1=6a2θ2+2b2

[0033] 6a2θ3+2b2=6a3θ3+2b3

[0034] …

[0035] 6a n-1 θ n +2b n-1 =6a n θ n +2b n

[0036] 6a n θ1+2b n =6a1θ1+2b1

[0037] Solve the above 4n equations to get n piecewise function of 4n unknown coefficients a1, b1, c1, d1, …, a n , b n , cn , d n .

[0038] In an embodiment of the present application, in step S3, the load at any position of the lining is calculated according to the cubic spline interpolation result, and the corresponding load is applied in the finite element numerical model, and the finite element model is iterated to convergence to obtain the whole circumference deformation of the tunnel lining inner contour under the given load, the lining ring is divided into 2n equal parts to obtain a 2n-dimensional deformation vector D; the finite element model calculation is repeated m times to obtain m deformation vectors D.

[0039] In an embodiment of the present application, in step S4, a BP neural network proxy model is constructed with the 2n-dimensional deformation vector D as input and the n-dimensional load vector P as output, the BP neural network proxy model is trained with the m samples obtained in step S3 to obtain a neural network proxy model P=f(D) for predicting the load vector P from the lining deformation vector D.

[0040] In an embodiment of the present application, in step S5, the three-dimensional laser scanning technology is used to obtain the contour point cloud data of the site tunnel, and the inner contour of the current state of the lining is obtained after processing; the whole circumference deformation of the tunnel under the current state is obtained by comparing with the inner contour data of the tunnel just after construction; and the deformation vector d of the site tunnel is obtained according to the division position in step S3.

[0041] In an embodiment of the present application, in step S6, the neural network proxy model P=f(D) trained in step S4 is input with the site deformation vector d to obtain the load vector p=f(d) of the site tunnel.

[0042] The present application also provides a shield tunnel irregular load inversion system based on a neural network proxy model, characterized by comprising a memory, a processor, and computer program instructions stored on the memory and capable of being executed by the processor, when the processor executes the computer program instructions, the method steps as described above can be realized.

[0043] The present application also provides a computer readable storage medium having computer program instructions capable of being executed by a processor stored thereon, when the processor executes the computer program instructions, the method steps as described above can be realized.

[0044] Compared with the prior art, the present application has the following beneficial effects: the present application discards the assumption of the load distribution mode in the traditional lining load inversion method, and the inversion result can reveal the irregular load distribution mode of the site tunnel, providing technical support for the service safety of the tunnel; the inversion is based on the lining whole circumference deformation obtained by the three-dimensional laser scanning technology, which significantly increases the amount of site data, and as n increases, the dimension of the deformation vector can be increased sufficiently, avoiding the defect that the traditional inversion method easily leads to non-unique inversion result. Attached Figure Description

[0045] Figure 1 This is a flowchart of the lining load inversion algorithm of the present invention;

[0046] Figure 2 This is a schematic diagram of the cross-section of a subway shield tunnel;

[0047] Figure 3 A refined three-dimensional finite element numerical model for tunnel lining;

[0048] Figure 4 The constitutive relations of the segments and bolts in the finite element model;

[0049] Figure 5 For training, testing, and validating BP neural networks;

[0050] Figure 6 The results are from on-site shield tunnel deformation monitoring and load inversion. Detailed Implementation

[0051] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings.

[0052] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0053] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0054] like Figure 1 As shown, this invention provides a method for inverting irregular loads in shield tunnels based on a neural network surrogate model, comprising:

[0055] Step S1: Based on the actual tunnel engineering's geometric dimensions, joint structure, and material mechanical parameters, establish a refined three-dimensional finite element model of the shield tunnel lining;

[0056] Step S2: Divide the lining into n equal parts along the circumference to form n load inversion target points, resulting in an N-dimensional load inversion vector P = (P1, P2, ..., P...). n Based on the Monte Carlo method, m load vectors P are randomly generated.

[0057] Step S3, the load determined in step S2 is applied to the lining, and is input into the finite element model established in step S1 to calculate the whole circumference deformation of the lining corresponding to each load vector P; the lining deformation is equally divided into 2n sections, and the lining deformation of each section is obtained to form m 2n-dimensional deformation vectors D;

[0058] Step S4, a neural network surrogate model is selected, the deformation vector D is taken as the model input, the load vector P is taken as the model output, and the neural network surrogate model is trained and verified by the m samples obtained in step S3;

[0059] Step S5, the three-dimensional laser scanning technology is used to obtain the whole circumference deformation of the actual tunnel; and the lining deformation vector d of the engineering site is determined according to the division method in step S3;

[0060] Step S6, the neural network surrogate model trained in step S4 is used to calculate the load vector p actually acting on the lining based on the field lining deformation vector.

[0061] Taking the current subway shield tunnel as an example, the related geometric dimensions are shown in Figure 2 . The outer diameter of the tunnel is 3.1 m, the lining thickness is 0.35 m, and the ring width is 1.2 m. The lining ring is assembled by 6 segments, including 1 top sealing block (F block), 2 adjacent blocks (L1 block and L2 block), 2 standard blocks (B1 block and B2 block) and 1 bottom sealing block (D block), and the segments are connected by straight bolts.

[0062] Step S1, according to the geometric dimensions provided by the field tunnel, a refined numerical model as shown in Figure 3 is established in the finite element analysis software ABAQU, which simulates the joint bolt and bolt hand hole. In the three-dimensional finite element model, the lining is a concrete material, and the three-fold line strain-strain relationship as shown in Figure 4 (a) is used, and the related mechanical parameters are the elastic modulus E1, the yield stress σ b , the plastic modulus E r , the ultimate stress σ c and the ultimate strain ε d of the concrete. The stress-strain relationship of the bolt is a double-fold line as shown in Figure 4 (b), and the related mechanical parameters are the elastic modulus E0, the yield stress σ s , the plastic modulus E r , the ultimate strain ε c (or the ultimate stress σ c ). The friction contact between the segments and the segments, and between the bolt hole and the bolt is simulated by the penalty function, the friction coefficient between the segments is 0.4, and the friction coefficient between the bolt hole and the bolt is 0.1. The segments and the bolts are both C3D8 solid elements.

[0063] Step S2, take the tunnel vault as θ=0 (0°), rotate counterclockwise as positive, the tunnel arch bottom as θ=π (180°), divide the lining into 36 parts, form a 36-dimensional load inversion vector P=(P1, P1, …, P 36 ) and the corresponding position coordinate vector θ=(0, 18 / π, 9 / π, …, π, …, 35 / 18π). Between adjacent load inversion target points, use a cubic spline interpolation of the form y=ax 3 +bx 2 +cx+d. According to the continuity and smoothness requirements of the spline function, the difference coefficients a1, b1, c1, d1, …, a 36 , b 36 , c 36 , d 36 of each segment function need to meet the following 144 equations,

[0064]

[0065]

[0066]

[0067]

[0068] Solve the above equations to get 144 spline interpolation coefficients.

[0069] Step S3, randomly generate 500 36-dimensional load vectors P=(P1, P1, …, P 36 ), each load component P i obeys the uniform distribution of (0kN, 100kN). According to the results of cubic spline interpolation, the load at any position of the lining is calculated, and the above load is applied in the finite element numerical model. Iterative finite element model to convergence, get the tunnel lining inner contour deformation under the given load. Divide the lining ring into 360 equal parts to get a 360-dimensional deformation vector D. Perform 500 finite element model calculations to obtain 500 deformation vectors D. Here, the horizontal displacement of the inner surface of the lining is selected.

[0070] Step S4, take the 360-dimensional deformation vector D as input, and the 36-dimensional load vector P as output, build a BP neural network surrogate model, train the BP neural network with the 500 samples obtained in step S3 to get a surrogate model P=f(D) that predicts the load vector P from the lining deformation vector D.

[0071] The BP neural network in the above step S4 is configured with 20 hidden layer neurons, a hyperbolic tangent function (Tanh function) as the activation function, a linear function (purelin function) as the output layer activation function, a mean square error (MSE) of the predicted load vector and the sample load vector as the loss function, and a gradient descent algorithm is used for training, and the training target is reached after 178 iterations. The training, testing and verification of the BP neural network are as shown in Figure 5

[0072] In the step S5, the profile point cloud data of the on-site tunnel is obtained by using the three-dimensional laser scanning technology, and the inner profile of the current state of the lining is obtained after processing. By comparing with the inner profile data of the tunnel just after the construction is completed, the tunnel full-circle deformation in the current state is obtained, and the lining horizontal displacement distribution is as shown in Figure 6 (a).

[0073] In the step S6, the proxy model P=f(D) obtained by training in the step S4 is input with the on-site deformation vector d to obtain the load vector p=f(d) of the on-site tunnel, and the tunnel load distribution obtained by inversion is as shown in Figure 6 (b).

[0074] The application further provides a shield tunnel irregular load inversion system based on a neural network proxy model, which is characterized by comprising a memory, a processor and computer program instructions stored in the memory and capable of being executed by the processor, and when the processor executes the computer program instructions, the method steps as described above can be realized.

[0075] The application further provides a computer readable storage medium, which stores computer program instructions capable of being executed by a processor, and when the processor executes the computer program instructions, the method steps as described above can be realized.

[0076] Those skilled in the art should understand that the embodiments of the application can be provided as a method, a system or a computer program product. Therefore, the application can be in the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the application can be in the form of a computer program product implemented on one or more computer usable storage media containing computer usable program codes (including but not limited to disk storage, CD-ROM, optical storage, etc.).

[0077] ​The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flow or blocks Figure 1 one or more flow or blocks

[0078] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flow or blocks Figure 1 one or more flow or blocks

[0079] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flow or blocks Figure 1 one or more flow or blocks

[0080] The above description is only preferred embodiments of the present application, not intended to limit other forms of the application. Any person familiar with the art can make changes or modifications to the above-mentioned technical content of the disclosure as equivalent embodiments. However, any simple modification, equivalent change and modification of the above embodiments without departing from the technical solution of the present application, according to the technical essence of the present application, still belongs to the protection scope of the technical solution of the present application.

Claims

1. A shield tunnel irregular load inversion method based on a neural network proxy model, characterized in that, The method comprises the following steps: Step S1, according to the geometric size, joint structure and material mechanics parameters of the actual tunnel engineering, a refined shield tunnel lining three-dimensional finite element model is established; Step S2: Divide the lining into n equal parts along the circumference to form n load inversion target points, resulting in an N-dimensional load inversion vector P = (P1, P2, ..., P...). n m load vectors P are randomly generated based on the Monte Carlo method; the loads between the target points are obtained using cubic spline interpolation. Step S3, the load determined in step S2 is applied to the lining, and is input into the finite element model established in step S1 to calculate the lining deformation corresponding to each load vector P; the lining deformation is divided into 2n equal parts, and the lining deformation of each equal part is obtained to form m 2n-dimensional deformation vectors D; specifically, the load at any position of the lining is calculated according to the cubic spline interpolation result, and the corresponding load is applied in the finite element numerical model, and the finite element model is iterated to convergence to obtain the whole circumference deformation of the tunnel lining under the given load; the lining ring is divided into 2n equal parts to obtain a 2n-dimensional deformation vector D; the finite element model calculation is repeated m times to obtain m deformation vectors D; Step S4, a neural network surrogate model is selected, the deformation vector D is taken as the input of the model, the load vector P is taken as the output of the model, and the neural network surrogate model is trained and verified by the m samples obtained in step S3; specifically, the 2n-dimensional deformation vector D is taken as the input, and the n-dimensional load vector P is taken as the output to construct a BP neural network surrogate model; the BP neural network surrogate model is trained by the m samples obtained in step S3 to obtain the neural network surrogate model P=f(D) for predicting the load vector P from the lining deformation vector D; Step S5, a three-dimensional laser scanning technology is used to obtain the whole circumference deformation of the actual tunnel; according to the division method in step S3, the lining deformation vector d of the engineering site is determined; specifically, the three-dimensional laser scanning technology is used to obtain the contour point cloud data of the site tunnel, and the inner contour of the current state of the lining is obtained after processing; the whole circumference deformation of the tunnel under the current state is obtained by comparing with the inner contour data of the tunnel just after construction; according to the division position in step S3, the deformation vector d of the site tunnel is obtained; Step S6, the neural network surrogate model trained in step S4 is used to calculate the load vector p actually acting on the lining based on the site lining deformation vector.

2. The shield tunnel irregular load inversion method based on a neural network proxy model according to claim 1, characterized in that, In step S1, the characteristics of the tunnel lining three-dimensional finite element numerical model are as follows: a geometric model is established according to the diameter, lining thickness and lining block condition of the actual tunnel engineering; the joints of different block segments are connected by bolts; the segments are connected by friction contact; the constitutive model of the concrete lining is a three-fold line elastic-plastic model, and the mechanical parameters are selected according to the concrete grade; the joint bolt is a double-fold line elastic-plastic model, and the mechanical parameters are selected according to the bolt type.

3. The shield tunnel irregular load inversion method based on a neural network proxy model according to claim 1, characterized in that, In step S2, the load between the inversion target points is obtained as follows: For the i-th load inversion target point, its polar coordinates are θ i A load P acts on it. i , forming data points (θ) i ,P i For adjacent data points, a cubic polynomial y = ax² is used. 3 +bx 2 Interpolation is performed using +cx+d; according to the continuity requirement, each segment of the cubic function must satisfy: 6a1θ2+2b1=6a2θ2+2b2 6a2θ3+2b2=6a3θ3+2b3 In step S6, according to the neural network surrogate model P=f(D) trained in step S4, the site deformation vector d is input to obtain the load vector p=f(d) of the site tunnel. ​ … 6a n-1 θ n +2b n-1 = 6a n θ n +2b n 6a n θ1+2b n = 6a1θ1+2b1 Solving the above 4n equations simultaneously, we obtain the 4n unknown coefficients a1, b1, c1, d1, …, an, bn, cn, dn of the n piecewise functions. n n n n ​​​​ 4. The shield tunnel irregular load inversion method based on a neural network proxy model according to claim 1, characterized in that, ​ 5. A shield tunnel irregular load inversion system based on a neural network proxy model, characterized in that, A computer program product comprising a memory, a processor and computer program instructions stored on the memory and executable by the processor, whereby execution of the computer program instructions by the processor implements the method steps of any of claims 1-4.

6. A computer readable storage medium having stored thereon computer program instructions executable by a processor, whereby execution of the computer program instructions by the processor implements the method steps of any of claims 1-4.

Images