A method for constructing a structural elastic wave propagation characteristics model based on spectral unit parallel computing
Through the three-dimensional time domain spectral unit method and parallel computing technology, the problem of elastic wave propagation complexity in orthogonal opposite-sex steel bridge decks is solved, efficient structural damage identification and health monitoring is achieved, and an accurate elastic wave propagation characteristic model is provided.
Patent Information
- Application Number
- CN202310297406.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-03-24
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2043-03-24
AI Technical Summary
The geometric structure of orthogonal opposite-sex steel bridge deck panels is complex, resulting in uncertain propagation paths of elastic waves, reflection, diffraction and dispersion phenomena, and accurate structural damage recognition cannot be achieved.
The three-dimensional time domain spectral unit method based on Legendre orthogonal polynomial is used to discrete civil engineering structures, combine parallel calculations to simplify the solution of dynamic equations, construct the elastic wave propagation characteristic model of the structure, and analyze the propagation path and wave field distribution of elastic waves.
It improves computing efficiency and accuracy, reduces the demand for computing resources, can effectively simulate the propagation process of high-frequency acoustic-emitting elastic waves in large structures, and provides accurate structural damage positioning and health monitoring support.
Smart Images

Figure CN116227001B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a method for constructing a structural elastic wave propagation characteristic model based on spectral unit parallel calculation, and belongs to the technical field of civil engineering structure monitoring and simulation. Background Art
[0002] In recent years, nondestructive testing technology based on elastic waves has been widely studied in the field of civil engineering structural health monitoring, and is suitable for accurate damage identification of key structural components. However, the propagation path of elastic waves in actual structures is very complex, with multiple reflections, diffraction and dispersion phenomena, and even coupling with existing damage. In order to achieve accurate damage identification, it is necessary to have a deep understanding of the propagation behavior of elastic waves in various complex structures and their interaction mechanism with damage. Therefore, it is crucial to effectively simulate the propagation characteristics of elastic waves in structures and construct elastic wave propagation characteristic models to serve the location of structural damage based on elastic waves, which is crucial to the development of complex structural health monitoring methods based on elastic waves.
[0003] Numerical simulation of elastic wave propagation can be performed using methods such as the finite element method, the finite difference method, and the boundary element method. Although the finite element method is simple in solving elastic waves, due to the high frequency and short wavelength of elastic waves, the size of the finite element grid needs to be no larger than one tenth of the minimum wavelength to ensure the calculation accuracy, which greatly reduces the calculation efficiency. The calculation resources consumed for simulating slightly larger structures will be so large that it is unrealistic. The finite difference method is a simple and effective numerical solution, but it cannot guarantee the solution accuracy and precision when the material structure is complex. The boundary element method is easier to handle structures with more complex boundaries, but it is more effective when calculating the divergence of elastic waves in infinite media.
[0004] In civil engineering bridge structures, orthotropic steel bridge decks have become a commonly used bridge deck structure for large-span bridges at home and abroad due to their advantages such as good integrity, large bearing capacity, light weight, fast construction, and beautiful structure. However, they are very prone to fatigue cracks, which poses a great safety hazard to the operation and maintenance of bridge structures. Orthotropic steel bridge decks are composed of top plates, U-shaped longitudinal ribs and transverse ribs, etc., and have a complex spatial structure. Effectively simulating the propagation characteristics of elastic waves in orthotropic steel bridge decks and constructing elastic wave propagation characteristic models to serve structural damage location is crucial to understanding the propagation path and wave field distribution of elastic waves and developing complex structural health monitoring methods based on acoustic emission elastic waves. Summary of the invention
[0005] Technical issues: The geometric structure of orthotropic steel bridge decks is extremely complex, with varying thicknesses and three-dimensional intersections. It is difficult for elastic waves to maintain a single propagation path, and reflection and diffraction are likely to occur between the plates. In addition, elastic waves have obvious frequency dispersion in plate structures, and different frequency components have different propagation speeds and energy dissipation degrees. These reasons make it impossible for nondestructive testing technology based on acoustic emission elastic waves to provide accurate damage identification results for orthotropic steel bridge decks.
[0006] The spectral element method adopts Gauss-Lobatto-Legendre polynomial node distribution, and has the flexibility of finite element method in dealing with boundaries and the fast convergence of spectral method. It is a fast, efficient and high-precision high-order finite element method. It overcomes the accuracy requirement that the finite element grid size needs to be less than one-tenth of the minimum wavelength, greatly reducing the demand for computing resources. It is capable of simulating the propagation process of ultra-high frequency acoustic emission elastic waves in large structures. At present, elastic wave simulation based on spectral element method is mainly concentrated in flat plate structures, and the numerical simulation research on three-dimensional space structures with complex geometric structures needs to be strengthened. The present invention aims to use three-dimensional time-domain spectral element method to simulate the propagation process of high-frequency acoustic emission elastic waves in orthotropic steel bridge decks, and accelerate the solution speed through parallel computing, analyze the elastic wave propagation characteristics, and construct the elastic wave propagation characteristics model of the structure to serve the structural damage location based on elastic waves. It is of great significance to understand the propagation path and wave field distribution of elastic waves and to assist in improving the engineering application potential of acoustic emission technology.
[0007] Technical solution: In order to solve the above technical problems, the present invention proposes a method for constructing a structural elastic wave propagation characteristic model based on parallel calculation of spectrum units, which method includes the following steps:
[0008] S1, discretizing a civil engineering structure or a civil engineering structure member into a finite number of elements using a three-dimensional time-domain spectral element based on Legendre orthogonal polynomials;
[0009] S2, using elastic dynamics control equations to establish the motion equations of a finite number of three-dimensional time-domain spectrum units after discretization, calculate the unit mass matrix, unit stiffness matrix and unit damping matrix, store the unit mass matrix and unit damping matrix in a strictly diagonal form in vector form and participate in subsequent calculations, assemble the unit mass matrix, unit stiffness matrix and unit damping matrix of a finite number of three-dimensional time-domain spectrum units to obtain the control equations of the system;
[0010] S3, using the central difference method and the element-by-element method, simplifies the process of assembling the unit stiffness matrix into the overall stiffness matrix and the operation of the large-scale overall stiffness matrix into the calculation and vector assembly of equivalent nodal forces within a finite number of three-dimensional time-domain spectrum units, so that the solution of the system's control equations is simplified to vector operations to implement GPU parallel acceleration calculation to solve the system's dynamic response;
[0011] S4, allocating different CPU computing cores to the working conditions of different excitation positions of the civil engineering structure or civil engineering structural component in step S1, performing elastic wave propagation simulation of the working conditions of different excitation positions through CPU parallel computing, and obtaining elastic wave response and structural wave field;
[0012] S5, calculating the arrival time of the elastic wave at the response point, and constructing a time difference diagram model of the elastic wave propagation characteristics of the civil engineering structure or the civil engineering structure component in step S1.
[0013] Furthermore, in step S1, the specific steps of discretizing the civil engineering structure or civil engineering structural member into a finite number of units using a three-dimensional time-domain spectrum unit based on Legendre orthogonal polynomials are as follows:
[0014] S11, dividing the overall spatial structure into a number of hexahedral parts according to the geometric characteristics and dimensions of the civil engineering structure or the civil engineering structural member;
[0015] S12, establishing spatial models in the same three-dimensional coordinate system according to the geometric dimensions of the plurality of hexahedral parts, discretizing the plurality of hexahedral parts using three-dimensional time-domain spectrum units, and sequentially numbering all three-dimensional time-domain spectrum units and nodes generated by the plurality of hexahedral parts;
[0016] S13, modifying the same nodes at the connection points of the geometric elements between the several hexahedral parts to the same number to assemble into an overall spatial model of the civil engineering structure or the civil engineering structural component, and ensuring that the numbers of all the three-dimensional time domain spectrum units and nodes after the assembly are continuous and non-repetitive;
[0017] S14, all three-dimensional time-domain spectrum units are based on Legendre orthogonal polynomials. Any shape of three-dimensional time-domain spectrum units in the global coordinate system can be obtained by coordinate mapping of standard units in the local coordinate system (ξ, η, γ). The coordinates of each main direction interpolation node in the standard domain are determined by the following formula:
[0018]
[0019] Where P n ′(ξ),P n ′(η) and P n ′(γ) are the first-order reciprocals of the n-order Legendre polynomials in the three main directions in the local coordinate system, where the recursive formula P(x) of the Legendre polynomial is:
[0020]
[0021] The coordinates of all nodes in the three-dimensional time-domain spectrum unit are determined by the above formula, and the displacement field inside the three-dimensional time-domain spectrum unit is obtained by interpolating the node displacement and the shape function.
[0022] Furthermore, the steps in step S2 are as follows:
[0023] S21, take a microelement from the civil engineering structure or civil engineering structural component for force analysis. According to the D'Alembert principle, the equilibrium equation of the microelement is:
[0024]
[0025] The geometric equation is:
[0026]
[0027] The physical equation is:
[0028] σ=Dε
[0029] Where f is the body force vector, ρ is the mass density, u is the displacement vector, t is the time, ε is the strain vector, σ is the stress vector, and D is the elastic matrix. for The transpose of The statement is as follows:
[0030]
[0031] Without considering the body force f, the wave equation for the three-dimensional problem is:
[0032]
[0033] S22, the unit mass matrix M of the three-dimensional time-domain spectrum unit e and the element stiffness matrix K e It can be expressed as:
[0034]
[0035]
[0036] In the formula, ω i ,ω j and ω k denote the Gauss-Lobatto-Legendre integral weight factors in the ξ, η, and γ directions, respectively. The integral weight factors are greater than 0, and the unit mass matrix is in a strictly diagonal form. (ξ i ,η j ,γ k ) represents the coordinates of the integration points, n represents the number of integration points, N is the shape function matrix, B is the geometry matrix, J eThe Jacobian matrix representing the mapping between global coordinates (x, y, z) and local coordinates (ξ, η, γ), the unit damping matrix C e The mass damping proportional coefficient η and the unit mass matrix are defined as follows:
[0037] C e =ηM e
[0038] The unit mass matrix and unit damping matrix in strict diagonal form are stored in vector form and participate in subsequent calculations;
[0039] S23, assembling the mass matrix, stiffness matrix and damping matrix of each unit can obtain the global mass matrix, stiffness matrix and damping matrix. According to the finite element theory, the system motion equation is:
[0040]
[0041] In the formula, M, C, K, F, q, and They are the overall mass matrix, overall damping matrix, overall stiffness matrix, equivalent nodal force vector, nodal displacement vector, nodal velocity vector and nodal acceleration vector of the structure respectively.
[0042] In step S2, the unit mass matrix M of the three-dimensional spectrum unit is formed e and the element damping matrix C e It is a strict diagonal form. Storing it in vector form and participating in subsequent calculations can reduce the consumption of computing resources and effectively improve computing efficiency.
[0043] Furthermore, in step S3, the method for solving the system motion equation obtained based on the assembly is as follows:
[0044] Using the explicit dynamics calculation method, the velocity at time t and the acceleration at time t The displacement q at time t+Δt t+Δt and the displacement q at time t-Δt t-Δt Interpolation yields:
[0045]
[0046]
[0047] Where Δt represents the time step of integration. Under zero initial conditions, the iterative format of the system motion equation in step S23 is:
[0048]
[0049] In the above formula, F tis the equivalent nodal force vector at time t, q t is the node displacement vector at time t. After the unit damping matrix and unit mass matrix are assembled into the overall damping matrix and the overall mass matrix in step S22, we have:
[0050] C=ηM
[0051] Substituting the above formula into the iterative format of the system motion equation, it is expressed as:
[0052]
[0053] in
[0054]
[0055]
[0056] In the above formula, the total mass matrix M is in a strictly diagonal form. In the solution process, the matrix Store and participate in calculations in the form of vectors;
[0057] In the above formula, the overall stiffness matrix K is a sparse large matrix. The element-by-element method is used to implement parallel acceleration calculation on the GPU. The parallel approach is: perform the unit stiffness matrix K on the three-dimensional time domain spectrum unit. e and the element node displacement vector The product operation is performed, and this operation of all m three-dimensional time-domain spectrum units is evenly distributed to each GPU computing core, so that the number of three-dimensional time-domain spectrum units and the number of unit stiffness matrices allocated to each computing core are the same, and the result output is summarized and assembled after all GPU computing core tasks are completed. The idea of GPU parallel acceleration calculation of the element-by-element method is expressed as follows:
[0058]
[0059] In the above formula is the static force vector of the unit equivalent node at time t.
[0060] In step S3, the assembly process of the overall stiffness matrix and large-scale matrix operations are simplified to the calculation of equivalent nodal forces in each unit and vector assembly, and the solution of the system's dynamic equations is simplified to vector operations, which effectively reduces the memory requirements during the iteration process. In addition, the use of parallel processing maximizes the solution speed. Therefore, this method can solve the propagation problem of elastic waves in the structure with extremely low computational cost.
[0061] Furthermore, in step S4, for the load conditions at multiple excitation positions of the civil engineering structure or civil engineering structural component in step S1, the load condition at each excitation position needs to be solved by the three-dimensional time-domain spectral element method, and the central processing unit CPU is used to allocate a CPU computing core to each load condition at the excitation position to perform the three-dimensional time-domain spectral element method solution of the condition, and the elastic wave response and structural wave field under multiple conditions are obtained simultaneously through CPU parallel calculation, and the calculation of all conditions is waited for to be completed, and the results are output and summarized, and the waveform diagrams of each measuring point and the wave field diagrams at each time under the corresponding conditions are drawn to analyze the elastic wave propagation characteristics.
[0062] In step S4, elastic wave propagation simulations are performed at different excitation positions of the structure through parallel computing, and the system dynamic response under the corresponding working conditions is solved at the same time, so as to minimize the total computing time.
[0063] Furthermore, in step S5, the Akaike information criterion is used to calculate the arrival time of elastic waves at each response point of the civil engineering structure or civil engineering structural member in step S1 under load conditions at different excitation positions, and a time difference diagram model of elastic wave propagation characteristics of the structure is constructed for locating and identifying structural damage; the calculation formula of the Akaike information criterion is as follows:
[0064] AIC(t)=tlog 10 (var(R(1,t)))+(Tt-1)log 10 (var(R(t+1,T)))
[0065] Among them, the entire time series T is divided into two time windows: the first half (1, t) and the second half (t+1, T) with maximum likelihood estimation. var(R) represents the variance of R, R(1, t) is the time window from 1 to t, and R(t+1, T) is the time window from t+1 to T. The time t corresponding to the minimum value of the equation is the arrival time of the elastic wave.
[0066] Beneficial effects: Compared with the prior art, the technical solution of the present invention has the following beneficial technical effects:
[0067] (1) The fitting and calculation accuracy of the discrete model is high: the present invention uses a three-dimensional time-domain spectrum unit based on Legendre orthogonal polynomials to discretize and calculate the structure, which can well fit complex spatial shapes. The high-order interpolation nodes are more densely distributed at the unit boundary, which can improve the interpolation accuracy at the unit boundary and effectively suppress the error caused by the Runge phenomenon, thereby achieving accurate high-order displacement field interpolation, ensuring the accuracy of the discretized model, and having both the flexibility of the finite element method in processing boundaries and the rapid convergence of the spectral method.
[0068] (2) Low memory requirement: The present invention overcomes the accuracy requirement that the finite element grid size must be less than one-tenth of the minimum wavelength through high-order displacement field interpolation, thereby reducing the unknown quantities generated after the structure is discretized and greatly reducing the memory requirement; the mass and damping matrices formed by the present invention are in a strictly diagonal form, and the memory requirement is also reduced by storing and participating in calculations in the form of vectors; the present invention adopts the element-by-element method to avoid large-scale matrix operations, which also effectively reduces the memory requirement.
[0069] (3) Fast solution speed: The present invention simplifies the solution of the system's dynamic equations into vector operations and adopts parallel processing to maximize the solution speed. For different excitation position conditions, the elastic wave propagation simulation of the structure at different excitation positions is performed through parallel calculation, and the system dynamic response under the corresponding working conditions is solved at the same time, thereby minimizing the total calculation time.
[0070] (4) Wide scope of application: The present invention is applicable to the simulation of elastic wave propagation in civil engineering structures or civil engineering structural components, and is particularly capable of simulating the propagation process of ultra-high frequency acoustic emission elastic waves in large structures.
[0071] (5) Great application potential: The present invention can simulate the propagation process of high-frequency elastic waves in civil engineering structures or civil engineering structural components with extremely low computational cost to construct an elastic wave propagation characteristic model of the structure, which is crucial for understanding the propagation path and wave field distribution of elastic waves, serving the location of structural damage based on elastic waves, and developing structural health monitoring methods based on acoustic emission elastic waves. BRIEF DESCRIPTION OF THE DRAWINGS
[0072] Figure 1 is a flow chart of the steps of the method of the present invention;
[0073] Figure 2 Orthotropic steel bridge deck model and its spectral element method meshing (unit: mm);
[0074] Figure 3 is the acoustic emission source excitation function;
[0075] Figure 4 is the test area and its excitation point grid (unit: mm);
[0076] Figure 5 The elastic waves received by each sensor and their arrival time under the excitation of node 1: (a) S1, (b) S2, (c) S3, (d) S4;
[0077] Figure 6 Wave field diagram of the model excited by node 1: (a) t = 0.01ms, (b) t = 0.1ms, (c) t = 0.15ms;
[0078] Figure 7Wave field diagram of the model excited by node 2: (a) t = 0.01ms, (b) t = 0.1ms, (c) t = 0.15ms;
[0079] Figure 8 Time difference diagram of the test area (unit: μs): (a) S1-S2, (b) S1-S3, (c) S1-S4, (d) S2-S3, (e) S2-S4, (f) S3-S4. DETAILED DESCRIPTION
[0080] In order to make the purpose and technical solution of the embodiment of the present invention clearer, the technical solution of the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all of the embodiments. Based on the described embodiment 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.
[0081] like Figure 1 As shown, the present invention proposes a method for constructing a structural elastic wave propagation characteristic model based on spectral unit parallel calculation, comprising the following steps:
[0082] S1, discretizing a civil engineering structure or a civil engineering structure member into a finite number of elements using a three-dimensional time-domain spectral element based on Legendre orthogonal polynomials;
[0083] S11, dividing the overall spatial structure into a number of hexahedral parts according to the geometric characteristics and dimensions of the civil engineering structure or the civil engineering structural member;
[0084] S12, establishing spatial models in the same three-dimensional coordinate system according to the geometric dimensions of the plurality of hexahedral parts, discretizing the plurality of hexahedral parts using three-dimensional time-domain spectrum units, and sequentially numbering all three-dimensional time-domain spectrum units and nodes generated by the plurality of hexahedral parts;
[0085] S13, modifying the same nodes at the connection points of the geometric elements between the several hexahedral parts to the same number to assemble into an overall spatial model of the civil engineering structure or the civil engineering structural component, and ensuring that the numbers of all the three-dimensional time domain spectrum units and nodes after the assembly are continuous and non-repetitive;
[0086] S14, all three-dimensional time-domain spectrum units are based on Legendre orthogonal polynomials. Any shape of three-dimensional time-domain spectrum units in the global coordinate system can be obtained by coordinate mapping of standard units in the local coordinate system (ξ, η, γ). The coordinates of each main direction interpolation node in the standard domain are determined by the following formula:
[0087]
[0088] Where P n ′(ξ),P n ′(η) and P n ′(γ) are the first-order reciprocals of the n-order Legendre polynomials in the three main directions in the local coordinate system, where the recursive formula P(x) of the Legendre polynomial is:
[0089]
[0090] The coordinates of all nodes in the three-dimensional time-domain spectrum unit are determined by the above formula, and the displacement field inside the three-dimensional time-domain spectrum unit is obtained by interpolating the node displacement and the shape function.
[0091] S2, using elastic dynamics control equations to establish the motion equations of a finite number of three-dimensional time-domain spectrum units after discretization, calculate the unit mass matrix, unit stiffness matrix and unit damping matrix, store the unit mass matrix and unit damping matrix in a strictly diagonal form in vector form and participate in subsequent calculations, assemble the unit mass matrix, unit stiffness matrix and unit damping matrix of a finite number of three-dimensional time-domain spectrum units to obtain the control equations of the system;
[0092] S21, take a microelement from the civil engineering structure or civil engineering structural component for force analysis. According to the D'Alembert principle, the equilibrium equation of the microelement is:
[0093]
[0094] The geometric equation is:
[0095]
[0096] The physical equation is:
[0097] σ=Dε (5)
[0098] Where f is the body force vector, ρ is the mass density, u is the displacement vector, t is the time, ε is the strain vector, σ is the stress vector, and D is the elastic matrix. for The transpose of The statement is as follows:
[0099]
[0100] Without considering the body force f, the wave equation for the three-dimensional problem is:
[0101]
[0102] S22, the unit mass matrix M of the three-dimensional time-domain spectrum unit e and the element stiffness matrix Ke It can be expressed as:
[0103]
[0104]
[0105] In the formula, ω i ,ω j and ω k denote the Gauss-Lobatto-Legendre integral weight factors in the ξ, η, and γ directions, respectively. The integral weight factors are greater than 0, and the unit mass matrix is in a strictly diagonal form. (ξ i ,η j ,γ k ) represents the coordinates of the integration points, n represents the number of integration points, N is the shape function matrix, B is the geometry matrix, J e The Jacobian matrix representing the mapping between global coordinates (x, y, z) and local coordinates (ξ, η, γ), the unit damping matrix C e The mass damping proportional coefficient η and the unit mass matrix are defined as follows:
[0106] C e =ηM e (10)
[0107] The unit mass matrix and unit damping matrix in strict diagonal form are stored in vector form and participate in subsequent calculations;
[0108] S23, assembling the mass matrix, stiffness matrix and damping matrix of each unit can obtain the global mass matrix, stiffness matrix and damping matrix. According to the finite element theory, the system motion equation is:
[0109]
[0110] In the formula, M, C, K, F, q, and They are the overall mass matrix, overall damping matrix, overall stiffness matrix, equivalent nodal force vector, nodal displacement vector, nodal velocity vector and nodal acceleration vector of the structure respectively.
[0111] In step S2, the unit mass matrix M of the three-dimensional spectrum unit is formed e and the element damping matrix C e It is a strict diagonal form. Storing it in vector form and participating in subsequent calculations can reduce the consumption of computing resources and effectively improve computing efficiency.
[0112] S3, using the central difference method and the element-by-element method, simplifies the process of assembling the unit stiffness matrix into the overall stiffness matrix and the operation of the large-scale overall stiffness matrix into the calculation and vector assembly of equivalent nodal forces within a finite number of three-dimensional time-domain spectrum units, so that the solution of the system's control equations is simplified to vector operations to implement GPU parallel acceleration calculation to solve the system's dynamic response;
[0113] Using the explicit dynamics calculation method, the velocity at time t and the acceleration at time t The displacement q at time t+Δt t+Δt and the displacement q at time t-Δt t-Δt Interpolation yields:
[0114]
[0115]
[0116] Where Δt represents the time step of integration. Under zero initial conditions, the iterative format of the system motion equation in step S23 is:
[0117]
[0118] In the above formula, F t is the equivalent nodal force vector at time t, q t is the node displacement vector at time t. After the unit damping matrix and unit mass matrix are assembled into the overall damping matrix and the overall mass matrix in step S22, we have:
[0119] C=ηM (15)
[0120] Substituting the above formula into the iterative format of the system motion equation, it is expressed as:
[0121]
[0122] in
[0123]
[0124]
[0125] In the above formula, the total mass matrix M is in a strictly diagonal form. In the solution process, the matrix Store and participate in calculations in the form of vectors;
[0126] In the above formula, the overall stiffness matrix K is a sparse large matrix. The element-by-element method is used to implement parallel acceleration calculation on the GPU. The parallel approach is: perform the unit stiffness matrix K on the three-dimensional time domain spectrum unit. e and the element node displacement vector The product operation is performed, and this operation of all m three-dimensional time-domain spectrum units is evenly distributed to each GPU computing core, so that the number of three-dimensional time-domain spectrum units and the number of unit stiffness matrices allocated to each computing core are the same, and the result output is summarized and assembled after all GPU computing core tasks are completed. The idea of GPU parallel acceleration calculation of the element-by-element method is expressed as follows:
[0127]
[0128] In the above formula is the static force vector of the unit equivalent node at time t.
[0129] In step S3, the assembly process of the overall stiffness matrix and large-scale matrix operations are simplified to the calculation of equivalent nodal forces in each unit and vector assembly, and the solution of the system's dynamic equations is simplified to vector operations, which effectively reduces the memory requirements during the iteration process. In addition, the use of parallel processing maximizes the solution speed. Therefore, this method can solve the propagation problem of elastic waves in the structure with extremely low computational cost.
[0130] S4, for the load conditions at multiple excitation positions of the civil engineering structure or civil engineering structural component in step S1, each load condition at each excitation position needs to be solved by the three-dimensional time domain spectral unit method, and a central processing unit CPU is used to allocate a CPU computing core to each load condition at each excitation position to perform the three-dimensional time domain spectral unit method solution of the condition, and the elastic wave response and structural wave field under multiple conditions are simultaneously obtained through CPU parallel calculation, and the calculation of all conditions is waited for to be completed, and the results are output and summarized, and the waveform diagram of each measuring point and the wave field diagram at each time under the corresponding condition are drawn, and the elastic wave propagation characteristics are analyzed.
[0131] In step S4, elastic wave propagation simulations are performed at different excitation positions of the structure through parallel computing, and the system dynamic response under the corresponding working conditions is solved at the same time, so as to minimize the total computing time.
[0132] S5, using the Akaike information criterion to calculate the arrival time of elastic waves at each response point of the civil engineering structure or civil engineering structural member in step S1 under load conditions at different excitation positions, and constructing a time difference diagram model of elastic wave propagation characteristics of the structure for locating and identifying structural damage; the calculation formula of the Akaike information criterion is as follows:
[0133] AIC(t)=tlog 10 (var(R(1,t)))+(Tt-1)log 10 (var(R(t+1,T))) (20)
[0134] Among them, the entire time series T is divided into two time windows: the first half (1, t) and the second half (t+1, T) with maximum likelihood estimation. var(R) represents the variance of R, R(1, t) is the time window from 1 to t, and R(t+1, T) is the time window from t+1 to T. The time t corresponding to the minimum value of the equation is the arrival time of the elastic wave.
[0135] Embodiment 1:
[0136] In order to make the features and advantages of this patent more obvious and easy to understand, the following is a detailed description in conjunction with the embodiments and the accompanying drawings of the specification as follows:
[0137] The main content of this embodiment is to simulate the propagation of acoustic emission elastic waves in orthotropic steel bridge decks using a three-dimensional time-domain spectral unit method and analyze the propagation characteristics, and to construct a time-difference diagram model of the elastic wave propagation characteristics of the structure, in order to verify the effectiveness and practicality of the method for constructing a structural elastic wave propagation characteristic model based on spectral unit parallel calculation proposed in this patent. Consider Figure 2 The orthotropic steel bridge deck structure shown in the figure is welded by a bridge deck and a single longitudinal rib. The size of the bridge deck is 1.5m×1.5m, the plate thickness is 0.02m, the top width of the U-shaped longitudinal rib is 0.3m, the bottom width is 0.2m, the height is 0.25m, and the thickness is 0.008m. The material is Q345D steel, and the elastic modulus E=2.06×10 11 Pa, Poisson's ratio υ=0.28, density ρ=7850kg / m 3 By defining damping to simulate the energy dissipation of elastic waves during propagation, the mass damping proportional coefficient η = 50000 can effectively reduce the impact of reflected waves. The above materials and dimensions are consistent with the steel bridge deck of a suspension bridge.
[0138] According to the geometric characteristics and size of the orthotropic steel bridge deck, the overall spatial structure can be divided into four hexahedral parts: top plate, left web plate, right web plate and bottom plate. The overall numerical model of the orthotropic steel bridge deck is established step by step. The three-dimensional time-domain spectral element method based on Legendre orthogonal polynomials is used to simulate the propagation process of acoustic emission elastic waves in the orthotropic steel bridge deck. The model is discretized into 8400 three-dimensional time-domain spectral elements (such as 20 mm) with a global size of 20 mm. Figure 2 As shown in Figure 2, the number of displacement interpolation nodes for all units in the three directions of x, y, and z axes is 4. Compared with the finite element method, the maximum frequency of the acoustic emission elastic wave of interest is set to f max =200Khz, the finite element mesh size should not be greater than 1.6mm, which means a total of 13863096 units are discretized, which is far beyond the computing power of ordinary computers.
[0139] The acoustic emission source is simulated by applying a transient triangular pulse load in the normal direction of the 3D time-domain spectrum unit node at the excitation location, such as Figure 3As shown. We only focus on the complex ribbed surface acoustic emission elastic wave propagation characteristics, such as Figure 4 As shown in the figure, the central area of the specimen with an area of 0.8m×1.2m is selected as the test area. The unfolded size of the ribbed surface is 1.2m×1.2m. Sensors S1, S2, S3, and S4 are set at the four vertices to receive elastic waves. The test area is divided into grids with a spacing of 0.1m, with a total of 13×13 grid nodes. Excitation is applied to all grid nodes except the four sensors in turn to simulate the acoustic emission source at each point.
[0140] The motion equations of 8400 three-dimensional time-domain spectrum units after discretization are established by using the elastic dynamics control equations, and the unit mass matrix, unit stiffness matrix and unit damping matrix are calculated. The unit mass matrix and unit damping matrix in strict diagonal form are stored in vector form and participate in subsequent calculations. The unit mass matrix, unit stiffness matrix and unit damping matrix of 8400 three-dimensional time-domain spectrum units are assembled to obtain the control equations of the system. The central difference method and the element-by-element method are used to simplify the assembly process of the overall stiffness matrix and the large-scale matrix operation into the calculation of equivalent nodal forces and vector assembly in 8400 three-dimensional time-domain spectrum units, so that the solution of the control equation of the system is simplified to vector operation to implement GPU parallel accelerated calculation to solve the dynamic response of the system. The parallel approach is: perform the product operation of the unit stiffness matrix and the unit node displacement vector on the three-dimensional time-domain spectrum unit, and evenly distribute this operation of all 8400 three-dimensional time-domain spectrum units to each GPU computing core, so that the number of three-dimensional time-domain spectrum units and the number of unit stiffness matrices allocated to each computing core are the same, wait for the completion of all GPU computing core tasks, and summarize and assemble the result outputs.
[0141] right Figure 4 All grid nodes except four sensors are excited, and the load condition at each excitation position needs to be solved by the three-dimensional time-domain spectral element method. The central processing unit CPU is used to allocate a CPU computing core to each load condition at the excitation position to solve the three-dimensional time-domain spectral element method for the condition. The elastic wave response and structural wave field under multiple conditions are obtained through CPU parallel calculation at the same time. After waiting for the completion of the calculation of all conditions, the results are output and summarized to minimize the total calculation time.
[0142] The solution results at the four sensors are extracted and random white noise is superimposed at a signal-to-noise ratio of 25 dB to obtain the acoustic emission elastic waves detected at the four sensors. Figure 4 Taking node No. 1 at the left web of the U-shaped longitudinal rib as an example, it is located between the weld and the chamfer of the U-rib and on the diagonal of the test area. The response result after applying excitation at this node can fully reflect the influence of the U-rib on the propagation of elastic waves. Figure 5The response waveforms detected by the four sensors and the arrival times determined by the Akaike Information Criterion are displayed. It can be seen that the elastic waves show complex dispersion phenomena in the orthogonal anisotropic steel bridge deck. The response amplitude detected by the S1 sensor closest to the excitation source is the largest and peaks at around 0.1ms. The response amplitude detected by the S3 sensor farthest from the excitation source is the smallest and peaks at around 0.18ms, which is consistent with the characteristics of elastic waves attenuating with propagation distance. It can be seen from the figure that the waveform frequency detected by the S1 and S4 sensors is significantly higher than the waveform frequency detected by the S2 and S3 sensors. This is because the path of the elastic wave propagating from node 1 to the S2 and S3 sensors is more complicated, with multiple reflections and diffractions. The resulting dispersion phenomenon makes different frequency components have different energy dissipation degrees, and the high-frequency components decay faster during the propagation process. Further comparison of the elastic waves detected by the S1 and S4 sensors shows that the elastic waves of the S4 sensor show more obvious dispersion phenomena. This is because the propagation paths of the two sensors are similar, but the S4 sensor is farther away from the excitation source. The high-frequency components have a faster propagation speed and gradually distance themselves from the low-frequency components as the propagation distance increases. In addition, observations Figure 5 The time for the elastic wave to reach the four sensors is S1, S4, S2, and S3, which is consistent with the propagation distance. It should be noted that although the plane expansion distance between the excitation source and the S2 sensor is the same as that of the S4 sensor, the propagation path of the former is more complicated, so the elastic wave arrival time of the S2 sensor is slightly later than that of the S4 sensor.
[0143] Figure 6 and Figure 7 The wave fields of node 1 on the left U-rib and node 2 on the left plate at three different times of 0.01ms, 0.1ms and 0.15ms were compared, covering representative features such as the elastic wave reaching the weld between the U-rib and the plate, about to reach the edge of the left plate, reaching the edge of the left plate and reflecting inward, and crossing the U-rib to reach the right side of the plate. Figure 6 In (b)-(c), most of the energy of the elastic wave excited by node 1 is concentrated in the U-rib. This is caused by the reflection at the connection between the U-rib and the plate weld and the change in plate thickness, which causes the energy of the elastic wave to decay sharply when it propagates to the part outside the U-rib. Similarly, Figure 7 In (b)-(c), the energy of the elastic wave excited by the two nodes is mostly concentrated in the plate, and energy attenuation and reflection occur when crossing the U-rib. Figure 6 (c) and Figure 7 In (b)-(c), more significant dispersion, reflection and diffraction phenomena can be observed at the left and lower edges of the plate, the U-rib weld, and the U-rib chamfer.
[0144] right Figure 4Excitation is applied to all grid nodes except four sensors in turn. The elastic wave response obtained under each excitation condition is superimposed with 10 random white noises at a signal-to-noise ratio of 25 dB to simulate 10 lead breaking tests in the actual operation environment of the bridge structure. A total of 1650 sets of acoustic emission data are obtained. The Akaike information criterion is used to calculate the arrival time of the acoustic emission elastic waves received by the four sensors, and the arrival time difference of the six sensor pairs is calculated to obtain a (6×1650) arrival time difference matrix. The cubic spline function is used to interpolate the obtained time difference matrix with an interpolation spacing of 20 mm to construct a time difference diagram model of the elastic wave propagation characteristics of the structure, as shown in the following figure. Figure 8 The time difference diagrams of the S1-S2, S1-S3, S1-S4, S2-S3, S2-S4, and S3-S4 sensor pairs shown can be used to locate and identify fatigue crack damage in orthotropic steel bridge decks.
[0145] From the above simulation, it can be found that acoustic emission elastic waves have significant reflection and dispersion phenomena in orthotropic steel bridge decks. This embodiment uses a three-dimensional time-domain spectrum unit method based on Legendre orthogonal polynomials and parallel processing technology to effectively simulate the propagation behavior of acoustic emission elastic waves in orthotropic steel bridge decks to construct a time difference diagram model of elastic wave propagation characteristics of the structure, which helps to reveal the propagation characteristics of elastic waves in complex spatial structures and their interaction with damage, which is crucial to understanding the propagation path and wave field distribution of elastic waves, serving the location of structural damage based on elastic waves, and developing complex structural health monitoring methods based on acoustic emission elastic waves.
[0146] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.
Claims
1. A method for constructing a structural elastic wave propagation characteristic model based on parallel calculation of spectral units. It is characterized in that The following steps are involved: S1, using a three-dimensional time-domain spectrum unit based on Legendre orthogonal polynomials to discretize the civil engineering structure or civil engineering structural member into a finite number of units; the specific steps are as follows: S11, dividing the overall spatial structure into a number of hexahedral parts according to the geometric characteristics and dimensions of the civil engineering structure or the civil engineering structural member; S12, establishing spatial models in the same three-dimensional coordinate system according to the geometric dimensions of the plurality of hexahedral parts, discretizing the plurality of hexahedral parts using three-dimensional time-domain spectrum units, and sequentially numbering all three-dimensional time-domain spectrum units and nodes generated by the plurality of hexahedral parts; S13, modifying the same nodes at the connection points of the geometric elements between the several hexahedral parts to the same number to assemble into an overall spatial model of the civil engineering structure or the civil engineering structural component, and ensuring that all the three-dimensional time domain spectrum units and node numbers after assembly are continuous and non-repetitive; S14, all three-dimensional time-domain spectrum units are based on Legendre orthogonal polynomials. Any shape of three-dimensional time-domain spectrum units in the global coordinate system can be obtained by coordinate mapping of standard units in the local coordinate system (ξ, η, γ). The coordinates of each main direction interpolation node in the standard domain are determined by the following formula: Where P n ′(ξ),P n ′(η) and P n ′(γ) are the first-order reciprocals of the n-order Legendre polynomials in the three main directions in the local coordinate system, where the recursive formula P(x) of the Legendre polynomial is: The coordinates of all nodes in the three-dimensional time-domain spectrum unit are determined by the above formula, and the displacement field inside the three-dimensional time-domain spectrum unit is obtained by interpolating the node displacement and the shape function; S2, using elastic dynamics control equations to establish the motion equations of a finite number of three-dimensional time-domain spectrum units after discretization, calculate the unit mass matrix, unit stiffness matrix and unit damping matrix, store the unit mass matrix and unit damping matrix in a strictly diagonal form in vector form and participate in subsequent calculations, assemble the unit mass matrix, unit stiffness matrix and unit damping matrix of a finite number of three-dimensional time-domain spectrum units to obtain the control equations of the system; the steps are as follows: S21, take a microelement from the civil engineering structure or civil engineering structural component for force analysis. According to the D'Alembert principle, the equilibrium equation of the microelement is: The geometric equation is: The physical equation is: σ=Dε (5) Where f is the body force vector, ρ is the mass density, u is the displacement vector, t is the time, ε is the strain vector, σ is the stress vector, and D is the elastic matrix. for The transpose of The statement is as follows: Without considering the body force f, the wave equation for the three-dimensional problem is: S22, the unit mass matrix M of the three-dimensional time-domain spectrum unit e and the element stiffness matrix K e It can be expressed as: In the formula, ω i ,ω j and ω k denote the Gauss-Lobatto-Legendre integral weight factors in the ξ, η, and γ directions, respectively. The integral weight factors are greater than 0, and the unit mass matrix is in a strictly diagonal form. (ξ i ,η j ,γ k ) represents the coordinates of the integration points, n represents the number of integration points, N is the shape function matrix, B is the geometry matrix, J e The Jacobian matrix representing the mapping between global coordinates (x, y, z) and local coordinates (ξ, η, γ), the unit damping matrix C e The mass damping proportional coefficient η and the unit mass matrix are defined as follows: C e =ηM e (10) The unit mass matrix and unit damping matrix in strict diagonal form are stored in vector form and participate in subsequent calculations; S23, assembling the unit mass matrix, unit stiffness matrix and unit damping matrix can obtain the global mass matrix, Stiffness matrix and damping matrix, according to finite element theory, the system motion equation is: In the formula, M, C, K, F, q, and They are the overall mass matrix, overall damping matrix, Overall stiffness matrix, equivalent nodal force vectors, nodal displacement vectors, nodal velocity vectors, and nodal acceleration vectors; S3, using the central difference method and the element-by-element method, simplifies the process of assembling the unit stiffness matrix into the overall stiffness matrix and the operation of the large-scale overall stiffness matrix into the calculation and vector assembly of equivalent nodal forces within a finite number of three-dimensional time-domain spectrum units, so that the solution of the system's control equations is simplified to vector operations to implement GPU parallel acceleration calculation to solve the system's dynamic response; S4, allocating different CPU computing cores to the working conditions of different excitation positions of the civil engineering structure or civil engineering structural component in step S1, performing elastic wave propagation simulation of the working conditions of different excitation positions through CPU parallel computing, and obtaining elastic wave response and structural wave field; S5, calculating the arrival time of the elastic wave at the response point, and constructing a time difference diagram model of the elastic wave propagation characteristics of the civil engineering structure or the civil engineering structure component in step S1.
2. The method for constructing a structural elastic wave propagation characteristic model based on spectral unit parallel computing according to claim 1, It is characterized in that The solution method of the system motion equation based on the assembly is as follows: Using the explicit dynamics calculation method, the velocity at time t and the acceleration at time t The displacement q at time t+Δt t+Δt and the displacement q at time t-Δt t-Δt Interpolation yields: Where Δt represents the time step of integration. Under zero initial conditions, the iterative format of the system motion equation in step S23 is: In the above formula, F t is the equivalent nodal force vector at time t, q t is the node displacement vector at time t. After the unit damping matrix and unit mass matrix are assembled into the overall damping matrix and the overall mass matrix in step S22, we have: C=ηM (15) Substituting the above formula into the iterative format of the system motion equation, it is expressed as: in In the above formula, the total mass matrix M is in a strictly diagonal form. In the solution process, the matrix Store and participate in calculations in the form of vectors; In the above formula, the overall stiffness matrix K is a sparse large matrix, and the element-by-element method is used to implement parallel acceleration calculation of the graphics processor GPU. The element stiffness matrix K is calculated on the three-dimensional time-domain spectral element. e and the element node displacement vector The product operation is performed, and this operation of all m three-dimensional time-domain spectrum units is evenly distributed to each GPU computing core, so that the number of three-dimensional time-domain spectrum units and the number of unit stiffness matrices allocated to each computing core are the same, and the results are output and assembled after all GPU computing core tasks are completed. The idea of GPU parallel acceleration calculation of the element-by-element method is expressed as follows: In the above formula is the static force vector of the unit equivalent node at time t.
3. The method for constructing a structural elastic wave propagation characteristic model based on spectral unit parallel computing according to claim 1, It is characterized in that In step S5, the Akaike information criterion is used to calculate the arrival time of elastic waves at each response point of the civil engineering structure or civil engineering structural member in step S1 under load conditions at different excitation positions, and a time difference diagram model of elastic wave propagation characteristics of the structure is constructed for locating and identifying structural damage; the calculation formula of the Akaike information criterion is as follows: AIC(t)=tlog 10 (var(R(1,t)))+(T-t-1)log 10 (var(R(t+1,T))) (20) The entire time series T is divided into the first half (1, t) and the second half (t+1, T) with maximum likelihood estimation. Two time windows, var(R) represents the variance of R, R(1,t) is the time window from 1 to t, and R(t+1,T) is the time window from t+1 to T. The time t corresponding to the minimum value of the equation is the arrival time of the elastic wave.
