Welding structure crack propagation life analysis method based on support vector machine model

Abstract

The invention provides a method for analyzing crack propagation life of a welding structure based on a support vector machine model, which comprises the following steps of: discretizing a geometric model of the welding structure by adopting a hexahedron element, generating a finite element model and performing statics simulation analysis; an initial crack is implanted in the center of the weld joint, and a sub-model grid containing the initial crack is re-divided by using a tetrahedron unit; obtaining a stress intensity factor numerical solution of the tip of the crack through finite element calculation under values of different half-crack lengths, crack positions and front edge positions; obtaining an analytical solution by using a stress intensity factor calculation formula, and calculating a difference value between the numerical solution and the analytical solution; taking the half-crack length, the crack position and the front edge position as input, taking the difference value as output, establishing a support vector machine regression model, and obtaining a stress intensity factor correction formula according to a stress intensity factor calculation formula; and substituting the stress intensity factor correction formula into a Paris crack propagation formula to obtain a Paris correction formula, and predicting the fatigue life of the welding structure.

Description

Technical Field

[0001] The invention belongs to the field of aerospace computing, and in particular relates to a method for analyzing crack propagation life of a welded structure based on a support vector machine model. Background Art

[0002] Fatigue is an important factor affecting the service life of engineering components. During the manufacturing process, engineering components will inevitably produce some initial defects, which are very easy to be damaged when facing unstable loads, initiating initial cracks, and causing damage to engineering components over time. Today, the structure of spacecraft tanks has problems such as insufficient operational stability and complex service conditions. It is urgent to carry out research on a structural life assessment technology to improve the level of my country's aerospace transportation equipment design and meet the needs of fatigue life prediction of tank structures in spacecraft during service.

[0003] People started studying fracture mechanics relatively late, and its research object is components with cracks. Fracture mechanics can quantitatively evaluate the fatigue life of components and provide theoretical guidance for engineering problems. Applying the principles of fracture mechanics to the analysis of fatigue crack propagation and estimating the fatigue life of components is an important part of fracture mechanics research.

[0004] The classic formula used for fatigue crack growth is the Paris formula: In recent years, many correction methods have emerged for the Paris formula. Most of the existing corrections to the Paris formula are aimed at the correction of the formula expression, such as introducing the influence of component shape, considering the closure effect, etc. These correction terms are added to the Paris formula in the form of multipliers or the expression of the Paris formula is modified. For example, Forman corrected the Paris formula Under the influence of working conditions, material properties, etc., the expression of fatigue crack growth formula is complex and changeable, with low applicability, and it is difficult to find a general formula that can be widely used in various situations. At the same time, the correction of Paris formula usually requires a large number of physical tests to obtain crack growth data, but physical tests of aerospace components are difficult to carry out, expensive, time-consuming, and accidental. Using physical test data to correct the formula requires a lot of cost and is not realistic. Therefore, how to establish a suitable calculation model and combine the Paris formula to predict the fatigue life of components has important academic and engineering application value. Summary of the invention

[0005] In view of the above problems, an embodiment of the present invention provides a method for analyzing the crack growth life of a welded structure based on a support vector machine model, which is used to improve the accuracy and efficiency of fatigue fracture life calculation of a welded structure.

[0006] An embodiment of the present invention provides a method for analyzing crack growth life of a welded structure based on a support vector machine model, which comprises the following steps:

[0007] Step 1: Use hexahedral units to discretize the geometric model of the welded structure to obtain a finite element model, and perform static simulation analysis to obtain node displacement and stress distribution calculation results;

[0008] Step 2: implanting an initial crack at the weld center of the finite element model, dividing the finite element model into a sub-model containing the initial crack and other partial models, re-discretizing the sub-model using tetrahedral units, and encrypting the crack tip, retaining hexahedral units in the other partial models, retaining the nodes of the contact surface between the sub-model and the other partial models, and using the node displacement of the contact surface as the boundary condition of the sub-model;

[0009] Step 3: Perform crack propagation finite element calculation on the sub-model to obtain the numerical solution K of the stress intensity factor at the crack tip under different values ​​of half crack length a, crack position d and front edge position λ. e ;

[0010] Step 4: Use the stress intensity factor calculation formula Calculate the analytical solution and calculate the numerical solution K e With the analytical solution K t The difference K * , K * =K e -K t ;

[0011] Step 5: Take the half crack length a, the crack position d and the front edge position λ as input, and use the difference K * As the output, establish a support vector machine regression model According to the stress intensity factor calculation formula Get the stress intensity factor correction formula

[0012] Step 6: Substitute the stress intensity factor correction formula K'(a, d, λ) into the Paris crack growth formula Get the Paris correction formula The structural residual strength criterion is selected as the criterion to obtain the fatigue life prediction value of the welded structure.

[0013] The method for analyzing the crack growth life of a welded structure based on a support vector machine model provided by an embodiment of the present invention has the following advantages:

[0014] (1) In the embodiment of the present invention, a static analysis is performed on the overall finite element model of the structure to obtain the calculation results of its node displacement and stress distribution; a key area is selected, the model of the area is cut, and a sub-model and other partial models are obtained; the node displacement at the intersection of the sub-model and the other partial models is extracted and used as boundary conditions for interpolation calculation; the aforementioned boundary conditions are applied to the sub-model, and finally the sub-model is solved. The above process uses a sub-model for local structural analysis. The overall finite element model of the structure is large in size and it is difficult to directly simulate small initial defects. It is often very limited when performing local analysis. The sub-model can realize refined analysis and boundary condition conversion between different units.

[0015] (2) The traditional crack propagation research method uses J-integral to calculate the numerical solution of the crack tip stress intensity factor. Compared with it, the embodiment of the present invention uses the mutual integration method, namely the M-integral method, to solve the stress intensity factor. The use of symmetric grids reduces local discrete errors, and the mutual integration method is independent of the path, which can bypass the crack tip singular area, thereby improving the accuracy of the stress intensity factor calculation, thereby improving the accuracy of the crack propagation life study.

[0016] (3) The embodiment of the present invention modifies the calculation formula of the stress intensity factor at the crack tip of the welded structure. Based on this, a correction term is introduced The final stress intensity factor calculation formula is obtained This correction retains the original physical information of the theoretical formula and improves the accuracy of the stress intensity factor by introducing correction terms related to the half-crack length, crack position and front edge position.

[0017] (4) The embodiment of the present invention uses a support vector machine regression model to perform regression analysis on the stress intensity factor correction term, and establishes the relationship between the three parameters of half crack length, crack position and front edge position and the stress intensity factor correction term K * The mapping relationship And into the stress intensity factor correction formula The support vector machine regression model can guarantee more reliable calculation results under small sample and multi-dimensional parameter conditions. At the same time, it can replace the complex and time-consuming finite element simulation to obtain the solution of the correction term, thereby improving the efficiency of the calculation of the crack propagation life of the welded structure. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A simplified flow chart of a method for analyzing crack growth life of a welded structure based on a support vector machine model according to an embodiment of the present invention;

[0019] Figure 2 A schematic diagram of a three-dimensional flat plate welding structure model according to an embodiment of the present invention;

[0020] Figure 3 It is a schematic diagram of dividing the sub-model of the three-dimensional flat plate welding structure according to an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to make the above-mentioned purposes, features and advantages of the embodiments of the present invention more obvious and understandable, the technical solutions 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 only 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 making creative work belong to the scope of protection of the present invention.

[0022] The embodiment of the present invention takes a three-dimensional plate including a welded structure as an example to specifically illustrate the welded structure crack growth life analysis method in the embodiment of the present invention, in which the welded structure stress intensity factor is corrected. In addition, the welded structure crack growth life analysis method can also be extended to the analysis of other components containing welded structures.

[0023] The embodiment of the present invention provides a method for analyzing the crack growth life of a welded structure based on a support vector machine model, which specifically includes the following steps:

[0024] Step 1: Use hexahedral units to discretize the geometric model of the welded structure to obtain a finite element model, and perform static simulation analysis to obtain the node displacement and stress distribution calculation results.

[0025] In the embodiment of the present invention, the geometric model of the three-dimensional plate with a welding structure (hereinafter referred to as the plate) is as follows: Figure 2 In this geometric model, the plate includes three areas, namely the base material area, the heat affected zone and the weld area. The weld area is a trapezoid with an upper surface width of 4 mm and a lower surface width of 6 mm; the heat affected zone is a parallelepiped with an upper and lower surface width of 8 mm; the rest is the base material area, the overall thickness of the model is 2 mm, the total width is 40 mm, and the total height is 50 mm. The material properties are set as shown in Table 1:

[0026] Table 1. Material parameter settings of flat plate with welded structure

[0027]

[0028] The C3D8R eight-node linear hexahedral element is used to Figure 1The geometric model shown is discretized to obtain a finite element model containing 4000 units. This finite element model is a global model for simulation. A fixed support boundary condition is applied to the bottom surface of the geometric model, and a tensile uniform surface load P = 1MPa is applied to the upper surface of the geometric model. On this basis, static finite element calculation is carried out to obtain the corresponding static analysis results.

[0029] Step 2: implant an initial crack at the center of the weld of the finite element model, divide the finite element model into a sub-model containing the initial crack and other partial models, re-discretize the sub-model using tetrahedral units, and encrypt the crack tip. Retain hexahedral units in other partial models, retain the nodes of the contact surface between the sub-model and other partial models, and use the node displacement of the contact surface as the boundary condition of the sub-model.

[0030] In the embodiment of the present invention, an initial crack is implanted in the weld center (i.e., the center of the weld zone) of the finite element model established in step 1. The initial crack is, for example, an initial through crack, and the overall finite element model is divided into a sub-model containing the initial crack and other partial models. Figure 3 As shown, since the crack tip stress has r -0.5 Singularity, the mesh of the sub-model is re-divided using tetrahedral elements, and the mesh is encrypted for the crack tip area to adapt to the stress field at the crack tip. The other parts of the model still retain hexahedral elements. The nodes (i.e., mesh nodes) of the contact surface between the sub-model and other parts of the model remain unchanged, and the node displacement of the above contact surface calculated by the finite element model in step 1 is used as the boundary condition for the crack extension analysis of the sub-model.

[0031] Step 3: Perform crack propagation finite element calculation on the sub-model to obtain the numerical solution K of the stress intensity factor at the crack tip under different values ​​of half crack length a, crack position d and front edge position λ. e .

[0032] In an embodiment of the present invention, three physical quantities that are strongly correlated with the stress intensity factor, namely, half-crack length a, crack position d, and front edge position λ, are used as key factors affecting the stress intensity factor at the crack tip for calculation. The initial through-crack implanted in the embodiment of the present invention is a middle through-crack, which has two front edges in the positive and negative directions along the normal of the weld. Here, the positive front edge is taken as the research object, that is, the crack front edge close to the load application surface is studied and calculated, and the numerical solution of the stress intensity factor at the crack tip is calculated. Specifically, when the half-crack length a, crack position d, and front edge position λ are respectively {1.0mm, 1.5mm}, {0.0mm, ±5.0mm}, {0, 0.5, 1.0}, the numerical solution K of the stress intensity factor at the crack tip is calculated. e , as shown in Table 2:

[0033] Table 2 Numerical solution of stress intensity factor K at crack tip of plate with welded structure e

[0034]

[0035] Step 4: Use the stress intensity factor calculation formula Calculate the analytical solution and calculate the numerical solution K e And the analytical solution K t The difference K * , K * =K e -K t .

[0036] Specifically, the stress intensity factor calculation formula is Given the half crack length a, crack position d and leading edge position λ in step 3, the corresponding stress intensity factor analytical solution K is calculated. t , In the embodiment of the present invention, the stress intensity factor calculation formula should be The corresponding calculation results are shown in Table 3:

[0037] Table 3 Analytical solution of stress intensity factor K at crack tip of plate with welded structure t

[0038]

[0039] According to the numerical solution of stress intensity factor shown in Table 2, the difference K between the two solutions is calculated. * =K e -K t , the difference is shown in Table 4:

[0040] Table 4 Difference K between numerical solution and analytical solution of stress intensity factor at crack tip of plate with welded structure *

[0041]

[0042] Step 5: Take the half crack length a, crack position d and front edge position λ as input and use the difference K * As the output, establish a support vector machine regression model According to the stress intensity factor calculation formula Get the stress intensity factor correction formula

[0043] Specifically, the three key parameters of half crack length a, crack position d and front edge position λ in step 3 are used as input, and the difference K in step 4 is used as input. *As the output, a support vector machine regression model is established. Since the fitting accuracy of the support vector machine regression model depends on the number of training sets, based on the specific values ​​of the half crack length a and crack position d in step three, the values ​​{2.0mm} and {±2.5mm} are added. The value of the leading edge position λ is given by the finite element calculation, and steps three and four are repeated to calculate the corresponding difference K * , a total of 345 sample points.

[0044] The embodiment of the present invention selects the support vector machine regression model (SVR) as the proxy model for the difference K * Fitting is performed to obtain the regression relationship between it and the three input variables of half crack length a, crack position d and front edge position λ. The detailed process is as follows:

[0045] How will the difference K be constructed? * The functional relationship between the half crack length a, the crack position d and the front edge position λ is transformed into the problem of establishing the optimal hyperplane. is the regression prediction value of the stress intensity factor difference, which can be expressed by the following function:

[0046]

[0047] Among them, w represents the weight vector in the initial sample space, and b represents the offset.

[0048] According to the structural risk minimization principle, in the embodiment of the present invention, the establishment of the optimal hyperplane is transformed into the following minimization problem:

[0049]

[0050] in, ] part represents the error accuracy, L represents the loss function, and the SVR model uses the ε-insensitive loss function ε represents the bandwidth of the ε-insensitive loss function interval. The error precision of samples within the bandwidth is recorded as zero; for samples outside the bandwidth, y i -f(x i )>ε, the error accuracy is recorded as y i -f(x i )<-ε, the error accuracy is recorded as ξ i . ||w|| 2 The part represents the confidence range, and the bandwidth of the ε-insensitive interval band is 2ε / (||w||) 2 , maximizing bandwidth is to minimize ||w||. K i * represents the true value, represents the predicted value of the support vector machine regression (SVR) model, and ||·|| represents the quadratic norm. The penalty coefficient C (C>0) is responsible for coordinating the balance between the error accuracy and confidence range of the SVR model.

[0051] Convert the problem of how to construct the optimal hyperplane into a minimization problem:

[0052]

[0053] The process of using the Lagrange multiplier method to solve the above minimization problem with inequality constraints is as follows:

[0054] Introducing Lagrange multipliers The above minimization problem with inequality constraints is transformed into the Lagrangian equation, using replace The abbreviation of is the same as follows:

[0055]

[0056] For variables Taking the derivative and simplifying we get:

[0057]

[0058] According to the Kuhn-Tucker condition (KKT), we get:

[0059]

[0060] Then, based on the Lagrange equation obtained by transforming the aforementioned minimization problem with inequality constraints, the problem of solving the SVR model is finally transformed into solving the quadratic programming problem shown in the following equation:

[0061]

[0062] φ(x i ) T φ(x j ) represents the inner product operation of two samples in the original space. For nonlinear regression problems, φ(x i ) T φ(x j ) is too computationally intensive, so the kernel function Kernal(x i ,x j )=φ(x i ) T φ(x j ), map the data to a high-dimensional feature space so that the data is approximately linearly separable in the high-dimensional feature space, and then perform linear regression.

[0063] The embodiment of the present invention selects a radial basis function (RBF) kernel to perform the above transformation, and the elements in the kernel matrix Kernal are:

[0064]

[0065] Among them, the width parameter γ is automatically selected to avoid the complexity of manually adjusting the parameters and provide reasonable performance. The kernel matrix Kernal of the embodiment of the present invention is calculated as:

[0066]

[0067] To solve the Lagrange multiplier The quadratic programming method is used to solve the problem and transform it into a matrix form of quadratic programming:

[0068]

[0069] Among them, the physical meaning of Lagrange multipliers is explained: when When , the sample points are located in a strip area with a width of ε, referred to as the ε interval zone. Usually, most of the α i and The value of is zero; when Sometimes, there is At this time, the sample point is at the upper boundary of the ε interval band. Similarly, when When , the sample point is at the lower boundary of the ε-interval band; when When , the sample point is outside the upper boundary of the ε interval band. , the sample point is outside the lower boundary of the ε interval band.

[0070] Each sample point can be considered as a vector. In the embodiment of the present invention, the half crack length a, crack position d and leading edge position λ of the sample point form a sample vector x=[ad λ]. and The vectors corresponding to the samples that are not equal to 0 at the same time are used as support vectors. Support vectors are used to determine the key points of the decision boundary. They determine the position and shape of the decision boundary, thereby affecting the prediction performance of the support vector machine regression model.

[0071] In the embodiment of the present invention, the support vector corresponding to the sample that meets the conditions is:

[0072]

[0073] At the same time, the response value K corresponding to the sample point * for: [0.13856226 0.13929272 0.16590188 0.14313055 0.15501498 0.200413650.20290985]

[0075] Solving the above minimization problem, we get the Lagrange multiplier as:

[0076] [-0.02023254 -0.05546061 0.05493472 -0.06356209 0.06610319 0.00103360.01718373]

[0077] Substitute the matrix form of the quadratic programming for the above problem of solving the Lagrange multiplier into the variable In the derived and simplified formula, the KKT conditions are combined to obtain the calculation formula for the offset b:

[0078]

[0079] The embodiment of the present invention solves and obtains the intercept, ie, the offset b=0.1590.

[0080] Correspondingly, the regression function equation for establishing the SVR model is:

[0081]

[0082] In the embodiment of the present invention:

[0083]

[0084] In the embodiment of the present invention, the RBF kernel is selected for calculation. For nonlinear SVR, the explicit regression formula will not be output, but the trained model can be saved, including data such as support vectors and kernel function parameters, and this model can be used to predict new data.

[0085] Mean square error (MSE) is one of the commonly used loss functions and evaluation indicators in regression problems. It represents the average value of the square of the difference between the model prediction value and the true value. The calculation formula is In the embodiment of the present invention, for the above 345 sample points and the corresponding difference The mean square error of the fitted SVR model is calculated to be 0.002168, indicating that the difference between the model prediction value and the true value is very small, and the prediction performance of the SVR model in the embodiment of the present invention is relatively good.

[0086] Finally, the stress intensity factor calculation formula used in the embodiment of the present invention is and the support vector machine regression equation Combined with the above, the final stress intensity factor correction formula is as follows:

[0087]

[0088] This stress intensity factor correction formula is oriented to welded structures and can be used in the next step of fatigue crack growth analysis and fatigue life analysis.

[0089] In the above steps, the correction of the crack growth rate of the welded structure is concentrated on the basic analytical formula of the stress intensity factor The correction is made by introducing K in the formula * As a correction term, the support vector machine model is selected to fit the correction term K*. The key parameters of half crack length, crack position and front edge position are used as input, and the stress intensity factor at the crack tip is used to numerically solve K e Analytical solution of stress intensity factor The difference between the two values ​​is used as the output, and a support vector machine model is established. The analytical solution is further combined to obtain a correction formula for calculating the stress intensity factor suitable for welded structures. This correction formula takes into account the influence of three key parameters on the stress intensity factor, improves the calculation accuracy of the stress intensity factor, and at the same time, this correction formula can be called to solve the stress intensity factor involved in the fatigue crack growth life calculation, without repeating the tedious steps of finite element modeling calculation, thereby improving the calculation efficiency of fatigue crack analysis.

[0090] Step 6: Substitute the stress intensity factor correction formula K'(a, d, λ) into the Paris crack growth formula Get the Paris correction formula The structural residual strength criterion is selected as the criterion to obtain the fatigue life prediction value of the welded structure.

[0091] Among them, the structural residual strength criterion is: under the action of typical alternating loads, the maximum stress intensity factor K at the crack tip is max Reach the fracture toughness K of the welding material at the crack tip c When the initial crack rapidly expands and forms a fracture, the number of load cycles N at this time is the predicted value of the fatigue life of the welded structure. The above typical alternating load refers to the sinusoidal cyclic load.

[0092] In the embodiment of the present invention, when fatigue crack growth analysis is performed, a sinusoidal cyclic load with a stress ratio R=0.1 is applied, and the peak value is consistent with the aforementioned load P=1MPa. The load and stress intensity factor correction formula K' is substituted into Where C mater ,m ial is a constant related to the material, and its value here is C material =8*10 -12 ,m material =3, Paris correction formula is:

[0093]

[0094] By shifting the terms and integrating the Paris correction formula, we can obtain:

[0095]

[0096] The number of load cycles N is an integer, and the stress intensity factor K' is updated once when N increases by one step. Since the increase in half crack length a is extremely small when N increases by one step, i.e., when the stress cycle is one time, K' is regarded as a constant in this process to calculate the crack growth rate, and the integration process can be converted into cumulative summation:

[0097]

[0098] Among them, a 0 is the initial half crack length, when 2a 0 +∑da=a max When the crack length is 2a 0 +∑da reaches the material fracture toughness K c The corresponding a max , output the corresponding number of load cycles N, which is the predicted value of fatigue life of the welded structure.

[0099] In the embodiment of the present invention, a is set respectively 1max =1.000005mm,a 2max =1.500005mm, fatigue crack growth analysis and prediction are carried out, and the results and relative errors are shown in Table 5:

[0100] Table 5 Fatigue crack growth analysis results and relative errors

[0101]

[0102] In this embodiment, the model results are tested by selecting parameters within the maximum value range of the training set, but not within the training set. Similarly, when selecting the parameter range in step 3, the values ​​that include the target working condition or the common working condition are selected as the maximum value of the value interval as much as possible, which can effectively improve the subsequent fitting accuracy.

[0103] In summary, the method for analyzing the crack propagation life of a welded structure based on a support vector machine model provided in an embodiment of the present invention is based on fracture mechanics and is intended for special engineering structures (e.g., structures containing welded structures) to achieve rapid calculation of the stress intensity factor at the crack tip. Among them, based on the static analysis results of the global model, the boundary conditions of the sub-model are obtained, and the mesh of the sub-model is redrawn to obtain a mesh unit that adapts to the stress field at the crack tip. According to the static analysis results and the characteristics of the welded structure, an initial crack is implanted at the weld to simulate the stable propagation of cracks in real components. The finite element method is used to calculate the numerical solution of the stress intensity factor, and the difference between the numerical solution and the analytical solution of the stress intensity factor is obtained under different half-crack lengths, crack positions, and front edge position parameters, and the support vector machine regression model is used to fit it. The stress intensity factor calculation formula and support vector machine regression model Combined with the above formula, the stress intensity factor correction formula is obtained. According to the Paris formula and the stress intensity factor correction formula, a rapid prediction of crack growth life is achieved. This method takes into account the influence of different parameters on the stress intensity factor. Compared with the theoretical calculation method, the crack growth life has a higher calculation accuracy. At the same time, the support vector machine regression model is used instead of the finite element simulation to further improve the calculation efficiency of the crack growth life.

[0104] In the description of this specification, the description of reference terms such as "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiments or examples are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0105] 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 method for analyzing crack growth life of welded structures based on a support vector machine model, characterized in that: include: Step 1: Use hexahedral units to discretize the geometric model of the welded structure to obtain a finite element model, and perform static simulation analysis to obtain node displacement and stress distribution calculation results; Step 2: implanting an initial crack at the weld center of the finite element model, dividing the finite element model into a sub-model containing the initial crack and other partial models, re-discretizing the sub-model using tetrahedral units, and encrypting the crack tip, retaining hexahedral units in the other partial models, retaining the nodes of the contact surface between the sub-model and the other partial models, and using the node displacement of the contact surface as the boundary condition of the sub-model; Step 3: Perform finite element calculation of crack propagation on the sub-model to obtain the numerical solution K of the stress intensity factor at the crack tip under different values ​​of half crack length a, crack position d and front edge position λ. e ; Step 4: Use the stress intensity factor calculation formula Calculate the analytical solution and calculate the numerical solution K e With the analytical solution K t The difference K * , K * =K e -K t ; Step 5: Take the half crack length a, the crack position d and the front edge position λ as input, and use the difference K * As the output, establish a support vector machine regression model According to the stress intensity factor calculation formula Get the stress intensity factor correction formula Step 6: Substitute the stress intensity factor correction formula K'(a, d, λ) into the Paris crack growth formula Get the Paris correction formula The structural residual strength criterion is selected as the criterion to obtain the fatigue life prediction value of the welded structure.

2. The analysis method according to claim 1, characterized in that The tetrahedral unit is a tetrahedral singular grid.

3. The analysis method according to claim 1, characterized in that The structural residual strength criterion is: under the action of typical alternating loads, the maximum stress intensity factor K at the crack tip is max The fracture toughness K of the welding material at the crack tip is reached c When the initial crack rapidly expands and forms a fracture, the corresponding number of load cycles N is the fatigue life prediction value of the welded structure.

4. The analysis method according to claim 1, characterized in that The step five comprises: According to the finite element calculation, the values ​​of the half crack length a, the crack position d and the leading edge position λ are increased, and the step three and the step four are repeated to obtain the corresponding difference K * ; The half crack length a, the crack position d and the front edge position λ are used as inputs, and the corresponding difference K * As output, the support vector machine regression model is established According to the support vector machine regression model And the stress intensity factor calculation formula Obtain the stress intensity factor calculation correction formula

Images