A curve fitting optimization method for fatigue stress-life method of welded joints

Abstract

The present invention relates to the technical field of welding fatigue analysis, and specifically discloses a curve fitting optimization method for fatigue stress-life method of welding joints, comprising the following steps: collecting welding joint fatigue test data, constructing a welding joint fatigue analysis database; preprocessing the data; calculating the Pearson correlation coefficient between each continuous attribute value in the welding joint fatigue analysis database; determining the key continuous attribute value feature division domain according to the result; first obtaining a traditional main S-N curve by the least squares fitting method, and then obtaining a family of S-N curves based on the feature division domain in the feature division domain by the least squares fitting method; performing goodness of fit statistical analysis on the traditional main S-N curve and the family of S-N curves based on the feature division domain by the goodness of fit statistical index, and comparing the goodness of fit statistical values ​​of the traditional main S-N curve and the family of S-N curves. The present invention can further reduce the dispersion of fatigue test samples and improve the prediction accuracy of fatigue life.

Description

Technical Field

[0001] The invention relates to the technical field of welding fatigue analysis. Background Art

[0002] Around the research of fatigue life analysis methods of welded joints, domestic and foreign scholars have done a lot of research work and proposed a variety of life prediction theories, models and analysis methods. Among them, the stress-life method is the most widely used life prediction method. The stress-life method is also called the SN curve method. The commonly used forms of the SN curve are the Basquin formula, power function, exponential function and three-parameter power function. According to the different basic stress parameters used, the SN curve method can be divided into the nominal stress method, the hot spot stress method, the notch stress method and the equivalent structural stress method (also known as the grid-insensitive structural stress method). Among them, the equivalent structural stress method solves the problem of consistency in stress calculation by defining a structural stress that is insensitive to the mesh size of the finite element analysis. Based on the fracture mechanics theory, a stress intensity factor expression including influencing factors such as joint plate thickness, load mode and stress concentration is established. The parameters in the expression are determined based on a large amount of welded joint fatigue test data to obtain a main SN curve that can characterize the fatigue life of the welded structure.

[0003] After the equivalent structural stress conversion is performed by comprehensively considering the joint plate thickness, load mode and stress concentration factors, the dispersion of the SN data sample points of the steel material welded joint characterized by the equivalent structural stress range-life has been greatly reduced, and all fatigue sample points are compressed into a narrow band. Compared with other methods, the equivalent structural stress method has the advantages of accurate structural stress calculation results and high fatigue life prediction accuracy, and can more accurately realize the analysis of the fatigue life of steel material welded joints. However, for aluminum alloy welded joints, the dispersion of the fatigue test samples of the main SN curve characterized by the equivalent structural stress range is still relatively high, resulting in the problem of unsatisfactory fatigue life prediction accuracy of welded joints. How to further reduce the dispersion of fatigue test samples and improve the fatigue life prediction accuracy is an urgent problem to be solved. Therefore, in view of the above-mentioned problems, it is very necessary to study and design a new type of welded joint fatigue stress-life method curve fitting optimization method to overcome the problem of unsatisfactory fatigue life prediction accuracy of existing welded joints. Summary of the invention

[0004] In order to solve the above problems existing in the existing prediction of fatigue life of welded joints, the present invention provides a curve fitting optimization method for fatigue stress-life method of welded joints.

[0005] The technical solution adopted by the present invention to achieve the above-mentioned purpose is a curve fitting optimization method of fatigue stress-life method of welded joints, comprising the following steps:

[0006] S1. Collect weld joint fatigue test data and build a weld joint fatigue analysis database;

[0007] S2. preprocessing the collected weld joint fatigue test data, eliminating erroneous data, and converting discrete attribute values ​​into continuous attribute values;

[0008] S3. Calculating the Pearson correlation coefficient between the continuous attribute values ​​in the weld joint fatigue analysis database by the Pearson correlation coefficient method;

[0009] S4. Determine the key continuous attribute value that affects the fatigue life of the welded joint according to the Pearson correlation coefficient result, and determine the feature division domain according to the key continuous attribute value;

[0010] S5. First, a traditional main SN curve is obtained by the least squares fitting method, and then the SN curve is fitted in the feature partition domain by the least squares fitting method to obtain a family of SN curves based on the feature partition domain;

[0011] S6. Perform goodness of fit statistical analysis on the traditional main SN curve and the family of SN curves based on feature partition domains through goodness of fit statistical indicators, and compare the goodness of fit statistical values ​​of the traditional main SN curve and the family of SN curves.

[0012] Preferably, in step S2, the continuous attribute values ​​in the weld joint fatigue analysis database include C1 material type, C2 welding method, C3 plate thickness, C4 stress ratio, C5 load type, C6 joint type, C7 equivalent structural stress range and D joint fatigue life.

[0013] Preferably, the Pearson correlation coefficient method in step S3 comprises the following steps:

[0014] S301. The continuous attribute values ​​in the weld joint fatigue analysis database are used as the decision information system S (C1, C2, C3, C4, C5, C6, C7, D), and the values ​​corresponding to each continuous attribute in the decision information system are used as a decision matrix and input;

[0015] S302. Normalizing the decision matrix by the maximum and minimum values ​​of each continuous attribute in the decision matrix;

[0016] S303. Calculate the covariance of the decision matrix after normalization, the formula is:

[0017]

[0018] Where x, y∈{C1,C2,...,C7} are any two continuous attributes in the decision information system, xi is the value of the i-th fatigue test sample in the decision information system on the continuous attribute x, y i is the value of the i-th fatigue test sample on the continuous attribute y in the decision information system; is the average value of the continuous attribute x for all fatigue test samples, is the average value of all fatigue test samples in the continuous attribute y, and n is the number of all fatigue test samples;

[0019] S304. Calculate the standard deviation of the decision matrix after normalization, the formula is:

[0020]

[0021] Among them, σ x is the standard deviation of the continuous attribute x, x i is the value of the i-th fatigue test sample on the continuous attribute x in the decision information system, is the average value of all fatigue test samples in the continuous attribute x, and n is the number of all fatigue test samples;

[0022] S305. Use covariance and standard deviation to calculate the Pearson correlation coefficient. The formula is:

[0023]

[0024] Among them, cov(x, y) is the covariance of the decision matrix after normalization, σ x is the standard deviation of the continuous attribute x, σ y is the standard deviation of the continuous attribute y.

[0025] Preferably, in step S302, the conversion function of the normalization process is:

[0026]

[0027] Among them, max{A i} is the maximum value corresponding to the continuous attribute, min{A i} is the minimum value corresponding to the continuous attribute, A i Represents the value of each continuous attribute in the decision matrix before normalization, B i Indicates the normalized value of each continuous attribute.

[0028] Preferably, in step S4, the key factors of welding fatigue life are plate thickness (C3) and stress ratio (C4).

[0029] Preferably, in step S4, the feature division domain is:

[0030] Feature domain 1: S1:{X∈U|X C3 =10and X C4 =0.1}

[0031] Feature domain 2: S2:{X∈U|X C3 =10and X C4 =0.5}

[0032] Feature domain three: S3:{X∈U|X C3 =2.5and X C4 =0.1}

[0033] Feature domain 4: S4:{X∈U|X C3 =3and X C4 =0}

[0034] Feature domain 5: S5:{X∈U|X C3 =4.7625and X C4 =0}

[0035] Among them, X∈U represents the fatigue test sample point; X C3 =10and X C4 =0.1 represents all fatigue test sample points where the plate thickness is 10 and the stress ratio is 0.1, and the meaning and definition rules of other parameters are the same.

[0036] Preferably, in step S5, the power function expression of the SN curve fitting is:

[0037] S m N=C

[0038] Among them, m and C are parameters related to stress ratio, material type, and load form properties;

[0039] Taking the logarithm of both sides of the equation gives the formula:

[0040] lgS=AlgN+B

[0041] Among them, A=-1 / m, B=lgC / m.

[0042] Preferably, in step S5, a power function is used as the expression form of the SN curve, and a traditional main SN curve in the traditional grid-insensitive structural stress method is obtained by the least squares fitting method. At the same time, the least squares fitting method is used to fit the SN curve based on the delineated feature domain to obtain a family of SN curves based on the feature partition domain.

[0043] Preferably, the goodness-of-fit statistical indicators in step S6 include the sum variance, the determination coefficient, the corrected determination coefficient and the fitting standard deviation of the regression coefficient.

[0044] Preferably, the calculation formula of the sum and variance is:

[0045]

[0046] The calculation formula of the determination coefficient is:

[0047]

[0048] SSR is the sum of the squares of the differences between the predicted data and the original data mean:

[0049]

[0050] SST is the sum of the squares of the differences between the original data and the mean:

[0051]

[0052] The calculation formula of the corrected determination coefficient is:

[0053]

[0054] The calculation formula of the regression coefficient is:

[0055]

[0056] Among them, y i is the value of the i-th fatigue test sample on the continuous attribute y in the decision information system, is the predicted value of the i-th fatigue test sample on the continuous attribute y in the decision information system, n is the number of all fatigue test samples, and j is the fitting coefficient.

[0057] A welding joint fatigue stress-life method curve fitting optimization method of the present invention constructs a welding joint fatigue life influencing factor analysis model based on the Pearson correlation coefficient method, reveals the coupling relationship between the various influencing factors of the welding joint fatigue life, determines the feature division domain aimed at reducing the dispersion level of fatigue test sample points, and performs SN curve fitting based on the determined feature division domain to obtain a family of SN curves. Compared with an SN curve in a traditional grid-insensitive structural stress method obtained by using the least squares fitting method, when a family of SN curves based on the feature division domain is used to predict the fatigue life of the welding joint or evaluate the cumulative damage, the dispersion of the fatigue test samples can be further reduced, thereby improving the prediction accuracy of the fatigue life. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 It is a flow chart for predicting fatigue life of welded joints based on equivalent structural stress;

[0059] Figure 2 is a technical flow chart of an embodiment of the present invention;

[0060] Figure 3 is a flow chart of the Pearson correlation coefficient method according to an embodiment of the present invention;

[0061] Figure 4 It is a main SN curve diagram in the embodiment of the present invention based on the traditional grid-insensitive structural stress method;

[0062] Figure 5 It is a family of SN curve diagrams based on feature partitioning domains in an embodiment of the present invention. DETAILED DESCRIPTION

[0063] The following embodiments of the present invention are described in further detail in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0064] The traditional fatigue life prediction process of welded joints based on equivalent structural stress is as follows: Figure 1 As shown in Figure 1, although the calculation results are relatively accurate and the fatigue life prediction accuracy is high, there is still a problem of high dispersion of fatigue test samples. In order to solve this problem, the present invention provides a welded joint fatigue stress-life method curve fitting optimization method as shown in Figure 1. Figure 2 As shown, the following steps are included:

[0065] S1. Collect weld joint fatigue test data and build a weld joint fatigue analysis database;

[0066] S2. Preprocess the collected weld joint fatigue test data, remove erroneous data, and convert discrete attribute values ​​into continuous attribute values ​​to facilitate the analysis of the impact of related attribute values;

[0067] The continuous attribute values ​​in the welded joint fatigue analysis database may include: C1 material type, C2 welding method, C3 plate thickness, C4 stress ratio, C5 load type, C6 joint type, C7 equivalent structural stress range, D joint fatigue life;

[0068] S3. Calculate the Pearson correlation coefficient between the continuous attribute values ​​in the weld joint fatigue analysis database by the Pearson correlation coefficient method, analyze the coupling relationship between the various influencing factors of the weld joint fatigue life based on the calculation results of the Pearson correlation coefficient, and reveal the implicit relationship between the various influencing factors;

[0069] like Figure 3 As shown, the Pearson correlation coefficient method in step S3 may include the following steps:

[0070] S301. The continuous attribute values ​​in the weld joint fatigue analysis database are used as the decision information system S (C1, C2, C3, C4, C5, C6, C7, D), and the values ​​corresponding to each continuous attribute in the decision information system are used as a decision matrix and input;

[0071] S302. Normalize the decision matrix by the maximum and minimum values ​​of each continuous attribute in the decision matrix;

[0072] S303. Calculate the covariance of the decision matrix after normalization, the formula is:

[0073]

[0074] Among them, x,y∈{C1,C2,...,C7} are any two continuous attributes in the decision information system, x i is the value of the i-th fatigue test sample in the decision information system on the continuous attribute x, y i is the value of the i-th fatigue test sample on the continuous attribute y in the decision information system; is the average value of the continuous attribute x for all fatigue test samples, is the average value of all fatigue test samples in the continuous attribute y, and n is the number of all fatigue test samples;

[0075] S304. Calculate the standard deviation of the decision matrix after normalization, the formula is:

[0076]

[0077] Among them, σ x is the standard deviation of the continuous attribute x, x i is the value of the i-th fatigue test sample on the continuous attribute x in the decision information system, is the average value of all fatigue test samples in the continuous attribute x, and n is the number of all fatigue test samples;

[0078] S305. Use covariance and standard deviation to calculate the Pearson correlation coefficient. The formula is:

[0079]

[0080] Among them, cov(x, y) is the covariance of the decision matrix after normalization, σ x is the standard deviation of the continuous attribute x, σ y is the standard deviation of the continuous attribute y;

[0081] S4. Determine the key continuous attribute value that affects the fatigue life of the welded joint according to the Pearson correlation coefficient result, and determine the feature division domain according to the key continuous attribute value;

[0082] S5. First, a traditional main SN curve is obtained by the least square fitting method, and then the SN curve is fitted in the feature partition domain by the least square fitting method to obtain a family of SN curves based on the feature partition domain;

[0083] The power function expression of SN curve fitting is:

[0084] S m N=C

[0085] Among them, m and C are parameters related to stress ratio, material type, and load form properties;

[0086] Taking the logarithm of both sides of the equation gives the formula:

[0087] lgS=AlgN+B

[0088] Wherein, A = -1 / m, B = lgC / m;

[0089] Taking the power function type as the expression form of SN curve, the least square fitting method is used to obtain a traditional main SN curve in the traditional grid-insensitive structural stress method. At the same time, the least square fitting method is used to fit the SN curve based on the delineated characteristic domain to obtain a family of SN curves based on the characteristic partition domain.

[0090] S6. Perform goodness of fit statistical analysis on the traditional main SN curve and the family of SN curves based on the feature partition domain by using goodness of fit statistical indicators, and compare the goodness of fit statistical values ​​of the traditional main SN curve and the family of SN curves;

[0091] Goodness-of-fit statistics can include the sum variance, coefficient of determination, corrected coefficient of determination, and standard deviation of the regression coefficient.

[0092] The calculation formula of sum and variance is:

[0093]

[0094] The formula for calculating the coefficient of determination is:

[0095]

[0096] SSR is the sum of the squares of the differences between the predicted data and the original data mean:

[0097]

[0098] SST is the sum of the squares of the differences between the original data and the mean:

[0099]

[0100] The formula for calculating the corrected coefficient of determination is:

[0101]

[0102] The regression coefficient is calculated as:

[0103]

[0104] Among them, y i is the value of the i-th fatigue test sample on the continuous attribute y in the decision information system, is the predicted value of the i-th fatigue test sample on the continuous attribute y in the decision information system, n is the number of all fatigue test samples, and j is the fitting coefficient.

[0105] In order to explain the above steps in more detail, the following describes the above process analysis and calculation results in detail by taking the construction of an aluminum alloy welded joint fatigue analysis database as an example.

[0106] Through collection and collation, relevant data of fatigue tests of internationally renowned associations, enterprises and scholars were obtained, and combined with fatigue test data obtained from laboratory fatigue tests, a fatigue analysis database of aluminum alloy basic welded joints was constructed. The fatigue analysis database of aluminum alloy welded joints contains attributes with discrete values ​​and attributes with continuous values. The attributes with discrete values ​​are: material type, welding method, load type and joint type. Among them, there are 4 welding methods, 5 material types, 2 load types and 4 joint types. The attributes with continuous values ​​are: plate thickness, stress ratio, stress concentration factor, nominal stress, equivalent structural stress and joint fatigue life. The obviously erroneous data were eliminated, and the discrete attribute values ​​were converted into continuous attribute values. Some information in the database is shown in Table 1, where: C1 is material type, C2 is welding method, C3 is plate thickness, C4 is stress ratio, C5 is load type, C6 is joint type, C7 is equivalent structural stress range, and D is joint fatigue life.

[0107] Table 1 Original database for fatigue analysis of aluminum alloy welded joints (partial)

[0108]

[0109] The fatigue analysis database of aluminum alloy welded joints is used as the decision information system S (C1, C2, C3, C4, C5, C6, C7, D), and the Pearson correlation coefficient method is used to calculate the correlation between each continuous attribute. The specific steps include:

[0110] Step 1. Take the continuous attribute values ​​in the weld joint fatigue analysis database as the decision information system S (C1, C2, C3, C4, C5, C6, C7, D), and take the values ​​corresponding to each continuous attribute in the decision information system as the decision matrix and input;

[0111] Step 2. Normalize the decision matrix by taking the maximum and minimum values ​​of each continuous attribute in the decision matrix;

[0112] Data normalization refers to scaling data so that the data falls into a small specific interval. It can remove the unit restriction of the data and convert it into a dimensionless pure value, so that indicators of different units or magnitudes can be compared and weighted. The most typical one is data normalization, that is, uniformly mapping the data to the interval [0, 1]. The method used in the present invention is data normalization, namely, min-max normalization, also called deviation normalization, and its conversion function is as follows:

[0113]

[0114] Among them, max{A i} is the maximum value corresponding to the continuous attribute, min{A i} is the minimum value corresponding to the continuous attribute, A i Represents the value of each continuous attribute in the decision matrix before normalization, B i Indicates the normalized value of each continuous attribute;

[0115] Step 3. Calculate the covariance of the decision matrix after normalization. The formula is:

[0116]

[0117] Among them, x,y∈{C1,C2,...,C7} are any two continuous attributes in the decision information system, x i is the value of the i-th fatigue test sample in the decision information system on the continuous attribute x, y i is the value of the i-th fatigue test sample on the continuous attribute y in the decision information system; is the average value of the continuous attribute x for all fatigue test samples, is the average value of all fatigue test samples in the continuous attribute y, and n is the number of all fatigue test samples;

[0118] Step 4. Calculate the standard deviation of the decision matrix after normalization. The formula is:

[0119]

[0120] Among them, σ x is the standard deviation of the continuous attribute x, x i is the value of the i-th fatigue test sample on the continuous attribute x in the decision information system, is the average value of all fatigue test samples in the continuous attribute x, and n is the number of all fatigue test samples;

[0121] step5. Use covariance and standard deviation to calculate the Pearson correlation coefficient. The formula is:

[0122]

[0123] Among them, cov(x, y) is the covariance of the decision matrix after normalization, σ x is the standard deviation of the continuous attribute x, σ y is the standard deviation of the continuous attribute y.

[0124] The Pearson correlation coefficient method is used to calculate the correlation between each continuous attribute, and the Pearson correlation coefficients between other continuous attributes (C1~C6) and the equivalent structural stress range attribute (C7) are: {0.12, 0.13, -0.14, -0.35, 0.13, 0.13}. Based on this, it can be judged that the equivalent structural stress range is greatly affected by the plate thickness (C3) and the stress ratio (C4). Therefore, according to the actual conditions of the plate thickness (C3) and the stress ratio (C4), the characteristic division domain is determined. Assuming that the domain of all fatigue test samples is U, the characteristic division domain is obtained as follows:

[0125] Feature domain 1: S1:{X∈U|X C3 =10 and X C4 =0.1}

[0126] Feature domain 2: S2:{X∈U|X C3 =10 and X C4 =0.5}

[0127] Feature domain three: S3:{X∈U|X C3 =2.5 and X C4 =0.1}

[0128] Feature domain 4: S4:{X∈U|X C3 =3 and X C4 =0}

[0129] Feature domain 5: S5:{X∈U|X C3 =4.7625 and X C4 =0}

[0130] Among them, X∈U represents the fatigue test sample point; X C3 =10 and X C4 =0.1 represents all fatigue test sample points where the plate thickness is 10 and the stress ratio is 0.1, and the meaning and definition rules of other parameters are the same.

[0131] The SN curve is also called the stress-life curve, which usually has three types: power function type, exponential type and three-parameter type. Among them, the power function type is the most common expression form, and the expression is:

[0132] S m N=C

[0133] Among them, m and C are parameters related to stress ratio, material type, and load form properties;

[0134] Taking the logarithm of both sides of the equation gives the formula:

[0135] lgS=AlgN+B

[0136] Among them, A=-1 / m, B=lgC / m.

[0137] The power function type is used as the expression form of the SN curve, and the least square fitting method is used to obtain the SN curve in the traditional grid-insensitive structural stress method (such as Figure 4 At the same time, the least squares fitting method is used to fit the SN curve based on the defined feature domain to obtain a family of SN curves (as shown in Figure 5 As shown in the figure, points of different shapes represent different fatigue test sample points and belong to different feature domains. The lines connecting points of different shapes represent the median SN curves in the corresponding feature domains. Feature domain one is Mean1, feature domain two is Mean2, feature domain three is Mean3, feature domain four is Mean4, and feature domain five is Mean5.

[0138] Depend on Figure 4 and Figure 5 It can be seen that in the traditional grid-insensitive structural stress method, the SN curve obtained by fitting based on the fatigue test sample points is one; however, based on the Pearson correlation coefficient analysis results, the characteristic domain is defined, and a family of SN curves can be fitted within the defined characteristic domain, which can further reduce the dispersion level of fatigue test samples and improve the accuracy of fatigue life prediction of welded joints, thereby achieving the optimization of the main SN curve fitting.

[0139] The effectiveness of the main SN curve fitting optimization method based on the feature partition domain is verified by comparing the calculation results of the goodness of fit statistical index values. The calculation of the goodness of fit statistical index values ​​is shown in Table 2. Among them, SSE is the sum variance, or the sum of squared errors. The closer SSE is to 0, the better the model selection and fitting are, and the more successful the data prediction is. R-square is the determination coefficient, which mainly characterizes the quality of the fit through the change of the data. Its value range is [0, 1]. Generally, the closer its value is to 1, the better the model fits the data. Adjusted R-square is called the corrected determination coefficient, which can obtain any value less than 1 or equal to 1. Generally, the closer its value is to 1, the better the fitting performance is. RMSE is called the fitting standard deviation of the regression coefficient, which is the square root of the mean square error. Similar to SSE, the closer its value is to 0, the better the fitting performance is, and the more successful the data prediction is.

[0140] Table 2 Comparison of goodness-of-fit statistics

[0141]

[0142] As shown in Table 2, according to the calculation results of the goodness of fit statistical index value, the sum of square errors (SSE) of a family of SN curves Mean1~Mean5 obtained based on the feature partition domain fitting is smaller than the sum of square errors of a traditional main SN curve Mean obtained by the traditional grid-insensitive structural stress method; except for Mean3, the correction coefficient (R-square) and adjusted correction coefficient (Adjusted R-square) of other SN curves obtained based on the feature partition domain fitting are closer to 1 than the traditional main SN curve Mean in the traditional grid-insensitive structural stress method; the standard deviation (RMSE) of the SN curve obtained based on the feature partition domain fitting is smaller than the standard deviation of a traditional main SN curve Mean in the traditional grid-insensitive structural stress method, so the SN curve obtained based on the feature partition domain fitting has better fatigue life prediction performance.

[0143] The statistical analysis results show that after defining the characteristic domain, the dispersion level of fatigue data samples has been further reduced, and the standard deviation of the SN curve obtained by fitting based on the characteristic partition domain has been further reduced overall compared with the SN curve directly fitted without defining the characteristic domain.

[0144] The embodiments of the present invention are given for the purpose of illustration and description, and are not intended to be exhaustive or to limit the invention to the disclosed forms. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present invention and to enable those of ordinary skill in the art to understand the present invention and thereby design various embodiments with various modifications suitable for specific uses.

Claims

1. A weld joint fatigue stress-life method curve fitting optimization method, characterized in that: The steps include: S1. Collect weld joint fatigue test data and build a weld joint fatigue analysis database; S2. preprocessing the collected weld joint fatigue test data, eliminating erroneous data, and converting discrete attribute values ​​into continuous attribute values; S3. Calculating the Pearson correlation coefficient between each continuous attribute value in the weld joint fatigue analysis database by the Pearson correlation coefficient method; S4. Determine the key continuous attribute value that affects the fatigue life of the welded joint according to the Pearson correlation coefficient result, and determine the feature division domain according to the key continuous attribute value; S5. First, a traditional main SN curve is obtained by the least squares fitting method, and then the SN curve is fitted in the feature partition domain by the least squares fitting method to obtain a family of SN curves based on the feature partition domain; S6. Perform goodness of fit statistical analysis on the traditional main SN curve and the family of SN curves based on feature partition domains through goodness of fit statistical indicators, and compare the goodness of fit statistical values ​​of the traditional main SN curve and the family of SN curves.

2. The weld joint fatigue stress-life method curve fitting optimization method according to claim 1, characterized in that: In step S2, the continuous attribute values ​​in the weld joint fatigue analysis database include C1 material type, C2 welding method, C3 plate thickness, C4 stress ratio, C5 load type, C6 joint type, C7 equivalent structural stress range and D joint fatigue life.

3. The weld joint fatigue stress-life method curve fitting optimization method according to claim 2, characterized in that: The Pearson correlation coefficient method in step S3 comprises the following steps: S301. The continuous attribute values ​​in the weld joint fatigue analysis database are used as the decision information system S (C1, C2, C3, C4, C5, C6, C7, D), and the values ​​corresponding to each continuous attribute in the decision information system are used as a decision matrix and input; S302. Normalizing the decision matrix by the maximum and minimum values ​​of each continuous attribute in the decision matrix; S303. Calculate the covariance of the decision matrix after normalization, the formula is: Where x, y∈{C1,C2,...,C7} are any two continuous attributes in the decision information system, x i is the value of the i-th fatigue test sample in the decision information system on the continuous attribute x, y i is the value of the i-th fatigue test sample on the continuous attribute y in the decision information system; is the average value of the continuous attribute x for all fatigue test samples, is the average value of all fatigue test samples in the continuous attribute y, and n is the number of all fatigue test samples; S304. Calculate the standard deviation of the decision matrix after normalization, the formula is: Among them, σ x is the standard deviation of the continuous attribute x, x i is the value of the i-th fatigue test sample on the continuous attribute x in the decision information system, is the average value of all fatigue test samples in the continuous attribute x, and n is the number of all fatigue test samples; S305. Use covariance and standard deviation to calculate the Pearson correlation coefficient. The formula is: Among them, cov(x, y) is the covariance of the decision matrix after normalization, σ x is the standard deviation of the continuous attribute x, σ y is the standard deviation of the continuous attribute y.

4. The method for optimizing the curve fitting of fatigue stress-life method of welded joints according to claim 3, characterized in that: In step S302, the conversion function of the normalization process is: Among them, max{A i } is the maximum value corresponding to the continuous attribute, min{A i } is the minimum value corresponding to the continuous attribute, A i Represents the value of each continuous attribute in the decision matrix before normalization. i Indicates the normalized value of each continuous attribute.

5. The weld joint fatigue stress-life curve fitting optimization method according to claim 3, characterized in that: In step S4, the key factors of welding fatigue life are plate thickness (C3) and stress ratio (C4).

6. The weld joint fatigue stress-life curve fitting optimization method according to claim 5, characterized in that: In step S4, the feature division domain is: Feature domain 1: S1:{X∈U|X C3 =10and X C4 =0.1} Feature domain 2: S2:{X∈U|X C3 =10and X C4 =0.5} Feature domain three: S3:{X∈U|X C3 =2.5and X C4 =0.1} Feature domain 4: S4:{X∈U|X C3 =3and X C4 =0} Feature domain 5: S5:{X∈U|X C3 =4.7625and X C4 =0} Among them, X∈U represents the fatigue test sample point; X C3 =10and X C4 =0.1 represents all fatigue test sample points where the plate thickness is 10 and the stress ratio is 0.1, and the meaning and definition rules of other parameters are the same.

7. The weld joint fatigue stress-life curve fitting optimization method according to claim 6, characterized in that: In step S5, the power function expression of the SN curve fitting is: S m N=C Among them, m and C are parameters related to stress ratio, material type, and load form properties; Taking the logarithm of both sides of the equation gives the formula: lg S=A lg N+B Among them, A = -1 / m, B = lg C / m.

8. The weld joint fatigue stress-life method curve fitting optimization method according to claim 7, characterized in that: In step S5, a power function type is used as the expression form of the SN curve, and a traditional main SN curve in the traditional grid-insensitive structural stress method is obtained by the least squares fitting method. At the same time, the least squares fitting method is used to fit the SN curve based on the defined feature domain to obtain a family of SN curves based on the feature partition domain.

9. The weld joint fatigue stress-life method curve fitting optimization method according to claim 8, characterized in that: The goodness of fit statistical indicators in step S6 include the sum variance, the determination coefficient, the corrected determination coefficient and the fitting standard deviation of the regression coefficient.

10. A weld joint fatigue stress-life method curve fitting optimization method according to claim 9, characterized in that: The calculation formula of the sum and variance is: The calculation formula of the determination coefficient is: SSR is the sum of the squares of the differences between the predicted data and the original data mean: SST is the sum of the squares of the differences between the original data and the mean: The calculation formula of the corrected determination coefficient is: The calculation formula of the regression coefficient is: Among them, y i is the value of the i-th fatigue test sample on the continuous attribute y in the decision information system, is the predicted value of the i-th fatigue test sample on the continuous attribute y in the decision information system, n is the number of all fatigue test samples, and j is the fitting coefficient.

Images