KR100706048B1 - A method of development of fatigue assessment of welding joint - Google Patents

Abstract

Fatigue strength evaluation methods are disclosed that determine various material constants required for a material damage model using genetic algorithm techniques.

In the fatigue strength evaluation method according to the present invention, after inputting only standard mechanical property values and using them, the basic data of the material damage model is constructed, and the final damage model is derived based on the comparison with the results of the uniaxial tensile test. Also, the step of constructing the basic data of the material damage model inputs standard mechanical property values to the inference engine, infers a material constant for the damage engine, and then verifies the material constant for the material damage evaluation. The relationship between fatigue life of fatigue characteristics is determined using standard deviations, and basic data of optimized material damage model using inference model and database are constructed.

Fatigue Strength, Weld Joints of Steel, Damage Mechanics, Genetic Algorithm

Description

A method of development of fatigue assessment of welding joint

1 is a block diagram schematically illustrating an optimization technique according to the present invention.

2 is a flowchart illustrating a material constant inference step in an inference engine.

3 is a flowchart showing a step of determining a material constant of an inexperienced steel.

4 is a graph showing the relationship between the Young's modulus and the visco-plastioc constants of the material in the correlation equation of the uniaxial tensile test data of the material constant.

5 is a graph showing the relationship between yield strength and viscous constant of a material.

FIG. 6 is a graph showing the relationship between ultimate tensile strength and viscous constant of materials.

FIG. 7 is a graph showing the relationship between the fracture elongation of materials and viscous firing.

8 is a graph showing the numerical relationship between the relationship between viscoplastic parameters (S) and material damage variables in the definition of plastic strain rate.

9 is a graph comparing empirical and actual inexperienced steels.

10 is a flowchart illustrating a process of evaluating structural health.

11 is a block diagram showing a structure for obtaining a residual stress estimate capable of estimating characteristics due to defects at the beginning of the structure.

12 is a flowchart illustrating a process of calculating a final damage amount after an initial defect is quantitatively calculated.

13 is an example of modeling for performing fatigue characteristic evaluation using geometric analysis data.

FIG. 14 is a conceptual diagram of performing basic structural analysis by inputting basic material properties, weld shape dimensions, and welding influence effects in order to predict weld fatigue horizontality. FIG.

FIG. 15 is a schematic table of fatigue life analysis performed at the point of maximum peripheral strain using Manson-Halford equation and SWT equation in this structural analysis.

16 is a method for applying a weld fatigue life prediction model applied in the present system.

17 is an example of modeling for verifying fatigue evaluation of a weld in the Manson-Halford equation and the SWT equation and the present system.

18 is a graph showing the results of fatigue analysis applied to the Manson-Halfordtlr and SWT equations.

Figure 19 shows the program fatigue evaluation results in order to compare the fatigue evaluation and damage-based evaluation techniques strain-based in Tso-Lianf Teng.

The present invention relates to a fracture problem analysis method using damage mechanics, and more particularly, to a technique for determining various material constants required for a material damage model using a genetic algorithm technique.

Recently, with the development of computer, the development of numerical analysis method for fatigue failure problem has been attempted in various fields. In addition, these attempts are basically based on finite element analysis, and there are practical difficulties due to the large amount of computation. Despite these difficulties, various theories such as fracture mechanics and damage mechanics have been proposed to deal with fatigue problems.

When the two methods are largely divided and described, first, fracture mechanics is related to the propagation behavior of cracks in an intact material with band-width parameters. Evaluated by fracture analysis. Fracture mechanics, however, is characterized by the fact that a large number of microcracks occur and develop in the material and they combine to form macroscopic cracks, when inelastic deformation is present at the crack tip, and the working load is mixed or non-proportional. In this case, there is a problem with application limitations.

On the other hand, damage mechanics, another approach to the problem of damage destruction of materials and structures, provides a complete theoretical basis because all dynamic variables are described within the scope of continuum mechanics. For this reason, it has been applied to various types of damage / destruction problems and contributed to the systematic development of numerical simulation methods for damage and destruction problems in materials or structures. Based on this theoretical basis, the method of numerically tracking the whole process from the initial intact state to the occurrence of a crack is called Local Approach to Fracture.

Damage mechanics itself has a relatively short history compared to the long history of solid mechanics, but recently, it is a technique that is actively employed in the damage-related problems of various structures. In particular, crack generation simulation through calculation of equations in virtual space has been recognized as a leading research field in the field of material strength, as it is recognized as a core technology for cost-saving effect and safety design concept for predicting related problems. The field of application is expanding.

However, the method of deriving the theory or constitutive equations that can describe the material properties remains a problem that must be solved, which in turn impedes the improvement of quantitative reliability of the analysis and simulation results and the expansion to design support technology. It is true.

The technical problem to be achieved by the present invention is to provide an evaluation method for fatigue strength evaluation of welded joints and an integrated system of GUI environment using the same, and to reduce the analysis time according to the system development and to predict the fatigue life depending on the long-term test. It is to provide a method and an integrated system that can be enabled in a short time.

Another technical problem to be solved by the present invention is to provide a method and system that can easily process modeling, analysis, and pre-analysis processing in a GUI environment.

In order to achieve the object of the present invention as described above, according to a feature of the present invention, a method for evaluating the fatigue strength of the welded joint of the steel, receiving a standard mechanical property value; Constructing basic data of a material damage model using the standard mechanical properties; And deriving a final damage model by comparing the basic data of the material damage model with the uniaxial tensile test results for the fatigue strength evaluation model of the welded portion, and constructing the basic data of the material damage model includes: Inputting standard mechanical properties into the inference engine; Inferring material constants for the damage engine; Verifying material constants for material damage assessment; Determining a fatigue life analysis (S-N) relationship of fatigue characteristic values using standard deviations; And constructing basic data of the optimized material damage model using the inference model and the database.

In order to fully understand the present invention, the advantages of the operability of the present invention, and the objects achieved by the practice of the present invention, reference should be made to the accompanying drawings which illustrate preferred embodiments of the present invention and the contents described in the accompanying drawings.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. Like reference numerals in the drawings denote like elements.

Optimization techniques using genetic algorithms, which are actively applied in various engineering problems, enable systematic derivation of theories and constitutive equations to describe material properties. In addition, even from the standpoint of the designer who does not have the expertise, the GUI type expert system consists of the precision damage model construction part necessary for the soundness evaluation and various strength characteristics evaluation technology applied to it. At the same time, it can be expected to serve as a practical design support tool.

Accordingly, the present invention uses the optimization technique to input only basic input data for steel materials, for example, standard mechanical properties, and to use the data to construct the basic data of the material damage model. By comparing the analytical properties of the material with that of the final damage model is obtained.

1 is a block diagram schematically illustrating an optimization technique according to the present invention.

In order to apply damage mechanics, 13 mechanical parameters are required. However, there is a problem that it takes too much experiment time to calculate all 13 mechanical properties of the steel. Therefore, if you use a database of known mechanical properties or know a few properties (for example, five variables), you can use the known mechanical properties to calculate the remaining mechanical properties through a formula.

Referring to FIG. 1, the optimization technique 100 estimates fatigue characteristics using mechanical characteristics based on the damage mechanics approach.

Specifically, first, the known mechanical property value is input to the inference engine (step 101). The inference engine then infers the material parameters for damage calculation. That is, to derive the final damage model, the S-S curve using the damage role is derived through the material properties (S-S curve) of the input material (step 103). The inference engine then verifies the material parameters for material damage assessment (step 105). The fatigue life analysis relationship of the fatigue characteristic values is determined using a standard deviation or the like (step 107), and the optimized data are reported through the estimation result and the database (step 109).

On the other hand, in the optimization technique 100 of FIG. 3, the error rate from the tensile test result, the fatigue test result, and the simulation analysis result, which are the simplest data representing the fracture characteristics of the material, is used. In other words, in step 103 and step 105, the inference engine receives the basic mechanical properties of the steel, and analyzes the material constants required for the finite element analysis program mounted therein in the tensile test results, fatigue test results, and simulation analysis results. These material constants are determined by mathematically optimizing the error rate.

2 is a flowchart illustrating a material constant inference step in an inference engine.

Referring to FIG. 2, the material constant inference step (step 103) of FIG. 1 infers the material constant through the following three inference steps.

First, using the known reasoning method (memory data) using the memory-based reasoning method to find the material constant for analysis (step 201). Secondly, the empirical formula is defined using the correlation between the mechanical characteristic value and the analysis material constant in the experimental numerical value in the data (step 203). Thirdly, the optimized analysis material constants are calculated using a genetic program to optimize the relationship between the memory-based analysis material constants and the analysis material constants modified by the empirical equations in the data. That is, in constructing an analysis material constant through existing known data (data), an error range is optimized by using a genetic program (genetic algorithm) to construct an analysis material constant of a material to be obtained ( Step 205).

First, a process of obtaining an analysis material constant using the memory-based inference of step 201 will be described.

Determination of the material constant using the memory-based reasoning method uses the information corresponding to the material constants of the existing class steels and various kinds of experienced steels, and thus it is easy to determine the material constants for inexperienced steels. It is easy to determine the final target constant because the constant is reduced through the decision relationship between new and inexperienced and experienced steels.

The first one finds similar cases from empirical reasoning and then finds and applies the necessary information of parameters by this empirical reasoning. This is Memory-Based Reasoning (MBR). Memory-based reasoning is a way of using experiences similar to direct data mining techniques. By keeping known records in the database, memory-based reasoning finds new neighbors and those neighbors are used for classification and prediction.

The advantage of memory-based reasoning is that the ability to reason about common constants uses the data itself. Unlike other data mining techniques, this method does not care about the format of the record. Memory-based reasoning methods focus on two things: A distance function that represents the distance between two records and a combination function that combines the results of reaching the answer. These functions are easily defined with the standard data types of almost all records. It can also handle data types that are difficult to handle with other analytical techniques.

Memory-based reasoning is also a good choice for relational data. That is, to deal with complex records such as material properties such as σ _Y , σ _T , E for each material, and combination functions for material constants, which are inherent damage mechanics, and sometimes missing values in some fields. Also used to.

3 is a flowchart showing a step of determining a material constant of an inexperienced steel.

3 is a flowchart showing the first order inference of the basic material constants of the memory based reasoning, which corresponds to step 201 of FIG. In step 201, a step of receiving a material constant of an experienced steel from an experienced steel database (step 301), estimating a memory-based steel material (step 303), and determining a material constant of an inexperienced steel (step 305).

On the other hand, the nearest neighbor approach is very local. Only the new storage record, which is the most similar to the new steel, serves as an empirical parameter for inexperienced steels. The relationship of the determination of mechanical specific variables for inexperienced steels in memory-based reasoning can primarily infer the necessary analysis material constants that are programmed in some modules of the system.

Next, the empirical definition process of step 202 will be described.

Step 202 is a step of defining an empirical equation using the correlation between the mechanical characteristic value and the analysis material constant in the experimental numerical value in the data, and first, it is necessary to infer the material integer correlation.

The relationship between each mechanical damage parameter (e.g., Young's Modules, Yield Strength, Tensile Strength, Fracture Elongation) for each empirical steel in the inference engine and each material damage variable can be derived using a mathematical empirical formula. . Here each parameter needs to accurately describe its mechanical properties. In one embodiment of the present invention, a technique for determining the various material constants required for a material damage model is empirical. In other words, input only basic input data for steel, for example, standard mechanical properties, and use it to compare basic data of material damage model and test results with basic information of each material. Induce.

Then, in the uniaxial tensile test, the results of the basic input information of each material and the test results can be compared and derived in experimental polynomial form. At this time, the maximum minimum relationship of the relational inference formula is calculated by applying the minimum and maximum relationships of existing data in the relationship index of the actual material.

FIG. 4 is a graph showing the relationship between Young's modulus and visco-plastioc constants of materials in the correlation of uniaxial tensile test data of material constants. 5 is a graph showing the relationship between yield strength and viscous constant of a material. FIG. 6 is a graph showing the relationship between ultimate tensile strength and viscous constant of materials. 7 is a graph showing the relationship between fracture elongation and viscous constant of a material.

In the fatigue analysis of materials, the material constants necessary to formulate the correlation between the mechanical properties of each material and the analysis material constants can be calculated based on the actual data for the experienced steels.

Equations 1 and 2 below are correlations between the viscoplastic parameters of the plastic strain rate and the damage coefficients S ₁ and X of the material.

8 is a graph showing the numerical relationship between the viscoplastic parameters (S) and the material damage variables in the definition of the plastic strain rate.

Through the relationship of FIG. 8, an inference equation of the relationship between the viscous parameter of the plastic strain rate and the damage coefficients S ₁ and Y of the material can be inferred, which can be expressed by Equations 3 and 4 below.

Table 1 shows the relationship between viscoplastic parameters and damage parameters.

S1 = exp (-1.152425223 × ln (x) -9.508897066) S2 = exp (-0.1980227915 × ln (x) +0.1685469124) Number of data points used = 45 Average ln (X) = -8.49781 Average ln (Y) = 0.284196 Residual sum of squares = 9.28923 Regression sum of squares = 47.8986 Coef of determination, R-squared = 0.837566 Residual mean square, sigma-hat -sq'd = 0.21602 Number of data points used = 37 Average ln (X) = -5.83534 Average ln (Y) = 1.32408 Residual sum of squares = 1.76709 Regression sum of squares = 3.44932 Coef of determination, R-squared = 0.661244 Residual mean square, sigma-hat -sq'd = 0.0504882 Alternate Y = pow (X, -1.152425223) * 7.418882197E-005 Alternate Y = pow (X, -0.1980227915) * 1.18358375

9 is a graph comparing empirical and actual inexperienced steels.

9, it can be seen that the parameters due to the plastic strain rate considering the damage effect in the mechanical properties of the inexperienced steel are in good agreement with the actual correlation.

On the other hand, after constructing the basic data of the material damage model, the step of evaluating the structural integrity of the steel proceeds.

10 is a flowchart illustrating a process of evaluating structural health.

Referring to FIG. 10, the procedure for slaughtering the final damage model (1000) includes assessing weld damage and strength characteristics of any structural member (step 1001) and soundness of material parameters through simulation in virtual space. Evaluating (step 1003).

In step 1001, in order to evaluate weld damage and strength characteristics, a model that can evaluate the fatigue strength of the weld in consideration of the weld residual stress is required.

FIG. 11 is a block diagram illustrating a structure for obtaining residual stress estimates capable of estimating characteristics due to defects at the beginning of a structure in order to consider all possible conditions in a welded structure in order to evaluate weld damage and strength characteristics of the structural member. to be.

Combination of strength characteristics of weld damage for each structural member by welding construction conditions, elasto-plastic model of residual stress by welding, residual stress at welded joint, residual stress using natural stress and idealized distribution By approximated equations, we estimate the characteristics due to defects at the beginning of the structure. In this case, the characteristic due to the defect of the initial stage of the structure is estimated through the structure as shown in FIG.

In the case of considering the defect characteristics at the beginning of the structure, as shown in FIG. 11, an elasto-plastic model estimation equation 1103, an idealized distribution approximation equation 1104, a welding construction condition estimation equation 1105, and a weld joint surface estimation equation ( 1106, and the intrinsic stress utilization estimation equation 1107 is used as the residual stress distribution process 1101.

The weld characteristics of the structural members include the heat input, the tendency of the initial defects of the structure requiring analysis, the geometrical classification of the structural members, the weld joints, the number of passes, etc. 1110 receives the weld property values of these structural members (1102), and can estimate the distribution of structural defects, that is, an estimation formula that takes into account the characteristics of the defects of these five structural initials, and calculates the derived equations. A residual stress distribution module can be established (1108). This is calculated as a parameter of important analysis of the fatigue strength characteristics evaluation of actual structural welds.

In step 1003, finite element modeling of a region of interest to evaluate the health through simulation in virtual space is a structure in which existing residual stresses are distributed during modeling in several existing commercial programs or systems currently under development. Approach the model with the initial initial defect. In this case, the region of interest can be modeled using existing commercial programs, Partran or Ansys. However, in the program according to the present invention, if a shape of a test specimen of an existing experimental standard is inputted, a module that automatically uses an existing mesh model without using a modeling process in an external preprocessor without a modeling process is mounted. It was. In addition, through the interface module on the system, it is possible to model the defects generated during welding by applying each external force to the zoom-in model while sub-modeling the actual structural problem. It can be evaluated quantitatively, and it is possible to define the inherent properties (breaks) of a material and perform structural analysis to finally evaluate the health of the material by calculating the damage amount of the material.

12 is a flowchart illustrating a process of calculating a final damage amount after an initial defect is quantitatively calculated.

Once the structural initial defect (ie residual stress) for the welding effect is calculated quantitatively, first, a material inherent fracture is defined using the foregoing (step 1201). Then, a structural analysis is performed (step 1203). The fatigue life assessment for any structure consists of three steps: First, the material constants in the structural response are obtained using memory-based reasoning. Secondly, under the given load, FEA is used to find the strain and stress of the structure. Finally, the damage and stress values in the stress concentration region are calculated through a predetermined damage generation equation.

Next, the fatigue characteristics evaluation is performed using the following analysis data as verification of the weld fatigue evaluation method. The test specimen was made to have the same shape as the fatigue test specimen, and stress concentration occurs depending on the shape of the weld bead, which greatly affects the fatigue life. Therefore, the average bead welding angle of 120 ° was considered by measuring the bead shape after fabrication of the test specimen. It was modeled as shown in FIG. We also compared the local strain approximation method through the elasto-plastic finite element analysis. First, the material properties for the three-dimensional analysis of the model are as follows. The analytical model uses solid 8node elements, and the model is divided into the base metal and the weld.

The load condition used in the analysis was the same as the experimental condition. The same tensile tensile load with R = 0 was used.If the maximum peripheral strain was the same, the finite element analysis was performed according to the maximum peripheral strain theory, which shows the same life under uniaxial and multiaxial conditions. The principal stress and the amount of change of the peripheral strain at the point representing the maximum peripheral strain at the time were obtained and fatigue life analysis was performed by applying Manson-Halford equations (Equations 5 and 6) and SWT equation (Equation 7).

FIG. 14 is a conceptual diagram illustrating basic structural analysis by inputting basic material properties, weld shape dimensions, and weld effect in order to predict weld fatigue level. FIG. 15 is a schematic table of fatigue life analysis at a point showing the maximum peripheral strain using the Manson-Halford equation and the SWT equation in this structural analysis.

16 is a method of applying a predictive model to the weld fatigue life applied in the system according to the present invention. It is a schematic diagram that predicts fatigue life of welds by calculating peripheral strain principal stresses in stress concentration and applying them to damage mechanic finite element codes in the same way as Manson-Halford and SWT equations.

A model as shown in FIG. 17 was set up to verify the fatigue evaluation of the weld using Manson-Halford equation, SWT equation and the present system. A finite element model was created for butt welding and the analysis was performed without considering the welding effects. Cyclic stress-strain characteristics are required for fatigue life prediction by local strain approximation, and cyclic stress-strain characteristics in Tso-Liang Teng's experiments are applied.

Cycle and Fatigue Properties σf (fatigue strength coefficient) 1014 εf (fatigue ductility coefficient) 0.271 b (fatigue strength exponent) -0.132 c (fatigue ductility exponent) -0.451 Kf (cyclic strength coefficient) 1097

Basquin-Mansontlr of Equation 8, which is widely used in the prediction of fatigue crack initiation life by local strain approximation, Morrow equation of Equation 9, Manson-Halford equation of Equation 10, and Equation 11 SWT (Smith, Watson, and Topper) can be applied to examine the suitability of life prediction.

18 is a graph showing the results of fatigue analysis applied to the Manson-Halfordtlr and SWT equations.

The load condition used in the analysis was the same as the experimental condition. The same tensile tensile load with R = 0 was used. The principal stress and the amount of change in peripheral strain of the contact, which represent the maximum peripheral strain at the time, were obtained and fatigue analysis was performed by applying the Manson-Halfordtlr and SWT equations as shown in FIG. Here, the value of local strain was used using ANSYS, a universal finite element code usable in the system according to the present invention.

Figure 19 shows the program fatigue evaluation results in order to compare the fatigue evaluation and damage-based evaluation techniques strain-based in Tso-Lianf Teng.

Although the present invention has been described with reference to one embodiment shown in the drawings, this is merely exemplary, and those skilled in the art will understand that various modifications and equivalent other embodiments are possible therefrom. . Therefore, the true technical protection scope of the present invention will be defined by the technical spirit of the appended claims.

According to the fatigue strength evaluation method according to the present invention, it is possible to compose the basic data by grasping the remaining mechanical properties required for the material damage model using a few standard mechanical properties, reducing the analysis time and the long time according to the system development It is possible to make prediction of fatigue life in a short time depending on the test.

Claims (8)

In the method for evaluating the fatigue strength of welded joints of steel, Receiving standard mechanical properties; Constructing basic data of a material damage model using the standard mechanical properties; And Comparing the basic data of the material damage model with the results of the uniaxial tensile test for the fatigue strength evaluation model of the welded portion to derive a final damage model, Comprising the basic data of the material damage model, Inputting the standard mechanical property values into an inference engine; Inferring material constants for the damage engine; Verifying material constants for material damage assessment; Determining a fatigue life analysis relationship of fatigue characteristic values using standard deviations; And Constructing basic data of the optimized material damage model using the inference model and the database Fatigue strength evaluation method of the welded joint of the steel, characterized in that it comprises a. The method of claim 1, Inferring a material constant for the damage engine, Obtaining an analysis material constant using known data; Defining an empirical formula based on a correlation between an experimental numerical value and a mechanical characteristic value and an analysis material constant in the known data; And Error-optimizing the relationship between the memory-based analysis material constants and the analysis material constants modified by empirical equations from the known data. Fatigue strength evaluation method of the welded joint of the steel, characterized in that it comprises a. The method of claim 2, The error optimization step is to optimize the error by using a genetic algorithm, the fatigue strength evaluation method of the welded joint of the steel, characterized in that to calculate the optimized analysis material constant. The method of claim 2, Obtaining the analysis material constants, Receiving a material constant of the experience steel from the experience steel database; Estimating a memory based steel material; And Determining Material Constants of Inexperienced Steels Fatigue strength evaluation method of the welded joint of the steel, characterized in that it comprises a. The method of claim 1, Deriving the final damage model, Evaluating weld damage and strength characteristics of any structural member; And Evaluating the health of material constants through simulation in virtual space Fatigue strength evaluation method of the welded joint of the steel, characterized in that it comprises a. The method of claim 5, Evaluating weld damage and strength characteristics of the structural member, The residual stress distribution is calculated by using the elasto-plastic model estimation equation, the idealized distribution approximation equation, the welding construction condition estimation equation, the weld joint surface estimation equation, and the intrinsic stress utilization estimation equation. Fatigue strength evaluation method of the welded joint of steel, characterized in that for establishing a residual stress distribution module based on the. The method of claim 5, Evaluating the integrity of the material constant through the simulation in the virtual space, quantitatively calculating the structural initial defects for the welding effect, defining the material inherent failure, and then performing a structural analysis to perform the final damage amount Fatigue strength evaluation method of the welded joint of the steel, characterized in that for calculating. A computer-readable medium suitable for a GUI (Graphic User Interface) environment equipped with an algorithm for evaluating the fatigue strength of welded joints of steel, The algorithm is Receiving standard mechanical properties; Constructing basic data of a material damage model using the standard mechanical properties; And Comparing the basic data of the material damage model with the results of the uniaxial tensile test for the fatigue strength evaluation model of the welded portion to derive a final damage model, Comprising the basic data of the material damage model, Inputting the standard mechanical property values into an inference engine; Inferring material constants for the damage engine; Verifying material constants for material damage assessment; Determining a fatigue life analysis relationship of fatigue characteristic values using standard deviations; And Constructing basic data of the optimized material damage model using the inference model and the database Computer-readable medium comprising a.

Images