CN115270615A - Method and system for predicting fatigue life of welding joint under multistage loading - Google Patents

Method and system for predicting fatigue life of welding joint under multistage loading Download PDF

Info

Publication number
CN115270615A
CN115270615A CN202210841674.9A CN202210841674A CN115270615A CN 115270615 A CN115270615 A CN 115270615A CN 202210841674 A CN202210841674 A CN 202210841674A CN 115270615 A CN115270615 A CN 115270615A
Authority
CN
China
Prior art keywords
whale
prediction model
fatigue
welding joint
loading
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210841674.9A
Other languages
Chinese (zh)
Inventor
邹丽
杨鑫华
杨一波
孙屹博
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Dalian Jiaotong University
Original Assignee
Dalian Jiaotong University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Dalian Jiaotong University filed Critical Dalian Jiaotong University
Priority to CN202210841674.9A priority Critical patent/CN115270615A/en
Publication of CN115270615A publication Critical patent/CN115270615A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/15—Vehicle, aircraft or watercraft design
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/17—Mechanical parametric or variational design
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Geometry (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Computer Hardware Design (AREA)
  • Medical Informatics (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Mathematical Analysis (AREA)
  • Computational Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Automation & Control Theory (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Investigating Strength Of Materials By Application Of Mechanical Stress (AREA)

Abstract

The invention discloses a method and a system for predicting the fatigue life of a welding joint under multistage loading, which are used for collecting fatigue data of the welding joint under multistage loading, establishing a fatigue analysis database of the welding joint, and establishing a training set and a test set; training the constructed support vector regression prediction model by using the training set, optimizing parameters of the prediction model based on a whale optimization algorithm in the training process to obtain the prediction model after parameter optimization, and testing the prediction model after parameter optimization by using the testing set; and predicting the fatigue life of the welding joint under the multilevel loading based on the obtained prediction model after the parameter optimization. The precision of the method is superior to that of a traditional accumulated damage model, and the fatigue life of the aluminum alloy welding joint can be better evaluated.

Description

Method and system for predicting fatigue life of welding joint under multistage loading
Technical Field
The invention relates to the technical field of welding fatigue analysis, in particular to a method and a system for predicting the fatigue life of a welding joint under multistage loading.
Background
In order to meet the development requirement of lightweight modern mechanical equipment, a welding structure is widely applied to actual engineering. During service, a welding structure is often subjected to the action of multi-level variable amplitude load, fatigue failure is easy to occur, the fatigue fracture mechanism is quite complex, the fatigue life of the welding structure is difficult to accurately estimate by using the traditional fatigue life estimation method, and great difficulty is caused to the design and maintenance of the structure.
The traditional fatigue life assessment method is represented by a Miner linear accumulated damage model, and has the advantages of simple form, easy calculation and wide application in engineering. Although the Miner's rule has the advantages of concise expression form, easy use and the like, the Miner's rule excessively simplifies fatigue failure mechanisms and criteria, so that the calculation accuracy is poor, and the actual requirements of engineering cannot be met.
Aiming at the defects of Miner's rule, related scholars put forward another mode of the damage accumulation theory, namely a nonlinear damage accumulation theory, by improving the damage evolution form and combining the fatigue failure basic mechanism. Richart and Newmark firstly put forward the concept of a damage curve, a damage-based nonlinear life assessment model is built based on Marco and Strakey, then a specific expression form of a damage index in the model is given by Manson and Halford, and the model is perfected based on an effective crack growth model. The Manson-Hall model (abbreviated as M-H model) gives a relation between material cracks and a cyclic ratio and a parameter calculation method based on a damage curve method. The M-H model can better consider the influence of load sequence effect on the service life, and the improved model-corrected M-H model can consider the influence of load interaction, but because the expression form is complex, the defects of low calculation efficiency, low engineering practicability and the like exist when the complex fatigue problem in the actual engineering is solved.
Because a great deal of randomness exists in the variable amplitude load process, the conventional accumulated damage model cannot consider the order effect among loads and the interaction effect among the loads, so that the prediction precision is low when the fatigue life is predicted. How to establish a high-precision fatigue life prediction and nonlinear accumulated damage model under the condition of a small sample is a bottleneck problem in the research of the conventional fatigue life prediction method.
Disclosure of Invention
Therefore, the invention provides a method and a system for predicting the fatigue life of a welding joint under multistage loading, and aims to solve the problems that the conventional fatigue life prediction method is low in prediction precision and cannot meet the actual requirements of engineering.
In order to achieve the above purpose, the invention provides the following technical scheme:
according to a first aspect of an embodiment of the present invention, a method for predicting fatigue life of a welding joint under multistage loading is provided, where the method includes:
collecting fatigue data of the welding joint under multi-stage loading, establishing a fatigue analysis database of the welding joint, and constructing a training set and a testing set;
training the constructed support vector regression prediction model by using the training set, optimizing parameters of the prediction model based on a whale optimization algorithm in the training process to obtain the prediction model after parameter optimization, and testing the prediction model after parameter optimization by using the testing set;
and predicting the fatigue life of the welding joint under the multilevel loading based on the obtained prediction model after the parameter optimization.
Further, establishing a welding joint fatigue analysis database, which specifically comprises:
fatigue test data of domestic and foreign welding fatigue analysis mechanisms and enterprises are collected and sorted, a laboratory fatigue test is carried out, welding joint fatigue analysis data under multistage loading are obtained, and a welding joint fatigue analysis database is constructed.
Further, the database comprises loading stress levels and stress mean values of all levels, the cycle times of the test piece under the action of the loading stress, and the residual fatigue life of the test piece under the action of the loading stress.
Further, constructing a training set and a test set, and specifically comprising:
and carrying out normalization processing on the sample data.
Further, training the constructed support vector regression prediction model by using the training set specifically comprises:
and taking fatigue parameter data comprising loading stress levels and stress mean values of all levels and cycle times of the test piece under the action of loading stress as input of a prediction model, taking the residual fatigue life of the test piece as expected output of the prediction model, and training the prediction model.
Further, optimizing the parameters of the prediction model based on a whale optimization algorithm in the training process to obtain the prediction model after parameter optimization, and specifically comprising the following steps:
setting a search range of SVR model parameters and whale optimization algorithm parameters, wherein the whale optimization algorithm parameters comprise the scale N of whale population, the dimension d of search space and the maximum iteration number tmax(ii) a Setting the initial iteration times t =1, initializing the population position by adopting a random method, and expressing the position of the ith whale in the d-dimensional space as
Figure BDA0003751339210000031
The optimal whale position, i.e. the prey position, represents a globally optimal solution to the problem;
calculating a Mean Square Error (MSE) according to a predicted value and a test value obtained in the training process, taking the MSE as a fitness function, calculating the fitness value of each whale, and selecting the whale position with the minimum fitness value as the current optimal position;
updating parameters a, A, C and l, wherein a is a convergence factor and is linearly reduced to 0 from 2 in the whale colony predation iterative process, and the calculation formula is a = 2-2T/TmaxIn which T is an element [ T, T ∈ [ ]max]Representing the current number of iterations, TmaxRepresenting the maximum number of iterations; a and C are coefficient vectors, implemented as a function of a, where A is [ -a, a [ -a [ ]]C is [0, a ]]The random values in (1) are used for balancing the global exploration and local development capability of the algorithm, and various optimal positions relative to the current position are realized by adjusting the values of A and CPosition, the formula is A =2. A. Rand1-a,C＝2·rand2Rand in the formula1,rand2Is [0,1 ]]A random number in between; l is a random number in the range of [ -1,1 [ ]]In the middle of; p is the probability of the algorithm selecting the attack or surrounding prey, randomly generated by the algorithm statement, ranging from 0,1]To (c) to (d);
if p is more than or equal to 0.5, attacking preys in a spiral motion mode according to a preset formula, and updating whale positions; if p is less than 0.5, judging whether the parameter | A | is less than 1, if so, performing shrinkage enclosure on the prey according to a preset formula, and updating the whale position, otherwise, performing global search according to the preset formula, and updating the whale position; at the moment, position updating is finished, the fitness value of the optimized whale population is calculated again and is compared with the previously reserved optimal position x*Making comparison if it is better than x*Then to x*Updating the position;
when the iteration times reach the maximum, finishing the optimization and outputting the currently optimized parameters, otherwise, iterating times t +1; and inputting the optimized parameters into the prediction model to obtain the optimized prediction model.
Further, the method further comprises:
if p is more than or equal to 0.5, according to the formula: x (t + 1) = (| x)*(t)-x(t)|)·ebl·cos(2πl)+x*(t) attacking prey in spiral motion mode to update whale position, wherein x (t + 1) is updated whale position, x*(t) represents the optimal whale position in the current whale group, x (t) represents the position of the current whale, | x*(t) -x (t) | represents the absolute value of the distance between the current whale position and the optimal whale position in the tth iteration, and b is a constant coefficient used for limiting the logarithmic spiral form;
if p < 0.5, then determine if the parameter | A | is less than 1, if yes, then according to the formula: x (t + 1) = x*(t)-A·(|C·x*(t) -x (t) |) performing shrinkage enclosure on the prey, and updating whale positions; otherwise, according to the formula: x (t + 1) = Xrand(t)-A·(|C·Xrand(t) -X (t) |) performing global search, updating whale position, wherein X israndIs carried out from the current fish schoolAnd selecting a whale individual position.
According to a second aspect of the embodiments of the present invention, there is provided a fatigue life prediction system for a weld joint under multi-stage loading, the system including:
the data set construction module is used for collecting fatigue data of the welding joint under the multi-stage loading, establishing a fatigue analysis database of the welding joint and constructing a training set and a test set;
the model training optimization module is used for training the constructed support vector regression prediction model by using the training set, optimizing parameters of the prediction model based on a whale optimization algorithm in the training process to obtain the prediction model after parameter optimization, and testing the prediction model after parameter optimization by using the test set;
and the prediction module is used for predicting the fatigue life of the welding joint under the multilevel loading based on the obtained parameter optimized prediction model.
Further, the model training optimization module is specifically configured to:
and taking fatigue parameter data comprising loading stress levels and stress mean values of all levels and cycle times of the test piece under the action of loading stress as input of a prediction model, taking the residual fatigue life of the test piece as expected output of the prediction model, and training the prediction model.
According to a third aspect of embodiments of the present invention, there is provided a computer storage medium having one or more program instructions embodied therein for executing a method as described in any one of the above by a multi-stage load-down weld joint fatigue life prediction system.
The invention has the following advantages:
the invention provides a method and a system for predicting the fatigue life of a welding joint under multistage loading, which are used for collecting fatigue data of the welding joint under multistage loading, establishing a fatigue analysis database of the welding joint, and establishing a training set and a test set; training the constructed support vector regression machine prediction model by using the training set, optimizing parameters of the prediction model based on whale optimization algorithm in the training process to obtain the prediction model with optimized parameters, and testing the prediction model with optimized parameters by using the test set; and predicting the fatigue life of the welding joint under the multilevel loading based on the obtained prediction model after the parameter optimization. The precision of the method is superior to that of the traditional accumulated damage model, and the fatigue life of the aluminum alloy welding joint can be better evaluated.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below. It should be apparent that the drawings in the following description are merely exemplary and that other implementation drawings may be derived from the provided drawings by those of ordinary skill in the art without inventive effort.
Fig. 1 is a schematic flowchart of a method for predicting fatigue life of a welding joint under multistage loading according to embodiment 1 of the present invention;
fig. 2 is a schematic flowchart of an implementation of a method for predicting fatigue life of a welding joint under multistage loading according to embodiment 1 of the present invention;
fig. 3 is a schematic diagram of a fitness curve in the method for predicting the fatigue life of a welding joint under multi-level loading according to embodiment 1 of the present invention;
fig. 4 is a mean value of fatigue life prediction errors of different models in the method for predicting the fatigue life of a welding joint under multistage loading according to embodiment 1 of the present invention.
Detailed Description
The present invention is described in terms of particular embodiments, other advantages and features of the invention will become apparent to those skilled in the art from the following disclosure, and it is to be understood that the described embodiments are merely exemplary of the invention and that it is not intended to limit the invention to the particular embodiments disclosed. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example 1
Welded structures in actual engineering are often subjected to multi-amplitude load during the working process of the welded structures. Because a great deal of randomness exists in the variable amplitude load process, the conventional cumulative damage model cannot consider the order effect among the loads and the interaction effect among the loads, so that the prediction precision is low when the fatigue life is predicted. How to establish a high-precision fatigue life prediction and nonlinear accumulated damage model under the condition of a small sample is a bottleneck problem in the research of the conventional fatigue life prediction method.
A Support Vector Regression (SVR) is based on VC dimension and structure risk minimization principle of statistical learning theory, and converts linear inseparable problem into linear problem by means of kernel function, thereby overcoming dimension disaster and over learning. The method has unique advantages when dealing with the interaction problem of the machine learning problem and the nonlinear characteristic under a small sample. When the SVR model is used for regression prediction, the punishment parameters and the kernel function parameters are influenced mutually, and the complexity and the generalization capability of the SVR model are jointly determined. The Whale Optimization Algorithm (WOA) is a global search algorithm deduced from the unique bubble net trapping behavior of the Whale in nature, and has the characteristics of few parameters, easiness in implementation and strong convergence. Compared with similar algorithms such as a particle swarm optimization algorithm, a gravity search algorithm and the like, the WOA algorithm has strong competitiveness and shows good performance in a nonlinear parameter optimization problem. The global search optimization of the penalty parameter and the kernel function parameter in the support vector machine model can be realized.
Aiming at the problems of the traditional accumulated damage model and the bottleneck problem in the research of the conventional fatigue life prediction method, the whale algorithm-based method for predicting the fatigue life of the welding joint under the condition of multi-stage loading of the optimized support vector machine is provided.
As shown in fig. 1, the present embodiment provides a method for predicting fatigue life of a welding joint under multistage loading, where the method includes the following steps:
s100, collecting fatigue data of the welding joint under multi-stage loading, establishing a fatigue analysis database of the welding joint, and constructing a training set and a test set;
s200, training the constructed support vector regression machine prediction model by using the training set, optimizing parameters of the prediction model based on whale optimization algorithm in the training process to obtain the prediction model with optimized parameters, and testing the prediction model with optimized parameters by using the test set;
s300, predicting the fatigue life of the welding joint under the multilevel loading based on the obtained parameter optimized prediction model.
As shown in fig. 2, the specific implementation process is as follows:
step1: the fatigue test data of domestic and foreign welding fatigue analysis mechanisms and enterprises are collected and sorted, a laboratory fatigue test is carried out, the data test data of the welding joint under the multi-stage loading are obtained, and a welding joint fatigue analysis database is constructed. For convenience of explanation, the fatigue analysis under two-stage loading is taken as an example in the following. The data attributes in the fatigue analysis database include: sigma1，σm1，R1，σ2，σm2，R2，n1，n2. Wherein σ1For first order loading of stress level, σm1Mean value of first stage loading stress, R1Is the first order load stress ratio, σ2For the second stage loading stress level, σm2Mean value of second stage loading stress, R2For second order loading stress ratio, n1For the test piece at σ1Number of cycles under action, n2For the test piece at sigma2Residual fatigue life under action.
Step2: and taking fatigue test samples in the welding joint fatigue analysis database as data samples for training the SVR, dividing a training set and a testing set, and carrying out normalization processing on the data. The calculation formula of the normalization process is as follows:
xnew＝(x-xmax)/(xmax-xmin)
wherein x is the sample data value before normalization, xmax，xminMaximum and minimum values, x, of the fatigue test sample on a certain data attributenewIs unified toThe transformed sample data values.
Step3: analysing σ in a database by fatigue1，σm1，R1，σ2，σm2，R2，n1Equal attribute value as input of SVR model, with n2And the attribute value is expected output of the SVR model, and the SVR model is trained.
Step4: in the SVR model training process, optimizing SVR model parameters (c, g) based on a WOA algorithm to obtain a parameter optimized SVR model, wherein a kernel function parameter g and a punishment parameter c. This model implements the slave sigma1，σm1，R1，σ2，σm2，R2，n1To n2pNon-linear mapping of, i.e.
n2p＝f(σ1,σm1,R1,σ2,σm2,R2,n1)
Wherein n is2pFor the test piece at sigma2Residual fatigue life prediction under action.
Wherein, the detailed steps of Step4 comprise:
step 4.1: setting the search range of the SVR model parameters and the WOA algorithm parameters, including the scale N of the whale population, the dimension d of the search space and the maximum iteration number tmax. Setting the initial iteration times t =1, initializing the position of the population by adopting a random method, and expressing the position of the ith whale in the d-dimensional space as
Figure BDA0003751339210000081
The optimal whale position (the position of prey) represents the globally optimal solution to the problem.
Step4.2: and calculating the Mean Square Error (MSE) according to the predicted value and the test value obtained in the training process, taking the MSE as a fitness function, and calculating the fitness value of each whale by using a calculation formula shown in the specification. Selecting whale position with minimum fitness value as current optimal position
Figure BDA0003751339210000082
Figure BDA0003751339210000083
In the formula (I), the compound is shown in the specification,
Figure BDA0003751339210000084
is the predicted value of the ith sample, yiThe test value of the ith sample is obtained.
Step4.3: updating parameters a, A, C and l, wherein a is a convergence factor and is linearly reduced to 0 from 2 in the whale colony predation iterative process, and the calculation formula is a = 2-2T/TmaxIn which T is an element [ T, T ∈ [ ]max]Representing the current number of iterations, TmaxThe maximum number of iterations is indicated. A and C are coefficient vectors, implemented as a function of a, where A is [ -a, a [ -a [ ]]C is [0, a ]]The random values in (1) are used for balancing the global exploration and local development capability of the algorithm, various positions of the optimal position relative to the current position can be realized by adjusting the values of A and C, and the calculation formula is A =2. A. Rand1-a,C＝2·rand2Rand in the formula1,rand2Is [0,1 ]]A random number in between. l is a random number in the range of [ -1,1 [ ]]In between. p is the probability of an algorithmic selection attack or enclosure on a prey, randomly generated by an algorithmic statement, ranging from [0,1 ]]In between.
If p is less than 0.5, then the process goes to Step4.4. Otherwise, according to the formula: x (t + 1) = (| x)*(t)-x(t)|)·ebl·cos(2πl)+x*(t) attacking prey in spiral motion mode to update whale position, wherein x (t + 1) is updated whale position, x*(t) represents the optimal whale position in the current whale group, x (t) represents the position of the current whale, | x*(t) -x (t) | represents the absolute value of the distance between the current whale position and the optimal whale position in the t iteration, and b is a constant coefficient used for limiting the logarithmic spiral form and takes the value of 1.
Step4.4: judging whether the parameter | A | is less than 1, if so, according to the formula: x (t + 1) = x*(t)-A·(|C·x*(t) -x (t) |) shrink-wrap the prey and update the whale position. Otherwise, according to the formula: x (t + 1) =Xrand(t)-A·(|C·Xrand(t) -X (t) |) performing global search, updating whale position, wherein X israndIs a whale individual position randomly selected from the current fish population.
Step4.5: at the moment, position updating is finished, the fitness value of the optimized whale population is calculated again and is compared with the previously reserved optimal position x*Making comparison if x is better*Then to x*The location is updated.
Step4.6: and when the iteration times reach the maximum, finishing optimization, outputting the currently optimized parameters, otherwise, t = t +1, and returning to Step4.2.
Step4.7: and inputting the optimized parameters into the SVR model to obtain the optimized SVR model.
Step5: known as σ1，σm1，R1，σ2，σm2，R2，n1Under the condition of welding joint fatigue test parameters, the prediction of the fatigue residual life of the welding joint under the level of the second-stage loading stress can be realized based on the SVR model after parameter optimization.
Experimental example:
and constructing a welding fatigue analysis database according to the collected and sorted fatigue test data of the high-speed motor train unit aluminum alloy vehicle body welding joint under the two-stage loading, and performing model experiment and analysis. The parent metal of the welding joint test piece is CRH2 type high-speed motor train unit aluminum alloy vehicle body material ENAW6005, the joint form comprises a butt joint form and an angle joint form, and the fatigue testing machine is a PLG-200 type high-frequency fatigue testing machine. The test adopts a high-low and low-high loading mode. The fatigue life of the butt joint under the stress levels of 104MPa, 89MPa and 74MPa is 549300 times, 880500 times and 1540100 times respectively. The fatigue life of the angle joint under the action of the stress level 93Mpa, 83Mpa and 73Mpa is 619800, 952300 and 1546100 times respectively.
Some of the data contained in the constructed weld joint fatigue analysis database are shown in table 1 below:
TABLE 1 weld joint fatigue analysis database
Figure BDA0003751339210000091
Setting the kernel function parameter g E of the SVR model to [0.01,100']And penalty parameter c is equal to 0.01,100]Maximum number of iterations Tm=50, the maximum number of populations is N =30. On the basis of a weld joint fatigue analysis database after normalization processing, a kernel function parameter g and a penalty parameter c of the SVR model are trained and optimized by using a WOA algorithm (the specific steps are shown in Step4.1-Step4.6). The fitness curve based on the fatigue test data of the butt joint and the angle joint obtained in the training process is shown in fig. 3. After training is finished, the optimal kernel function parameter g =2.15 and the optimal penalty parameter c =27.4043 of the butt joint SVR model are obtained. The optimal kernel function parameter g =1.3068 and the optimal penalty parameter c =7.6782 of the corner joint SVR model.
The SVR model after parameter optimization can be used for predicting the fatigue life of the diagonal joint and the butt joint under two-stage loading: will σ1，σm1，R1，σ2，σm2，R2，n1Inputting the fatigue test parameters of the aluminum alloy butt joint and angle joint welding joints into the SVR model after parameter optimization, predicting the fatigue life of the welding joints under the level of the second-stage loading stress, and recording the obtained predicted value as n2p。
In order to verify the prediction effect of the WOA-SVR model, the real fatigue life of the welding joint under the level of the second-stage loading stress in the fatigue test process is recorded as n2. Obtaining a fatigue life prediction value n of a fatigue test sample obtained by a WOA-SVR model under a second-stage stress level under two load modes of high-low and low-high corresponding to 8 groups of butt joint and angle joint welding joints2pAnd true test value n2The control ratios are shown in table 2. In the table, the calculation formulas of the fatigue life prediction error and the error percentage are respectively as follows:
error = | n2-n2p|
Percent error = | n2-n2p|/n2×100％.
TABLE 2 prediction results of fatigue life of WOA-SVR model under two-stage loading
Figure BDA0003751339210000101
The cumulative damage obtained by the WOA-SVR prediction model is compared with Miner, M-H and corrected M-H models, the results are shown in Table 3, fatigue life prediction is performed by using different models, and the cumulative damage error average value is shown in FIG. 4. Wherein, the calculation formula of the accumulated damage error percentage is as follows:
Figure BDA0003751339210000111
TABLE 3 comparison of cumulative damage results for different models under two-stage loading
Figure BDA0003751339210000112
As can be seen from Table 3 and FIG. 4, for the aluminum alloy butt joint under two-stage loading, the average errors of fatigue life estimation by adopting a Miner model, an M-H model and an improved M-H model are 31.98%, 14.05 and 14.22%, and the average error of the proposed WOA-SVR model is 0.62%, which is far lower than that of the traditional cumulative damage model. For the aluminum alloy angle joint under two-stage loading, the error of fatigue life prediction of the Miner model is the largest, the average error is 20.56%, the prediction results of the M-H model and the corrected M-H model are more in accordance with test data, the average errors are 2.58% and 2.47%, the prediction average error of the model provided by the invention is 0.58%, and the prediction precision is further improved. Therefore, through fatigue test analysis of aluminum alloy welding joints of different joint types under two-stage loads, the prediction result of the WOA-SVR model is more consistent with the test result, the accuracy of the WOA-SVR model is superior to that of a traditional Miner model, an M-H model and a corrected M-H model, and the fatigue life of the aluminum alloy welding joint can be better estimated.
Example 2
In correspondence with embodiment 1 described above, this embodiment proposes a fatigue life prediction system for a weld joint under multistage loading, the system including:
the data set construction module is used for collecting fatigue data of the welding joint under multi-stage loading, establishing a fatigue analysis database of the welding joint, and constructing a training set and a test set;
the model training optimization module is used for training the constructed support vector regression prediction model by using the training set, optimizing parameters of the prediction model based on a whale optimization algorithm in the training process to obtain the prediction model after parameter optimization, and testing the prediction model after parameter optimization by using the test set;
and the prediction module is used for predicting the fatigue life of the welding joint under the multilevel loading based on the obtained parameter optimized prediction model.
Further, the model training optimization module is specifically configured to:
and taking fatigue parameter data comprising loading stress levels and stress mean values of all levels and cycle times of the test piece under the action of the loading stress as input of a prediction model, taking the residual fatigue life of the test piece as expected output of the prediction model, and training the prediction model.
The functions performed by each component in the system for predicting the fatigue life of a welding joint under multistage loading according to the embodiment of the present invention are described in detail in embodiment 1, and therefore will not be described herein again.
Example 3
In accordance with the above embodiments, the present embodiments provide a computer storage medium having one or more program instructions embodied therein for executing the method of embodiment 1 by a multi-stage, under-load weld joint fatigue life prediction system.
Although the invention has been described in detail with respect to the general description and the specific embodiments, it will be apparent to those skilled in the art that modifications and improvements may be made based on the invention. Accordingly, such modifications and improvements are intended to be within the scope of the invention as claimed.

Claims (10)

1. A method for predicting fatigue life of a welding joint under multistage loading, which is characterized by comprising the following steps:
collecting fatigue data of the welding joint under multi-stage loading, establishing a fatigue analysis database of the welding joint, and constructing a training set and a testing set;
training the constructed support vector regression machine prediction model by using the training set, optimizing parameters of the prediction model based on whale optimization algorithm in the training process to obtain the prediction model with optimized parameters, and testing the prediction model with optimized parameters by using the test set;
and predicting the fatigue life of the welding joint under the multilevel loading based on the obtained prediction model after the parameter optimization.
2. The method for predicting the fatigue life of the welding joint under the condition of multistage loading according to claim 1, wherein the step of establishing a welding joint fatigue analysis database specifically comprises the following steps:
fatigue test data of domestic and foreign welding fatigue analysis mechanisms and enterprises are collected and sorted, a laboratory fatigue test is carried out, fatigue analysis data of the welding joint under multistage loading are obtained, and a welding joint fatigue analysis database is constructed.
3. The method for predicting the fatigue life of the welding joint under the multi-stage loading according to claim 1, wherein the database comprises loading stress levels and stress mean values of all stages, the cycle number of the test piece under the action of the loading stress, and the residual fatigue life of the test piece under the action of the loading stress.
4. The method for predicting the fatigue life of the welding joint under the condition of multi-stage loading according to claim 1, wherein a training set and a testing set are constructed, and the method specifically comprises the following steps:
and carrying out normalization processing on the sample data.
5. The method for predicting the fatigue life of the welding joint under the condition of multistage loading according to claim 1, wherein the training set is used for training the constructed support vector regression prediction model, and specifically comprises the following steps:
and taking fatigue parameter data comprising loading stress levels and stress mean values of all levels and cycle times of the test piece under the action of the loading stress as input of a prediction model, taking the residual fatigue life of the test piece as expected output of the prediction model, and training the prediction model.
6. The method for predicting the fatigue life of the welding joint under the multi-stage loading according to claim 1, wherein parameters of a prediction model are optimized based on a whale optimization algorithm in a training process to obtain the prediction model with the optimized parameters, and the method specifically comprises the following steps:
setting a search range of SVR model parameters and whale optimization algorithm parameters, wherein the whale optimization algorithm parameters comprise the scale N of whale population, the dimension d of search space and the maximum iteration number tmax(ii) a Setting the initial iteration times t =1, initializing the population position by adopting a random method, and expressing the position of the ith whale in the d-dimensional space as
Figure FDA0003751339200000021
The optimal whale position, i.e. the prey position, represents a globally optimal solution to the problem;
calculating a Mean Square Error (MSE) according to a predicted value and a test value obtained in the training process, taking the MSE as a fitness function, calculating the fitness value of each whale, and selecting the whale position with the minimum fitness value as the current optimal position;
updating parameters a, A, C and l, wherein a is a convergence factor and is linearly reduced to 0 from 2 in the whale colony predation iterative process, and the calculation formula is a = 2-2T/TmaxWhere T is an element [ T, T ∈ ]max]Representing the current number of iterations, TmaxRepresenting the maximum number of iterations; a and C are coefficient vectors, implemented as a function of a, where A is [ -a, a [ -a [ ]]C is [0, a ]]The two are used for the global exploration sum of the balance algorithmThe local development ability is realized by adjusting the values of A and C to realize various positions of the optimal position relative to the current position, and the calculation formula is A =2. A · rand1-a,C＝2·rand2Rand in the formula1,rand2Is [0,1 ]]A random number in between; l is a random number in the range of [ -1,1 [ ]]To (c) to (d); p is the probability of an algorithmic selection attack or enclosure on a prey, randomly generated by an algorithmic statement, ranging from [0,1 ]]To (c) to (d);
if p is more than or equal to 0.5, attacking prey in a spiral motion mode according to a preset formula, and updating the whale position; if p is less than 0.5, judging whether the parameter | A | is less than 1, if so, performing shrinkage enclosure on the prey according to a preset formula, and updating the whale position, otherwise, performing global search according to the preset formula, and updating the whale position; at the moment, position updating is finished, the fitness value of the optimized whale population is calculated again and is compared with the previously reserved optimal position x*Making comparison if x is better*Then to x*Updating the position;
when the iteration times reach the maximum, finishing the optimization and outputting the currently optimized parameters, otherwise, iterating times t +1; and inputting the optimized parameters into the prediction model to obtain the optimized prediction model.
7. The method for predicting the fatigue life of the welding joint under the multistage loading according to claim 1, further comprising:
if p is more than or equal to 0.5, according to the formula: x (t + 1) = (| x)*(t)-x(t)|)·ebl·cos(2πl)+x*(t) attacking prey in spiral motion mode to update whale position, wherein x (t + 1) is updated whale position, x*(t) represents the optimal whale position in the current whale group, x (t) represents the position of the current whale, | x*(t) -x (t) | represents the absolute value of the distance between the current whale position and the optimal whale position in the tth iteration, and b is a constant coefficient used for limiting the logarithmic spiral form;
if p < 0.5, then determine if the parameter | A | is less than 1, if yes, then according to the formula: x (t + 1) = x*(t)-A·(|C·x*(t)-x(t) |) to contract and enclose prey and update whale position; otherwise, according to the formula: x (t + 1) = Xrand(t)-A·(|C·Xrand(t) -X (t) |) performing global search, updating whale position, wherein X israndIs a whale individual position randomly selected from the current fish population.
8. A weld joint fatigue life prediction system under multi-level loading, the system comprising:
the data set construction module is used for collecting fatigue data of the welding joint under the multi-stage loading, establishing a fatigue analysis database of the welding joint and constructing a training set and a test set;
the model training optimization module is used for training the constructed support vector regression prediction model by using the training set, optimizing parameters of the prediction model based on a whale optimization algorithm in the training process to obtain the prediction model after parameter optimization, and testing the prediction model after parameter optimization by using the test set;
and the prediction module is used for predicting the fatigue life of the welding joint under the multilevel loading based on the obtained parameter optimized prediction model.
9. The system of claim 8, wherein the model training optimization module is specifically configured to:
and taking fatigue parameter data comprising loading stress levels and stress mean values of all levels and cycle times of the test piece under the action of loading stress as input of a prediction model, taking the residual fatigue life of the test piece as expected output of the prediction model, and training the prediction model.
10. A computer storage medium comprising one or more program instructions embodied therein for executing the method of any one of claims 1-7 by a multi-stage under load weld joint fatigue life prediction system.
CN202210841674.9A 2022-07-18 2022-07-18 Method and system for predicting fatigue life of welding joint under multistage loading Pending CN115270615A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210841674.9A CN115270615A (en) 2022-07-18 2022-07-18 Method and system for predicting fatigue life of welding joint under multistage loading

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210841674.9A CN115270615A (en) 2022-07-18 2022-07-18 Method and system for predicting fatigue life of welding joint under multistage loading

Publications (1)

Publication Number Publication Date
CN115270615A true CN115270615A (en) 2022-11-01

Family

ID=83767200

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210841674.9A Pending CN115270615A (en) 2022-07-18 2022-07-18 Method and system for predicting fatigue life of welding joint under multistage loading

Country Status (1)

Country Link
CN (1) CN115270615A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117688848A (en) * 2024-02-02 2024-03-12 湖南璟德科技有限公司 High-power stator vane coil fatigue life testing method

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117688848A (en) * 2024-02-02 2024-03-12 湖南璟德科技有限公司 High-power stator vane coil fatigue life testing method
CN117688848B (en) * 2024-02-02 2024-04-16 湖南璟德科技有限公司 High-power stator vane coil fatigue life testing method

Similar Documents

Publication Publication Date Title
CN110659722A (en) AdaBoost-CBP neural network-based electric vehicle lithium ion battery health state estimation method
CN112700060B (en) Station terminal load prediction method and prediction device
CN112101530A (en) Neural network training method, device, equipment and storage medium
CN110504676B (en) Power distribution network state estimation method based on APSO-BP
CN110874502B (en) Aerospace product reliability assessment method based on multi-stage test data reduction
CN110989342B (en) Real-time T-S fuzzy modeling method for combined cycle unit heavy-duty gas turbine
CN112597610B (en) Optimization method, device and equipment for lightweight design of mechanical arm structure
CN115270615A (en) Method and system for predicting fatigue life of welding joint under multistage loading
CN110110380B (en) Piezoelectric actuator hysteresis nonlinear modeling method and application
CN113469426A (en) Photovoltaic output power prediction method and system based on improved BP neural network
CN111931983A (en) Precipitation prediction method and system
CN114781692A (en) Short-term power load prediction method and device and electronic equipment
Chen et al. Grey Signal Predictor and Evolved Control for Practical Nonlinear Mechanical Systems.
CN115796358A (en) Carbon emission prediction method and terminal
CN111832911A (en) Underwater combat effectiveness evaluation method based on neural network algorithm
CN114722655A (en) Structural topology optimization method based on local limited life fatigue constraint condition
CN113868765A (en) Ship main scale parameter optimization method based on approximate model
CN110135085B (en) Fatigue crack evaluation method based on adaptive kernel density estimation auxiliary particle filter
CN110210072B (en) Method for solving high-dimensional optimization problem based on approximate model and differential evolution algorithm
CN111369072A (en) Nuclear minimum mean square time sequence online prediction model based on sparsification method
CN116227674A (en) Carbon emission prediction method based on online updating model
CN116230124A (en) High-temperature alloy performance prediction method based on ANN and UMAT coupling micro-tissue damage
CN114880792A (en) Deformation prediction-based omnibearing multi-angle optimization method
CN118094250B (en) Attention information enhanced radial basis function optimization design method
CN112016222B (en) Modal optimization method based on assembly finite element model and orthogonal test method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination