Rim welding process optimization method based on NSGA II algorithm and BP neural network
By combining NSGAⅡ algorithm and BP neural network, the rim welding process parameters are optimized, and the problem of high cracking rate of rim welds is solved, achieving high quality and efficient production of welded joints.
Patent Information
- Application Number
- CN202411566076.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-05
- Publication Date
- 2025-02-07
AI Technical Summary
After flash butt welding, the welds are prone to cracking defects, resulting in low welding quality and cracking rate as high as 27%, affecting production efficiency and cost.
The rim welding process optimization method based on NSGAⅡ algorithm and BP neural network is adopted to optimize welding process parameters through multi-objective optimization, establish a nonlinear mapping model between welding parameters and weld quality, and optimize the elongation and impact work of welds, thereby Improve the toughness and strength of welded joints.
It significantly improves the quality and performance of the welded joints, reduces the cracking rate of welds, solves the welding quality problems in the existing technology, and has the ability to adjust welding parameters in real time, improving production efficiency.
Smart Images
Figure CN119387784A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of welding process optimization, and in particular to a rim welding process optimization method based on NSGAⅡ algorithm and BP neural network. Background Art
[0002] Flash butt welding is a resistance welding method that uses the resistance heat generated by the end face of the workpiece to quickly melt the metal material, forming liquid metal on the butt surface. Pressure is applied through the tooling electrode. During the upsetting stage, oxides and impurities are squeezed out along with the liquid metal, and the solid metal contacts to form a welded joint. The schematic diagram and equipment are as follows: Figure 1 As shown, it belongs to the solid phase welding method. The flash process is the resistance heating process of the workpiece. The butt end faces are quickly heated to melt, and the lintel explodes and splashes out in the form of liquid, forming a flash. At the same time, the butt end faces of the weldment approach each other at a certain speed. At the end of the flash stage, the butt end faces quickly merge under the action of the upsetting force, and the flash is extinguished instantly. The welding transitions to the upsetting stage, extruding the overheated metal and oxide impurities. In the recrystallization process, common grains are formed between the two joints, and the welding is completed. However, the rim is prone to cracking defects in the weld after the expansion process after flash welding.
[0003] As an important structural part in the field of engineering vehicles and their parts manufacturing, the quality of the rim has a great impact on the safety of the vehicle during driving. Although its application in the field of manufacturing and forming has been continuously matured, due to its large size and complex welds, the residual stress distribution after welding is complex, the gradient is large, and it is difficult to eliminate, especially causing the phenomenon of flaring cracking. In actual production, the cracking rate of rim welds is as high as 27%, which has caused a decrease in production efficiency and waste of costs for enterprises. Therefore, how to avoid cracking of rim welds has become a problem that urgently needs to be solved. Summary of the invention
[0004] In view of the many problems existing in the above-mentioned prior art, the present invention provides a wheel rim welding process optimization method based on NSGAⅡ algorithm and BP neural network. The present invention combines BP neural network with NSGA-II algorithm, and performs multi-objective optimization on welding process parameters to establish a mapping relationship between welding parameters and weld quality, and finally realizes the production of high-quality welded joints. By optimizing welding process parameters, the toughness and strength of welded joints can be significantly improved.
[0005] like Fig.11 As shown, a rim welding process optimization method based on NSGAⅡ algorithm and BP neural network includes the following steps:
[0006] Collect process parameter data during welding, and pre-process and standardize the collected process parameter data;
[0007] A nonlinear mapping model between welding process parameters and weld quality is established using BP neural network, and the key quality response in the welding process is obtained through training;
[0008] The NSGA-II algorithm is used for multi-objective optimization, and the elongation and impact energy of the weld are set as optimization targets to obtain the optimal combination of welding process parameters.
[0009] The BP neural network model is combined with the NSGA-II algorithm in Matlab software, and the optimization results of welding process parameters are obtained by setting the algorithm and adjusting the parameters;
[0010] The optimized welding process parameters are experimentally verified, and the mechanical properties of the welded joints are evaluated through tensile tests and impact tests to ensure that the weld quality meets the predetermined standards.
[0011] Preferably, the process parameter data specifically include a burning distance of 12 to 18 mm, a burning speed of 0.8 to 1.2 mm per second, an upsetting distance of 8 to 16 mm and an electrified upsetting time of 0.2 to 0.6 seconds.
[0012] Preferably, the BP neural network model consists of at least three hidden layers and is trained using the Levenberg-Marquardt algorithm.
[0013] Preferably, the optimization objectives of the NSGA-II algorithm include the elongation of the weld, the elongation target of the weld is set between 20% and 30%, the impact energy target is set between 8 and 25 joules per square centimeter, and is evaluated by a Pareto front solution.
[0014] Preferably, the algorithm settings in the Matlab software include setting the population size to 200, the generation limit to 500 generations, and the algorithm convergence tolerance to 1e-100.
[0015] Preferably, the tensile test uses a standardized specimen having a size of 20 mm wide and 100 mm long, a test speed of 5 mm per minute, and is measured by a universal tensile testing machine.
[0016] Preferably, in the experimental verification, a metallographic analysis is performed on the microstructure of the weld, and the metallographic analysis is performed using a metallographic microscope to analyze the grain morphology and distribution of the weld area.
[0017] Preferably, the key quality response in the welding process includes the hardness of the weld, which is measured by a Rockwell hardness tester, and the toughness of the weld is obtained by an impact test.
[0018] Compared with the prior art, the advantages and beneficial effects of the present invention are:
[0019] The present invention achieves accurate optimization of rim welding process parameters by combining NSGA-II algorithm and BP neural network technology, significantly improving the quality and performance of welded joints. Specifically, the BP neural network is used to establish a nonlinear mapping relationship between welding parameters and weld quality, and the NSGA-II algorithm is used for multi-objective optimization, so that the elongation and impact energy of the weld meet the predetermined standards, effectively reducing the weld cracking rate and solving the welding quality problems in the prior art. In addition, the present invention has the ability to adjust welding parameters in real time, further improving production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] Figure 1 This is the principle diagram of flash butt welding;
[0021] Figure 2a-2b Schematic diagram of sampling of local cracking of weld in an embodiment of the present invention;
[0022] Figure 3a-3d They are respectively schematic diagrams of the fracture morphology of A1, A2, A3 and A4 under a high-power microscope in the embodiments of the present invention;
[0023] Figure 4a-4b Schematic diagram of the metallographic structure (100X) of the parent material in the embodiment of the present invention;
[0024] Figure 5a-5b Schematic diagram of the metallographic structure (100X) of the center of the weld in the embodiment of the present invention;
[0025] Figure 6a-6b Schematic diagram of the metallographic structure (100X) of the heat affected zone in an embodiment of the present invention;
[0026] Figure 7 Schematic diagram of a tensile specimen in an embodiment of the present invention;
[0027] Figure 8 This is a schematic diagram of an impact specimen in an embodiment of the present invention;
[0028] Fig. 9 This is a schematic diagram of a BP neural network model in an embodiment of the present invention;
[0029] Figure 10a-10b are respectively the test value and the predicted value of the elongation and the impact energy in the embodiment of the present invention; Fig.11 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION
[0030] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is obvious that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.
[0031] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.
[0032] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification, and should not be interpreted in an idealized or overly rigid manner.
[0033] In a specific embodiment, in order to study the cracking phenomenon, samples were taken at the crack. Figure 2a The sample was cut by wire cutting at the fracture position shown in the figure and cleaned with ultrasonic equipment. The morphology under high magnification microscope is shown in the figure Figure 3a and Figure 3b As shown. From a macroscopic perspective, there is almost no plastic deformation at the fractures of A1, A2, and A3, and they show metallic luster. Under a high-power microscope, they show cleavage steps and fan-shaped patterns. The cracks extend between different crystal planes to form cleavage steps. As the cracks extend, the cleavage steps gradually merge to form river patterns. The fan-shaped patterns originate from the expansion of the crystal from the inside to the outside to form a fan shape. This is related to the brittle inclusions on the grain boundaries. Figure 3d The A4 fracture at the end of the crack produces slight plastic deformation and is in a torn state. In addition to the cleavage steps and granular inclusions, there are also banded dimples, which are composed of a small number of equiaxed dimples and torn dimples, which are ductile fracture characteristics. The A4 fracture shows a brittle-to-ductile transition trend. Related studies have shown that the coarse grain size of the weld and the high dislocation accumulation during the plastic deformation of the expansion process can easily cause shear force and reduce the toughness of the welded joint. In addition, there is a linear relationship between the brittle-to-ductile transition temperature and the grain size of the material, and the increase in grain size often induces a brittle-to-ductile transition.
[0034] The sample was observed using a metallographic microscope. Figure 4a-4b It can be seen that the matrix structure is mainly composed of ferrite + pearlite, and the pearlite segregation is banded. Figure 5a-5bThere are black oxides in the center of the weld and severe Widmanstatten ferrite structure. The Widmanstatten ferrite seriously reduces the toughness of the butt weld, causing the rim to break under the force of expansion. Figure 6a-6b This is the metallographic structure distribution diagram of the weld heat affected zone. It can be seen that the coarse grain area is Widmanstatten structure, the fine grain area is ferrite + pearlite, and the pearlite segregation is banded. Therefore, Widmanstatten ferrite and coarse grain area mainly constitute the metallographic characteristics of the weld joint area, resulting in reduced toughness, which is also one of the reasons for the fracture in the rim expansion process.
[0035] In order to optimize the welding related parameters, tensile and shock wave tests were designed. In order to study the influence of rim flash welding parameters on the plastic deformation of the material weld area, the tensile test was used to understand the change law of the mechanical properties of Q355D steel before and after welding, and to provide data support for the subsequent research on welding process performance quality assessment and optimization. In the process of preparing the sample, the wire cutting equipment was used to cut the sample, and the tensile sample was as follows: Figure 7 In order to reduce the test error and sample preparation cost, two groups of tensile specimens and three groups of impact specimens were used to take their average values as the test results.
[0036] The impact test was conducted to study the effect of welding parameters on the toughness of the material, i.e. the ability to resist crack generation and expansion. The impact specimens of the rim welded joints were cut by wire cutting. According to the national standard CB / T2650-2008, the welding area is in the center of the entire specimen, and a V-shaped notch is machined in the center. The actual schematic diagram of the impact specimen is shown in the figure below. Figure 8 As shown. According to the national standard GB / T229-2020, the test was completed at room temperature using a pendulum impact tester, with a bending angle of 180°, a roller spacing of 20 mm, and a pressure head diameter of 8 mm. In order to reduce the test error and ensure the accuracy of the test, the average value of the impact energy absorbed by the three impact specimens was taken for evaluation. Table 1 lists the welding parameters and test results.
[0037] Table 1 Welding parameters and test results
[0038]
[0039] In order to optimize the welding parameters, a rim welding process optimization scheme based on NSGAⅡ algorithm and BP neural network is proposed. The BP neural network model used in the present invention is as follows: Fig. 9 shown.
[0040] Matlab was used to optimize welding process parameters using NSGAⅡ algorithm and BP neural network.
[0041] The NSGAⅡ algorithm function is as follows:
[0042] Function 1:
[0043] function y = abc(x);
[0044] y(1)=-(myNeuralNetworkFunction1(x));
[0045] y(2)=-(myNeuralNetworkFunction2(x));
[0046] end
[0047] Function 2:
[0048] clear
[0049] clc
[0050] fitnessfcn=@abc;%Function handle to the fitness function
[0051] nvars=4;%Number of decision variables
[0052] lb=[12,0.8,8,0.2];%Lowerbound
[0053] ub=[18,1.2,16,0.6];%Upperbound
[0054] A=[]; b=[]; %No linear inequality constraints
[0055] Aeq=[];beq=[];%No linear equality constraints
[0056] options=gaoptimset('ParetoFraction',0.3,'PopulationSize',200,'Generations',500,'StallGe
[0057] nLimit',200,'TolFun',1e-100,'PlotFcns',@gaplotpareto);
[0058] [x,fval]=gamultiobj(fitnessfcn,nvars,A,b,Aeq,beq,lb,ub,options);
[0059] The optimization parameters obtained by Matlab software are shown in Table 2. The experimental values are shown in Figure 10a-10b shown.
[0060] Table 2 Optimized welding parameters and test results
[0061]
[0062]
[0063] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware.
[0064] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.
Claims
1. A rim welding process optimization method based on NSGAⅡ algorithm and BP neural network, characterized in that: The following steps are involved: Collect process parameter data during welding, and pre-process and standardize the collected process parameter data; A nonlinear mapping model between welding process parameters and weld quality is established using BP neural network, and the key quality response in the welding process is obtained through training; The NSGA-II algorithm is used for multi-objective optimization, and the elongation and impact energy of the weld are set as optimization targets to obtain the optimal combination of welding process parameters. The BP neural network model is combined with the NSGA-II algorithm in Matlab software, and the optimization results of welding process parameters are obtained by setting the algorithm and adjusting the parameters; The optimized welding process parameters are experimentally verified, and the mechanical properties of the welded joints are evaluated through tensile tests and impact tests to ensure that the weld quality meets the predetermined standards.
2. The rim welding process optimization method based on NSGA-II algorithm and BP neural network according to claim 1 is characterized in that: The process parameter data specifically include a burning distance of 12 to 18 mm, a burning speed of 0.8 to 1.2 mm per second, an upsetting distance of 8 to 16 mm, and an electrified upsetting time of 0.2 to 0.6 seconds.
3. The rim welding process optimization method based on NSGA-II algorithm and BP neural network according to claim 1 is characterized in that: The BP neural network model consists of at least three hidden layers and is trained using the Levenberg-Marquardt algorithm.
4. The rim welding process optimization method based on NSGA-II algorithm and BP neural network according to claim 1 is characterized in that: The optimization objectives of the NSGA-II algorithm include the elongation of the weld, which is set between 20% and 30%, and the impact energy target is set between 8 and 25 joules per square centimeter and evaluated by the Pareto front solution.
5. The wheel rim welding process optimization method based on NSGA-II algorithm and BP neural network according to claim 1, characterized in that: The algorithm settings in the Matlab software include setting the population size to 200, the generation limit to 500 generations, and the algorithm convergence tolerance to 1e-100.
6. The rim welding process optimization method based on NSGA-II algorithm and BP neural network according to claim 1 is characterized in that: The tensile test uses a standardized specimen with a size of 20 mm in width and 100 mm in length, a test speed of 5 mm per minute, and is measured by a universal tensile testing machine.
7. The wheel rim welding process optimization method based on NSGA-II algorithm and BP neural network according to claim 1, characterized in that: In the experimental verification, a metallographic analysis is performed on the microstructure of the weld, and the metallographic analysis is observed using a metallographic microscope to analyze the grain morphology and distribution of the weld area.
8. The rim welding process optimization method based on NSGA-II algorithm and BP neural network according to claim 1 is characterized in that: The key quality responses in the welding process include the hardness of the weld, which is measured by a Rockwell hardness tester, and the toughness of the weld is obtained by an impact test.