A method for predicting residual stress considering ultrasonic impact process parameters of welded joints

Abstract

This invention proposes a method for predicting residual stress in welded joints considering ultrasonic impact process parameters. Specifically, the method involves: conducting thermo-mechanical coupling simulation of the welded structure using the finite element method to obtain the residual stress field, which is then used as the initial stress to establish a finite element model for ultrasonic impact treatment. Using hypercubic Latin sampling, parameter combination samples are extracted, and the corresponding residual stress data after ultrasonic impact are simulated. A residual stress prediction function model with undetermined coefficients is established using the ultrasonic impact parameters as independent variables. Based on particle swarm optimization, the undetermined coefficient values ​​that achieve the best prediction effect are obtained through training, completing the establishment of the prediction function model. This allows for the acquisition of the optimal combination of ultrasonic impact process parameters to improve the mechanical properties of the welded joint. This method for predicting residual stress in welded joints considering ultrasonic impact process parameters will provide a basis for improving the service performance of large-scale thin-walled welded structures.

Description

Technical Field

[0001] This invention belongs to the field of advanced manufacturing such as welding processing, and specifically relates to a method for predicting residual stress considering ultrasonic impact process parameters of welded joints. Background Technology

[0002] Thin-walled welded structures are widely used in high-end equipment fields such as aerospace, automotive, and radar. However, the uneven temperature field during welding generates thermal stress, leading to significant residual deformation and stress. Furthermore, welded joints may exhibit cracks and coarse microstructures, reducing the service performance of thin-walled welded structures. To address this issue, ultrasonic impact treatment is employed as an effective method to reduce welding residual stress. This involves using an impact head to bombard the metal surface at high frequency, inducing plastic deformation, thereby altering the original stress field and strengthening the impacted area. While most studies indicate that ultrasonic impact treatment effectively reduces welding residual stress, the relationship between ultrasonic impact process parameters and joint residual stress has received less attention. Therefore, this study establishes a coupled numerical model of welding and ultrasonic impact to investigate the influence of ultrasonic impact process parameters on welding residual stress. By establishing a predictive function model of the relationship between post-impact residual stress and ultrasonic impact process parameters, the optimal ultrasonic impact process parameters can be obtained, providing a basis for improving the service performance of large-scale thin-walled welded structures. Summary of the Invention

[0003] The purpose of this invention is to provide a method for predicting residual stress considering ultrasonic impact process parameters of welded joints, so as to obtain the optimal ultrasonic impact treatment process parameters and improve the mechanical properties of welded joints.

[0004] The technical solution of this invention is: a method for predicting residual stress considering ultrasonic impact process parameters of welded joints, comprising the following steps:

[0005] Step 1: Perform thermo-mechanical coupling simulation based on the finite element method to obtain the welding residual stress field, then proceed to Step 2.

[0006] Step 2: Based on thermo-mechanical coupling simulation, using the welding residual stress field as the initial stress, establish a finite element model for ultrasonic impact treatment, and proceed to Step 3.

[0007] Step 3: Use the Latin hypercube sampling method to extract parameters within each range of ultrasonic impact parameters to obtain parameter combination samples. Set up an ultrasonic impact finite element model based on the extracted parameter combination samples to obtain residual stress data after impact corresponding to different parameter combinations, and then proceed to Step 4.

[0008] Step 4: Based on the mapping relationship between different parameter combinations and their corresponding post-impact residual stress data, establish a prediction function model and proceed to Step 5.

[0009] Step 5: Based on the particle swarm optimization algorithm, the undetermined coefficients in the prediction function model are used as the optimization target. The fitness function is defined by the prediction error. The particle swarm optimization algorithm parameters are adjusted until the goodness of fit R is achieved. 2 Once the value reaches 0.99, the optimized prediction function model is obtained, and the process proceeds to step 6.

[0010] Step 6: Calculate the optimal ultrasonic shock process parameters based on the optimized prediction function model.

[0011] Compared with the prior art, the significant advantages of this invention are: by establishing a predictive function model between ultrasonic impact process parameters and residual stress, this invention can accurately predict the influence of different parameter combinations on residual stress, and can obtain the optimal impact process parameters through optimization algorithms, thereby improving the mechanical properties of welded joints, reducing experimental time and costs, and improving the design efficiency of the process. Attached Figure Description

[0012] Figure 1 This is a flowchart of a method for predicting residual stress in welded joints that takes into account ultrasonic impact process parameters, as described in this invention.

[0013] Figure 2 This is the finite element model of the T-shaped thin-walled component welding structure in the embodiment of the present invention.

[0014] Figure 3 This is a comparison diagram of the impact path stress curves before and after ultrasonic impact treatment of the welded structure in this embodiment of the invention.

[0015] Figure 4 This is a fitness evolution curve of the particle swarm optimization algorithm in an embodiment of the present invention.

[0016] Figure 5 This is a comparison chart of the predicted value and the actual value of the prediction function in an embodiment of the present invention.

[0017] Figure 6 These are grain size distribution diagrams of the weld before and after ultrasonic impact in the embodiments of the present invention, wherein (a) represents the weld before impact and (b) represents the weld after impact.

[0018] Figure 7 This is a curve showing the average hardness of the welded structure before and after ultrasonic impact in an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0020] Furthermore, the technical solutions of the various embodiments of the present invention can be combined with each other, but only if they are feasible to those skilled in the art. If the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0021] The following section will further introduce the specific implementation method, as well as the technical difficulties and inventive points of this invention, using this design example as an example.

[0022] Combination Figure 1 A method for predicting residual stress considering ultrasonic impact process parameters of welded joints, comprising the following steps:

[0023] Step 1: Perform thermo-mechanical coupling simulation based on the finite element method to obtain the welding residual stress field, as follows:

[0024] The uneven temperature field during welding will generate thermal stress, resulting in residual deformation and residual stress. Based on the finite element method, a welding finite element model is established to complete the thermo-mechanical coupling simulation and obtain the welding residual stress field.

[0025] Proceed to step 2.

[0026] Step 2: Based on thermo-mechanical coupling simulation, using the welding residual stress field as the initial stress, a finite element model for ultrasonic impact treatment is established, as follows:

[0027] Ultrasonic impact treatment, as an effective method to reduce residual stress in welding, uses an impactor to bombard the metal surface at high frequency, causing plastic deformation, thereby altering the original stress field and strengthening the impacted area. A finite element model of ultrasonic impact treatment is established based on the welding model. The residual stress field obtained from thermo-coupling simulation is used as the initial stress condition and applied to the model. A velocity load parallel to the weld direction is applied to the ultrasonic impactor to simulate its movement along the weld direction; a displacement load perpendicular to the weld surface is applied to simulate the continuous vibration of the impact pin.

[0028] Proceed to step 3.

[0029] Step 3: Using the Latin hypercube sampling method, parameters are extracted within the range of various ultrasonic impact parameters (including impact frequency, impact amplitude, and impact velocity) to obtain parameter combination samples. Based on the extracted parameter combination samples, an ultrasonic impact finite element model is set up to obtain the residual stress data after impact corresponding to different parameter combinations, as detailed below:

[0030] When building a prediction function model, the accuracy of the model depends on the coverage and representativeness of the training sample data. Latin hypercube sampling is an efficient sampling method that can select sample points relatively evenly in a multidimensional parameter space, avoiding sample concentration and bias problems, and making the prediction function model more applicable and accurate under different process conditions.

[0031] Within the range of ultrasonic shock treatment parameters, parameters are extracted using the Latin hypercube sampling method to obtain parameter combination samples. Ultrasonic shock simulation is then performed on all parameter combination samples using the established ultrasonic shock treatment finite element model. The residual stress data after impact corresponding to different parameter combination samples are obtained by using the stress at the maximum stress point before impact on the ultrasonic shock path as an indicator.

[0032] Proceed to step 4.

[0033] Step 4: Based on the mapping relationship between different parameter combinations and their corresponding post-impact residual stress data, establish a prediction function model, as follows:

[0034] First, the effects of impact frequency, impact amplitude, and impact velocity on the residual stress value after impact are comprehensively considered. The number of times the ultrasonic impactor strikes the specimen increases with increasing impact frequency and decreasing impact velocity, while the impact depth increases with increasing impact amplitude. Therefore, the impact intensity is considered to be directly proportional to the impact frequency and impact amplitude, and inversely proportional to the impact velocity. Thus, the comprehensive variable X is defined to satisfy the following formula:

[0035]

[0036] Where x1 is the impact frequency; x2 is the impact amplitude; and x3 is the impact velocity.

[0037] Plot the curve of the comprehensive variable X versus the relevant stress value. Based on the curve, it can be preliminarily determined that the two have an exponential relationship. Therefore, the following relationship function is defined:

[0038]

[0039] Where y is the residual stress after impact; x1 is the impact frequency; x2 is the impact amplitude; x3 is the impact velocity; and a, n, and m are all undetermined coefficients.

[0040] Subsequently, the correlation between stress values ​​and impact frequency, impact amplitude, and impact velocity was considered, and the variation of stress values ​​was analyzed when two parameters remained constant and the third parameter changed. According to the data curves, the relationship between impact amplitude, impact velocity, and stress values ​​approximates a parabola, while the relationship between impact frequency and stress values ​​fluctuates but is generally positively correlated. Therefore, the prediction function model is:

[0041]

[0042] Where y is the residual stress after impact; x1 is the impact frequency; x2 is the impact amplitude; x3 is the impact velocity; and a, b, c, d, e, f, g, h, k, n, and m are all undetermined coefficients.

[0043] Proceed to step 5.

[0044] Step 5: Based on the particle swarm optimization algorithm, the undetermined coefficients in the prediction function model are used as the optimization target. The fitness function is defined by the prediction error. The particle swarm optimization algorithm parameters are adjusted until the goodness of fit R is achieved. 2 The value reached 0.99 to obtain the optimized prediction function model, as follows:

[0045] The prediction function model is trained using the particle swarm optimization algorithm to meet the prediction requirements. In the particle swarm algorithm, particles move in the solution space and update their velocity and position by tracking the individual optimal value and the global optimal value of the group until the fitness no longer changes or the maximum number of iterations is reached. In order to balance the global search capability in the early stage and the convergence speed in the later stage, the velocity inertia weight of the particles is set to a linearly decreasing weight.

[0046] Using the undetermined coefficients as the optimization target, and assuming the prediction error is the reciprocal of the mean square error between the predicted stress value obtained by the prediction function model and the actual stress value of the sample, the prediction error is used as the fitness function to optimize and train the undetermined coefficients, as shown in the following formula:

[0047]

[0048] Where MSE is the mean squared error; H is the number of samples, and y actual,i Let y be the actual residual stress value of the i-th sample. predicted,i Let F be the model prediction value for the i-th sample parameter, and F be the fitness.

[0049] By adjusting the particle swarm optimization algorithm parameters until the fit R is achieved... 2 When the value reaches 0.99, the final output is the optimal coefficient that makes the predicted value match the actual value, thus obtaining the optimized prediction function model.

[0050] Proceed to step 6.

[0051] Step 6: Calculate the optimal ultrasonic impact process parameters based on the optimized prediction function model, as follows:

[0052] The optimized prediction function model can accurately predict residual stress under different ultrasonic impact parameters. By optimizing this function and outputting the impact frequency, impact amplitude, and impact velocity that result in the lowest function value, the optimal combination of process parameters can be obtained, leading to good mechanical properties of the welded joint.

[0053] Example 1

[0054] Combination Figure 1 The residual stress prediction method for welded joints considering ultrasonic impact process parameters, as described in this invention, comprises the following steps:

[0055] Step 1: Uneven temperature fields during welding generate thermal stress, resulting in significant residual deformation and residual stress. Based on the finite element method, a welding finite element model is established as follows: Figure 2 Complete the thermo-mechanical coupling simulation, obtain welding residual stress field data, and proceed to step 2.

[0056] Step 2: Ultrasonic impact treatment can effectively reduce residual welding stress and strengthen the impacted area, such as... Figure 6 , Figure 7 As shown, a finite element model of ultrasonic impact treatment is established based on thermo-mechanical coupling simulation to obtain the residual stress after impact, as shown in the figure. Figure 3 The details are as follows:

[0057] An ultrasonic impact model is established by adding an impact head to the welding model. The residual stress field obtained from the thermo-coupling simulation is used as the initial stress condition and applied to the model. A velocity load parallel to the weld direction is applied to the ultrasonic impact head to realize its movement along the weld direction; a displacement load perpendicular to the weld surface is applied to simulate the continuous vibration of the impact needle. The simulation can obtain the residual stress field after impact, and then proceed to step 3.

[0058] Step 3: Based on the Latin hypercube sampling method, parameter combination samples are extracted within the range of ultrasonic impact parameters. A finite element model for ultrasonic impact treatment is then set up based on the extracted parameter combination samples to obtain the corresponding residual stress data after impact, as detailed below:

[0059] When building a prediction function model, the accuracy of the model depends on the coverage and representativeness of the training sample data. Latin hypercube sampling is an efficient sampling method that can select sample points relatively evenly in a multidimensional parameter space, avoiding sample concentration and bias problems, and making the prediction function model more applicable and accurate under different process conditions.

[0060] In this example, the selected ultrasonic impact parameters are impact frequency, impact amplitude, and impact velocity. The impact frequency range is 12–24 kHz, the impact amplitude range is 1–6 μm, and the impact velocity is 30–80 mm / s. Within each range of ultrasonic impact parameters, 36 parameter combination samples are extracted using the Latin hypercube sampling method to ensure that the established prediction function model can more accurately describe the relationship between ultrasonic impact parameters and welding residual stress.

[0061] The ultrasonic impact model was established to simulate ultrasonic impact for all parameter combination samples. The stress at the point of maximum stress before impact on the ultrasonic impact path was used as an indicator to obtain the residual stress data after impact for each parameter combination sample.

[0062] Step 4: Based on the mapping relationship between different parameter combinations and their corresponding post-impact residual stress data, establish a prediction function model, as follows:

[0063] First, the effects of impact frequency, impact amplitude, and impact velocity on the residual stress value after impact are comprehensively considered. The number of times the ultrasonic impactor strikes the specimen increases with increasing impact frequency and decreasing impact velocity, while the impact depth increases with increasing impact amplitude. Therefore, the impact intensity is considered to be directly proportional to the impact frequency and impact amplitude, and inversely proportional to the impact velocity. Thus, the comprehensive variable X is defined to satisfy the following formula:

[0064]

[0065] Where x1 is the impact frequency; x2 is the impact amplitude; and x3 is the impact velocity.

[0066] Plot the curve of the comprehensive variable X versus the relevant stress value. Based on the curve, it can be preliminarily determined that the two have an exponential relationship. Therefore, the following relationship function is defined:

[0067]

[0068] Where y is the residual stress after impact; x1 is the impact frequency; x2 is the impact amplitude; x3 is the impact velocity; and a, n, and m are all undetermined coefficients.

[0069] Subsequently, the correlation between stress values ​​and impact frequency, impact amplitude, and impact velocity was considered, and the variation of stress values ​​was analyzed when two parameters remained constant and the third parameter changed. According to the data curves, the relationship between impact amplitude, impact velocity, and stress values ​​approximates a parabola, while the relationship between impact frequency and stress values ​​fluctuates but is generally positively correlated. Therefore, the prediction function model is:

[0070]

[0071] Where y is the residual stress after impact; x1 is the impact frequency; x2 is the impact amplitude; x3 is the impact velocity; and a, b, c, d, e, f, g, h, k, n, and m are all undetermined coefficients.

[0072] Step 5: Based on the particle swarm optimization algorithm, the undetermined coefficients in the prediction function model are used as the optimization target. The fitness function is defined by the prediction error. The particle swarm optimization algorithm parameters are adjusted according to the results of each run. This process is repeated until the fit R is achieved. 2The value reached 0.99, and the coefficient that ultimately yielded the best predictive performance for the model was determined as follows:

[0073] A particle swarm optimization (PSO) algorithm is used to train the prediction function model to meet the prediction requirements. In PSO, particles move in the solution space, updating their velocity and position by tracking individual optimal values ​​and the global optimum of the swarm, until the fitness no longer changes or the maximum number of iterations is reached. To balance the early global search capability with the later convergence speed, the particle velocity inertia weight is set to a linearly decreasing weight from 0.9 to 0.4.

[0074] Eleven undetermined coefficients (a, b, c, d, e, f, g, h, k, n, m) are used as optimization targets. The reciprocal of the mean square error between the predicted stress value obtained from the prediction function model and the actual stress value of the sample is used as the fitness function to optimize and train the undetermined coefficients. The formula is as follows:

[0075]

[0076] Where MSE is the mean squared error; H is the number of samples, and y actual,i Let y be the actual residual stress value of the i-th sample. predicted,i Let F be the model prediction value for the i-th sample parameter, and F be the fitness.

[0077] Based on the coefficient values ​​and mean squared error output from each program run, adjust the optimization algorithm parameters such as the search space range, number of evolutions, and population size, repeating this process. The final fitness curve is shown below. Figure 4 The curve of the prediction function is as follows Figure 5 As shown. At this point, the predicted value and the actual value are highly consistent, with a correlation coefficient R. 2 A score of 0.99 indicates that the established prediction function model has good predictive performance.

[0078] Step 6: Calculate the optimal ultrasonic impact process parameters based on the optimized prediction function model, as follows:

[0079] The optimized prediction function model can accurately predict residual stress under different ultrasonic impact parameters. By optimizing this function and outputting the impact frequency, impact amplitude, and impact velocity that result in the lowest function value, the optimal combination of process parameters can be obtained, leading to good mechanical properties of the welded joint.

Claims

1. A method for predicting residual stress considering ultrasonic impact process parameters of welded joints, characterized in that, The steps are as follows: Step 1: Perform thermo-mechanical coupling simulation based on the finite element method to obtain the welding residual stress field, then proceed to Step 2; Step 2: Based on thermo-mechanical coupling simulation, using the welding residual stress field as the initial stress, establish a finite element model for ultrasonic impact treatment, and proceed to Step 3; Step 3: Use the Latin hypercube sampling method to extract parameters within each range of ultrasonic impact parameters to obtain parameter combination samples. Set up a finite element model for ultrasonic impact treatment based on the extracted parameter combination samples to obtain residual stress data after impact corresponding to different parameter combinations, and then proceed to step 4. Step 4: Based on the mapping relationship between different parameter combinations and their corresponding post-impact residual stress data, establish a prediction function model, as follows: First, the effects of impact frequency, impact amplitude, and impact velocity on the residual stress value after impact are comprehensively considered. The number of times the ultrasonic impactor strikes the specimen increases with increasing impact frequency and decreasing impact velocity, while the impact depth increases with increasing impact amplitude. Therefore, the impact intensity is considered to be directly proportional to the impact frequency and impact amplitude, and inversely proportional to the impact velocity. Thus, the comprehensive variable X is defined to satisfy the following formula: Where x1 is the impact frequency; x2 is the impact amplitude; and x3 is the impact velocity. Plot the curve of the comprehensive variable X versus the relevant stress value. Based on the curve, it is preliminarily determined that the two have an exponential relationship. Therefore, the following relationship function is defined: Where y is the residual stress after impact; x1 is the impact frequency; x2 is the impact amplitude; x3 is the impact velocity; and a, n, and m are all undetermined coefficients. Subsequently, the correlation between stress values ​​and impact frequency, impact amplitude, and impact velocity was considered, and the variation of stress values ​​was analyzed when two parameters remained constant and the third parameter changed. According to the data curves, the relationship between impact amplitude, impact velocity, and stress values ​​is parabolic, while the relationship between impact frequency and stress values ​​fluctuates but is generally positively correlated. Therefore, the prediction function model is: Where y is the residual stress after impact; x1 is the impact frequency; x2 is the impact amplitude; x3 is the impact velocity; and a, b, c, d, e, f, g, h, k, n, and m are all undetermined coefficients. Proceed to step 5; Step 5: Based on the particle swarm optimization algorithm, the undetermined coefficients in the prediction function model are used as the optimization target. The fitness function is defined by the prediction error. The particle swarm optimization algorithm parameters are adjusted until the goodness of fit R is achieved. 2 Once the value reaches 0.99, the optimized prediction function model is obtained, and the process proceeds to step 6. Step 6: Calculate the optimal ultrasonic shock process parameters based on the optimized prediction function model.

2. The residual stress prediction method considering ultrasonic impact process parameters of welded joints as described in claim 1, characterized in that, In step 1, a thermo-mechanical coupling simulation is performed based on the finite element method to obtain the welding residual stress field, as detailed below: The uneven temperature field during welding will generate thermal stress, resulting in residual deformation and residual stress. Based on the finite element method, a welding finite element model is established to complete the thermo-mechanical coupling simulation and obtain the welding residual stress field.

3. The residual stress prediction method considering ultrasonic impact process parameters of welded joints as described in claim 2, characterized in that, In step 2, based on thermo-mechanical coupling simulation, a finite element model for ultrasonic impact treatment is established using the welding residual stress field as the initial stress, as detailed below: An ultrasonic impact model was established based on the welding model. The welding residual stress field obtained from the thermo-coupling simulation was used as the initial stress condition and applied to the ultrasonic impact model. An ultrasonic impact head was set up and a displacement load perpendicular to the weld surface and a velocity load parallel to the weld were applied to it to simulate the continuous vibration and horizontal movement of the impact needle, thus establishing a finite element model of ultrasonic impact treatment.

4. The residual stress prediction method considering ultrasonic impact process parameters of welded joints as described in claim 3, characterized in that, In step 3, the Latin hypercube sampling method is used to extract parameters within each range of ultrasonic impact parameters to obtain parameter combination samples. Based on the extracted parameter combination samples, an ultrasonic impact finite element model is set up to obtain the residual stress data after impact corresponding to different parameter combinations, as detailed below: Within the range of ultrasonic shock treatment parameters, parameters are extracted using the Latin hypercube sampling method to obtain parameter combination samples. Ultrasonic shock simulation is then performed on all parameter combination samples using the established ultrasonic shock treatment finite element model. The residual stress data after impact corresponding to different parameter combination samples are obtained by using the stress at the maximum stress point before impact on the ultrasonic shock path as an indicator.

5. The residual stress prediction method considering ultrasonic impact process parameters of welded joints as described in claim 4, characterized in that, The parameters for ultrasonic shock treatment include shock frequency, shock amplitude, and shock velocity.

6. The residual stress prediction method considering ultrasonic impact process parameters of welded joints as described in claim 5, characterized in that, In step 5, based on the particle swarm optimization algorithm, the undetermined coefficients in the prediction function model are used as the optimization target, and the fitness function is defined by the prediction error. The particle swarm optimization algorithm parameters are adjusted until the goodness of fit R is achieved. 2 The value reached 0.99 to obtain the optimized prediction function model, as follows: The prediction function model is trained using the particle swarm optimization algorithm to meet the prediction requirements. In the particle swarm algorithm, particles move in the solution space and update their velocity and position by tracking the individual optimal value and the global optimal value of the group until the fitness no longer changes or the maximum number of iterations is reached. In order to balance the global search capability in the early stage and the convergence speed in the later stage, the velocity inertia weight of the particles is set to a linearly decreasing weight. Using the undetermined coefficients as the optimization target, and assuming the prediction error is the reciprocal of the mean square error between the predicted stress value obtained by the prediction function model and the actual stress value of the sample, the prediction error is used as the fitness function to optimize and train the undetermined coefficients, as shown in the following formula: Where MSE is the mean squared error; H is the number of samples, and y actual,i Let y be the actual residual stress value of the i-th sample. predicted,i Let F be the model prediction value for the i-th sample parameter, and F be the fitness. By adjusting the particle swarm optimization algorithm parameters until the fit R is achieved... 2 When the value reaches 0.99, the final output is the optimal coefficient that makes the predicted value match the actual value, thus obtaining the optimized prediction function model.

7. The residual stress prediction method considering ultrasonic impact process parameters of welded joints as described in claim 6, characterized in that, In step 6, the optimal ultrasonic impact process parameters are calculated based on the optimized prediction function model, as follows: The optimized prediction function model is optimized by taking the parameter variables as the optimization objective, and the optimal combination of process parameters that minimizes the residual stress value after impact is obtained.

Images