Double-side composite friction stir welding core area temperature real-time monitoring method

Abstract

The invention belongs to the technical field of friction stir welding process monitoring, and relates to a method for monitoring the temperature of a double-side composite friction stir welding core area in real time. According to the method, friction heat production of an upper shaft shoulder and a lower shaft shoulder, friction heat production of a stirring needle and metal plastic deformation heat production in the welding process are comprehensively considered, a DS-FSW process simulation model is established by applying a coupled Euler-Lagrange simulation technology, and temperature distribution in the welding process is obtained; and constructing an incidence relation model between the extreme temperature of the core area and the temperature of the surface feature point of the edge of the shaft shoulder by combining an improved random dual coordinate rising algorithm, and realizing the representation of the extreme temperature of the core area based on the surface temperature. An on-line monitoring system integrating surface infrared real-time temperature measurement, parameter adjustment and visualization functions is developed, and accurate on-line monitoring of the temperature of the welding core area is achieved. According to the invention, a high-precision intelligent monitoring means is provided for real-time regulation and control of DS-FSW process parameters, and the welding quality and the process stability of the medium-thickness plate are remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of friction stir welding process monitoring, and relates to a method for real-time monitoring of the temperature in the core area of double-sided composite friction stir welding. It is a method for monitoring the temperature in the core area of double-sided composite friction stir welding (DS-FSW) based on the correlation between the temperature on the surface of the welded part and the core area. Background Art

[0002] Friction Stir Welding (FSW) is a solid-state plastic connection technology invented by the Welding Institute of the United Kingdom in the 1990s. It realizes the plastic connection of materials through the friction heat generated by the stirring head and the workpiece, and has been widely used in the fields of aerospace, rail transit, etc. However, traditional single-sided FSW has significant technical bottlenecks when welding medium and thick plate materials: there is a non-linear temperature gradient along the thickness direction, resulting in defects such as uneven microstructure in the thermo-mechanical affected zone, incomplete penetration at the root, and excessive welding deformation. Therefore, some scholars have proposed the DS-FSW technology, that is, synchronous welding of the stirring heads on both sides of the welded part to balance the temperature field along the thickness direction. Its two-way thermal and mechanical effects promote the full plastic flow of the material along the thickness, eliminating the defect of incomplete penetration at the root; at the same time, the symmetric heat input suppresses the asymmetric deformation, significantly improving the weld microstructure uniformity and dimensional stability, providing a new path for high-quality welding of thick plates.

[0003] The temperature in the core area of DS-FSW (including the weld nugget area and the thermo-mechanical affected zone located under the shoulder of the stirring head) will affect the plastic flow of the weld material and the evolution of the microstructure. Too high or too low temperature will cause welding defects, thus affecting the weld quality. Therefore, it is necessary to characterize the temperature in the core area for timely control. However, the effective monitoring of the temperature in the core area of DS-FSW faces the following technical challenges: (1) The double heat source coupling mechanism is complex, and the synchronous heat generation of the upper and lower shoulders and the stirring pin makes it difficult to model the temperature field; (2) Existing contact temperature measurement methods (such as embedded thermocouples) have inherent defects such as damaging the workpiece or the structure of the stirring head and response lag; (3) Non-contact infrared temperature measurement can only obtain the surface temperature data of a limited area at the edge of the shoulder due to the occlusion of the double-shoulder structure. The current domestic and foreign research has not yet broken through the in-situ measurement technology of the temperature field in the core area of DS-FSW. Summary of the Invention

[0004] Aiming at the problems of the existing technology, the present invention provides a method for real-time monitoring of the temperature in the core area of double-sided composite friction stir welding (DS-FSW) based on the temperature of surface feature points, and develops a temperature monitoring system for the core area of DS-FSW to achieve real-time monitoring of the temperature in the core area. The present invention innovatively integrates the temperature measurement technology of infrared thermal imagers, thermo-mechanical coupling finite element simulation and the improved stochastic dual coordinate ascent (ISDCA) algorithm to construct an intelligent temperature monitoring system for the core area of the DS-FSW process, and solves the technical problem of real-time characterization of the temperature in the core area during the welding process of medium and thick plates.

[0005] The technical solution of the present invention is as follows:

[0006] A method for real-time monitoring of the temperature in the core area of double-sided composite friction stir welding, the specific steps are as follows:

[0007] Step 1: Considering comprehensively the heat generation in the contact areas between the upper and lower tool shoulders and the base material of DS-FSW and the heat generation due to metal plastic deformation, establish a simulation model of the DS-FSW process based on the coupled Euler-Lagrange method;

[0008] Step 2: Analyze the temperature field distribution of the DS-FSW process simulation model, and extract the data set of the extreme temperatures (peak temperature and minimum temperature) in the core area and the temperatures of the corresponding surface feature points at the edge of the tool shoulder during the welding process;

[0009] Step 3: Establish a characterization model for the extreme temperatures in the core area based on the ISDCA algorithm, which model includes a basic prediction model and two parts of model adaptive optimization using a dynamic adapter;

[0010] The basic prediction model takes the temperature data of the surface feature points on the advancing side and the retreating side during the welding process as input, and preprocesses the original data through a normalization layer and a clipping layer to eliminate the dimensional difference between the data and limit the interference of outliers to the model. Then the model uses the stochastic coordinate ascent algorithm to optimize the parameters, and introduces an early stopping mechanism during the training process. By monitoring the change of the loss of the validation set in real time, when the validation loss has not been improved for multiple consecutive epochs, the training is automatically terminated to prevent the model from overfitting. After the training is completed, the basic model outputs the characterization result of the extreme temperatures in the welding core area, and the model parameters are kept stable through regularization constraints, providing a reliable benchmark prediction for the subsequent model adaptive optimization.

[0011] With the continuous accumulation of experimental data, in order to further improve the adaptability of the model to actual data, the dynamic adapter adaptively optimizes and corrects the basic prediction model through the surface feature point temperature measured in real time and the extreme temperature data in the welding core area. The dynamic adapter is innovatively designed as a structure combining a feature extractor and a parameter generator. Among them, the feature extractor is responsible for capturing key features (such as weights, biases, etc.) in the input data, and the parameter generator generates the required scale factors and bias adjustment terms in real time. These parameters are dynamically updated through the backpropagation algorithm with noise estimation to achieve the deep integration of the basic prediction model and the latest experimental data. Among them, the scale factor is strictly constrained in the interval [0.8, 1.2] after being processed by the Sigmoid function, and the bias adjustment term is constrained in the interval [-1, 1], so as to ensure that the characterization results of the model are always within a reasonable range. The temperature characterization model after adaptive optimization can more accurately characterize the extreme temperature in the core area during the actual welding process.

[0012] Step 4: Establish a DS-FSW core area temperature monitoring system, and the system integrates the functions of real-time measurement of the surface feature point temperature of the welded part based on an infrared thermal imager, parameter adjustment, and visualization;

[0013] Step 5: Package and embed the extreme temperature characterization model in the core area of DS-FSW into the DS-FSW core area temperature monitoring system;

[0014] Step 6: Conduct DS-FSW experiments. During the welding process, the infrared thermal imager measures the surface feature point temperature at the edge of the tool shoulder in real time, extracts the surface feature point temperature using global variables and inputs it into the extreme temperature characterization model in the core area. The model outputs the extreme temperature characterization value in the core area and displays it on the front-end interface of the monitoring system. When the extreme temperature in the core area exceeds the solid-liquid phase temperature threshold of the welded part material, the over-limit alarm function is triggered to provide a reference for welding process control.

[0015] Advantages of the present invention:

[0016] The present invention proposes a DS-FSW core area temperature monitoring method based on the correlation between the surface and core area temperatures of the welded part, establishes an extreme temperature characterization model in the core area based on the improved stochastic dual coordinate ascent algorithm, develops a DS-FSW core area extreme temperature monitoring system, realizes the in-situ characterization of the core area temperature in double-sided composite friction stir welding for the first time, realizes the real-time monitoring of the core area temperature during the welding process, and conducts on-site experimental verification to verify the effectiveness of the proposed method, which can provide a reference for welding process control. Description of the Drawings

[0017] Figure 1 It is a schematic diagram for extracting the extreme temperature in the core area;

[0018] Figure 2It is a schematic diagram for extracting the temperature of surface feature points;

[0019] Figure 3 It is a flow chart for establishing the extreme temperature characterization model of the core area;

[0020] Figure 4 It is the extreme temperature result of the core area characterized by the monitoring system. Specific implementation manners

[0021] The following further describes the specific implementation manners of the present invention in combination with the accompanying drawings and technical solutions.

[0022] In this embodiment, a process simulation model of 12 mm thick 2219 aluminum alloy DS-FSW is established, the extreme temperature data of the surface and the core area are extracted and the correlation between the two is established. Based on the correlation, an extreme temperature characterization model of the core area is established and a core area temperature monitoring system is built. Combining with the real-time measurement of the surface temperature based on an infrared thermal imager, the real-time monitoring of the core area temperature is realized. The specific steps are as follows:

[0023] (1) Establish a process simulation model of 12 mm thick 2219 aluminum alloy DS-FSW based on the coupled Euler-Lagrange method.

[0024] Use UG to establish the geometric models of the stirring head and the welded part. The geometric parameters of the stirring head are shown in Table 1.

[0025] Table 1 Geometric parameters of the stirring head

[0026]

[0027] The welded part model includes a Lagrangian solid with dimensions of 100 mm × 70 mm × 12 mm and an Eulerian body with dimensions of 100 mm × 70 mm × 14 mm. After the geometric models of the welded part and the stirring head are established, they are imported into the Abaqus finite element simulation software for assembly. During assembly, their bottom surfaces coincide, and the stirring head is located at the center line position 20 mm to the left of the welded part. The bottom surface of the stirring pin coincides with the upper surface of the Lagrangian solid. Among them, the Lagrangian solid is used for material assignment, and the Eulerian body participates in the simulation calculation. The mesh elements of the stirring head adopt tetrahedrons, with a total number of nodes of 21,712 and a total number of elements of 115,310. The mesh elements of the welded part adopt hexahedrons, with a total number of nodes of 341,901. The total number of meshes is reduced from 1.05 million to 300,000 through local refinement. The heat during the welding process comes from frictional heat generation and plastic deformation heat generation. The initial temperature is set to 25 °C, considering three heat dissipation forms: heat conduction (conductivity 100), heat convection, and heat radiation. The heat conduction coefficient between the welded part and the fixture is set to 100 W / m 2 °C, and the heat conduction coefficient between the welded part and the air is 300 W / m 2°C. The contact mode between the stirring head and the welded part is set as general contact. In the initial stage of welding, sliding friction is the main heat generation mode; when the temperature exceeds a certain specific value, the material becomes a viscous flow state and undergoes plastic flow, and shear friction becomes dominant. The Johnson-Cook constitutive model is selected to describe the relationship between the flow stress of 2219 aluminum alloy and temperature, strain, and strain rate. Finally, based on the established process simulation model, the temperature fields under multiple groups of different process parameters are obtained.

[0028] (2) Obtain the dataset of the extreme temperature in the core area and the temperature of the surface feature points at the edge of the shoulder during the welding process based on the DS-FSW process simulation model.

[0029] By analyzing the temperature field distribution under different process parameters, it is found that the peak temperature in the core area of DS-FSW is located below the shoulder of the stirring head on the cross-section of the weld, while the lowest temperature in the core area is located at the bottom surface of the stirring pin on the cross-section of the weld, as Figure 1 shown. The selected surface feature points are located on both sides of the shoulder, as Figure 2 shown. This position can ensure that the temperature measurement error caused by the thermal radiation of the stirring head is avoided when measuring the temperature with an infrared thermal imager during the welding process. 144 groups of surface feature point temperature data and the corresponding extreme temperature data in the core area are extracted as the dataset during the welding feed stage of the process simulation under different process parameters.

[0030] (3) Establish a characterization model for the extreme temperature in the core area based on the ISDCA algorithm.

[0031] The establishment process of this model is as Figure 3 shown. The basic prediction model is established using the random coordinate ascent algorithm, and its expression form is:

[0032]

[0033] where, is the extreme temperature in the core area output by the basic model, the input feature dimension n = 2, x i corresponds to the measured value of the surface feature point temperature, w i is the weight of the input quantity, and b is the bias.

[0034] During the training process, with the surface feature point temperature as the input and the extreme temperature in the core area as the output, the training / validation / test set is divided according to 100:22:22. The mean square error is used as the loss function, the convergence threshold is set to 0.001, the initial learning rate is set to 0.001, the weight initialization range is set to [0, 0.01], the initial value of the bias is set to 0, and the maximum number of iterations is 100. An early stopping mechanism is set, that is, the training of the model stops when the validation loss no longer decreases in 5 consecutive iterations to prevent overfitting of the model. Finally, the model parameters with the best generalization performance are saved as the basic prediction model.

[0035] To adapt to experimental data, a dynamic adapter is introduced to further improve the generalization performance of the model. As experimental data continues to be generated, the adapter generates a scale factor α(x) ∈ [0.8, 1.2] and a bias term β(x) ∈ [-1, 1] in a feed-forward manner to correct the output of the basic prediction model:

[0036]

[0037] The temperature characterization model trained by the above method can output the extreme temperature characterization value of the core area after adaptive optimization.

[0038] (4) Development and experimental verification of the monitoring system.

[0039] A bilateral composite FSW core area extreme temperature monitoring system is developed based on C#, and a class file for running the core area extreme temperature characterization model is embedded in the software of the system. During the welding process, the temperature of the surface feature points of the welded part measured by the calorimeter is extracted using global variables, and the core area extreme temperature characterization model is called and characterized on the front-end interface. When the extreme temperature of the core area exceeds the solid-liquid phase temperature threshold of the welded part material, the over-limit alarm function is triggered for timely regulation.

[0040] 12 mm thick 2219 aluminum alloy welding experiments under 5 different process parameters are carried out using a bilateral composite friction stir welding device. The surface temperature is collected by an infrared thermal imager, the extreme temperature of the core area of the bilateral composite FSW is characterized in situ by the monitoring system, and a thermocouple is used for verification. The extreme temperature data of the friction stir welding core area obtained from the thermocouple is compared with the extreme temperature results of the core area characterized by the monitoring system. As Figure 4 shown, the maximum relative error of the peak temperature in the welding core area is 3.96%, and the maximum relative error of the lowest temperature in the core area is 4.70%, verifying the effectiveness of the proposed monitoring method.

Claims

1. A method for real-time monitoring of the core temperature of double-sided composite stir friction welding, characterized in that: The specific steps are as follows: Step 1: Considering the heat generation in the contact area between the upper and lower shoulders and the base material of double-sided composite stir friction welding DS-FSW and the heat generation in the metal plastic deformation, a DS-FSW process simulation model based on the coupled Euler-Lagrangian method is established; Step 2: Analyze the temperature field distribution of the DS-FSW process simulation model and extract the data set of the extreme temperature of the core area and the temperature of the corresponding characteristic points on the shoulder edge surface during the welding process; Step 3: Establish a core area extreme temperature characterization model based on the improved random dual coordinate ascent algorithm ISDCA, including two parts: a basic prediction model and a dynamic adapter for model adaptive optimization; The basic prediction model takes the temperature data of the surface feature points on the forward and backward sides of the welding process as input, standardizes and preprocesses the raw data through the normalization layer and the cropping layer, then uses the random coordinate ascent algorithm to optimize the parameters, and introduces the early stopping mechanism in the training process; After training, the basic model outputs the extreme temperature characterization results of the welding core area. The model parameters are kept stable through regularization constraints, providing a reliable benchmark prediction for subsequent model adaptive optimization. The dynamic adapter adaptively optimizes and corrects the basic prediction model through real-time measurement of surface feature point temperature and extreme temperature data of the welding core area; the dynamic adapter includes a feature extractor and a parameter generator, wherein the feature extractor is responsible for capturing key features in the input data, including weights and biases; The parameter generator generates the required scale factors and bias adjustments in real time; these parameters are dynamically updated via a back-propagation algorithm with noise estimation; Step 4: Establish a DS-FSW core area temperature monitoring system, which integrates the real-time measurement, parameter adjustment and visualization functions of the characteristic point temperature on the weldment surface based on an infrared thermal imager; Step 5: Encapsulate the core area extreme temperature characterization model and embed it into the DS-FSW core area temperature monitoring system; Step 6: Conduct DS-FSW experiment. During the welding process, the infrared thermal imager measures the temperature of the characteristic points on the edge of the stirring head shoulder in real time. The surface characteristic point temperature is extracted using global variables and input into the core area extreme temperature characterization model. The model outputs the core area extreme temperature characterization value and displays it on the front-end interface of the monitoring system. When the core area extreme temperature exceeds the solid-liquid phase temperature threshold of the weldment material, the over-limit alarm function is triggered, providing a reference for welding process control.

2. A method for real-time monitoring of the core temperature of double-sided composite stir friction welding according to claim 1, characterized in that: In step 3, the basic prediction model is established using the random coordinate ascent algorithm, which is expressed as: in, The core area extreme temperature output by the basic model, input feature dimension n = 2, x i Corresponding to the temperature measurement value of the surface feature point, w i is the weight of the input, and b is the bias.

3. The method for real-time monitoring of the core temperature of double-sided composite stir friction welding according to claim 1 is characterized in that: In step 3, the basic prediction model output after dynamic adapter optimization and correction is:

4. The method for real-time monitoring of the core temperature of double-sided composite stir friction welding according to claim 1 is characterized in that: In step 3, the scale factor is constrained to the interval [0.8, 1.2] after being processed by the Sigmoid function, and the bias adjustment term is constrained to the interval [-1, 1].

5. The method for real-time monitoring of the core temperature of double-sided composite stir friction welding according to claim 1, characterized in that: In step 2, the core area extreme temperature includes the peak temperature and the minimum temperature.

Images