An online monitoring system and control method for friction stir welding process

Abstract

The application relates to an online monitoring system and control method for a friction stir welding process, belonging to the technical field of metal material welding, which comprises a visual module, a temperature measuring module, a feedback execution module, a performance prediction and evaluation module, the visual module is used for image information collection and transmits the collected image information to the feedback execution module, the temperature measuring module is used for real-time temperature collection and transmits the real-time temperature to the feedback execution module, the feedback execution module is used for material welding and process parameter adjustment and transmits all the obtained information to the performance prediction and evaluation module, and the performance prediction and evaluation module analyzes the obtained information to obtain a welding joint strength prediction and evaluation result. Through comparison and control of the real-time temperature data of the welding seam and the peak temperature of the temperature field, the application guarantees that the temperature fluctuation along the welding seam direction is within a controllable range, and greatly enhances the strength stability of the welding structure.

Description

Technical Field

[0001] This invention relates to the field of metal welding technology, and in particular to an online monitoring system and control method for friction stir welding processes. Background Technology

[0002] With the development of intelligent, precise, and lightweight manufacturing in aerospace, rail transportation, and shipbuilding industries, aluminum alloys have become the primary choice for lightweight structural design materials due to their high specific strength, good corrosion resistance, excellent processing performance, high plasticity, and recyclability. Friction stir welding (FSW) technology is widely used in aluminum alloy welding, offering advantages such as good mechanical properties of the welded parts, low cost, high efficiency, and environmental friendliness. It has significant advantages in welding lightweight alloys such as aluminum and magnesium alloys. Two key parameters affecting the quality of the weld joint in FSW are heat input and pressure. Heat input, as one of the important factors, is crucial for achieving high-performance and stable weld joints, and accurately obtaining the welding temperature value is of critical technical significance.

[0003] Common temperature detection methods are mainly divided into two categories: contact temperature measurement and non-contact temperature measurement. Infrared temperature measurement is widely used among non-contact methods due to its advantages such as rapid response and ease of operation. In the friction stir welding process, infrared temperature measurement has significant advantages, enabling autonomous control of the welding process. In traditional infrared temperature measurement, the emissivity of the target is usually treated as a constant, and its temperature is determined by measuring the target's radiation intensity according to Planck's blackbody radiation law. However, in aluminum alloy infrared temperature measurement, the emissivity changes with temperature, therefore traditional infrared temperature measurement methods cannot meet the ideal temperature measurement accuracy. On the other hand, contact temperature measurement methods, such as thermocouple temperature measurement, have the advantage of measurement accuracy. However, this method has disadvantages such as damaging the plate surface, cumbersome operation, high cost, and inability to monitor the welding process temperature in real time. Therefore, one of the key technical problems currently facing the friction stir welding process is finding a convenient, effective, and accurate temperature detection method for aluminum alloy friction stir welding. Existing technical solutions, such as the published invention patent CN111854968A, propose a method to measure the temperature of the same aluminum ingot using two infrared thermometers with different emissivities, and to establish a system of simultaneous equations based on the two different temperature values ​​and their respective emissivities to obtain the actual temperature of the measured aluminum ingot. However, this method does not consider the problem of aluminum alloy emissivity changing with temperature. On the other hand, invention patent CN116329299A obtains the actual strip temperature through contact temperature measurement, then inputs this temperature into an infrared thermometer to automatically calculate the emissivity, and records the calculated emissivity parameters. Although this method solves the problem of dynamic emissivity calculation, it does not consider the situation of aluminum alloy emissivity variation. In addition, invention patent CN115060376A proposes an aluminum alloy temperature field measurement method based on infrared thermal imaging and iterative algorithms. It calculates a stable emissivity field using data obtained from infrared thermal imaging and thermocouple measurements, and uses this field and radiation intensity to obtain the temperature of the measured target. However, this method cannot adjust the friction stir welding process parameters in real time. Therefore, it is of great significance to develop a method for online monitoring and parameter control of high-performance and stable friction stir welding process to ensure the strength consistency of welded joints along the weld direction. Summary of the Invention

[0004] To address the challenges of accurate temperature measurement in existing aluminum alloy sheet welding processes, which requires high precision and suffers from significant errors, this invention provides an online monitoring system and control method for the friction stir welding process. This method can detect the temperature cycle curve of characteristic points in the weld online and compare the peak temperature of these characteristic points with the temperature fluctuation range of the theoretical temperature field in the process-aided computer. By inputting the temperature difference into a temperature intelligent control model, the adjustment amount of the shoulder pressing force is obtained. The process-aided computer sends instructions to the parameter adjustment mechanism based on the pressing force adjustment, achieving real-time control of the welding process and ensuring stable welding of the aluminum alloy sheet. Simultaneously, the acquired temperature data, process parameters during welding, and weld surface image information are input into a performance prediction and evaluation module. This module obtains a predicted evaluation result of the weld joint strength to determine whether the current weld joint meets the design strength requirements, thereby avoiding the need for post-weld non-destructive testing.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] An online monitoring system for friction stir welding includes a vision module, a temperature measurement module, a feedback execution module, and a performance prediction and evaluation module. The vision module is used for image information acquisition and transmission of the acquired image information to the feedback execution module. The temperature measurement module is used for real-time temperature acquisition and transmission of the real-time temperature to the feedback execution module. The feedback execution module is used for material welding and process parameter adjustment and integrates all the acquired information and transmits it to the performance prediction and evaluation module. The performance prediction and evaluation module analyzes the obtained information to obtain a predicted evaluation result of the weld joint strength.

[0007] The vision module includes a CCD industrial camera and a spherical clamp. The CCD industrial camera is mounted under the pressure plate on the worktable of the friction stir welding machine via the spherical clamp. A first scale and a second scale are coaxially mounted at the connection position between the spherical clamp and the CCD industrial camera. The second scale is closer to the side of the CCD industrial camera. The spherical clamp includes a ball head, a guide rod, and a bolt. The guide rod passes through the center of the ball head, and the ball head and the guide rod are fixed by the bolt. There are two CCD industrial cameras, which are mounted on the lower right side of the pressure plate on the worktable of the friction stir welding machine.

[0008] The temperature measurement module includes a fiber optic infrared thermometer, a universal bamboo-joint tube, an electric air pump, a slide rail device, and a dual-wavelength infrared temperature measurement model installed in a computer. The slide rail device is installed below the pressure plate on the friction stir welding machine's workbench. An angle scale is engraved on the slide rail device, and two sliders are installed in the slide rail via bolts. Two fiber optic infrared thermometers are respectively installed below the sliders through the universal bamboo-joint tube. The slide rail device adopts a U-shaped guide rail design and is connected to the pressure plate on the welding machine's workbench via bolts. The electric air pump is placed on one side of the friction stir welding machine's workbench. One end of a soft PVC conduit is connected to the air outlet of the electric air pump, and the other end of the soft PVC conduit is placed next to the fiber optic infrared thermometer. The fiber optic infrared thermometer inputs the measured two-wavelength radiation intensity into the dual-wavelength infrared temperature measurement model to obtain the measured temperature.

[0009] The dual-wavelength infrared thermometry model is a formula model obtained by minimizing the cost function using calculus and solving for the model parameters. The formula is as follows:

[0010] T=α3×(r ΔDR (T) / C f r ε (T)) 3 +α2×(r ΔDR (T) / C f r ε (T)) 2 +α1×(r ΔDR (T) / C f r ε (T))+α0

[0011] In the formula r ΔDR (T) represents the ratio of the average radiation intensity of the two bands, C f As the correction factor is a constant, r ε (T) represents the ratio of the average emissivity of the two bands, and α3, α2, α1, and α0 are the model setting parameters.

[0012] The feedback execution module includes a friction stir welding machine, a friction stir welding machine worktable, a pressure sensor, a process auxiliary computer, a parameter adjustment mechanism, and an LCD display. The temperature measurement module transmits the measured real-time temperature to the process auxiliary computer, which contains a temperature intelligent control decision model. The real-time temperature is compared with the peak and minimum temperatures of the temperature cycle curve in the process auxiliary computer's database. The temperature cycle curve represents the temperature fluctuation range of the welded joint when obtaining an ideal weld surface. The peak and minimum temperatures are the boundaries of this temperature cycle curve. If the real-time temperature is within the range of the peak and minimum temperatures, welding continues. If the real-time temperature is not within the range of the peak and minimum temperatures, the temperature difference is calculated between the real-time temperature and the closest value between the peak and minimum temperatures. This temperature difference is transmitted to the temperature intelligent control decision model to obtain the downward pressure adjustment amount. The parameter adjustment execution mechanism includes a servo fine-tuning motor and a sliding vertical guide rail. The sliding vertical guide rail is positioned above the friction stir welding machine worktable. One side of the pressure plate on the friction stir welding machine worktable is installed in the sliding vertical guide rail. A servo fine-tuning motor is installed on the side of the pressure plate on the friction stir welding machine worktable that connects to the sliding vertical guide rail. The fine-tuning motor controls the pressure plate on the friction stir welding machine's worktable to move up and down in the sliding vertical guide rail. The parameter adjustment actuator is electrically connected to the friction stir welding machine and the process auxiliary computer. The process auxiliary computer issues instructions to the parameter adjustment actuator based on the downward pressure adjustment amount: when the real-time temperature of the welding area is within the range of peak and minimum temperatures, the current shoulder downward pressure parameter remains unchanged; if the temperature difference in the welding area is positive, the temperature difference is input into the temperature intelligent control decision model to obtain the adjustment amount to reduce the shoulder downward pressure; when the temperature difference in the welding area is negative, the temperature difference is input into the temperature intelligent control... In the decision-making model, to obtain the adjustment amount for increasing the shoulder pressure, a pressure sensor is installed at the connection between the welding machine spindle and the stirring head fixture to measure the stirring head pressure information. The stirring head pressure information is transmitted to the process auxiliary computer via a pressure data signal line. An LCD display is installed on the friction stir welding machine and electrically connected to the industrial auxiliary computer. If the pressure value exceeds the theoretical maximum critical pressure of heat input in the process auxiliary computer, the LCD display will display an alarm signal. The LCD display is used to display the current stirring head axial pressure, real-time temperature, and welding shoulder pressure adjustment amount information.

[0013] The intelligent temperature control decision model adopts a neural network prediction model, which contains a BP neural network structure with two hidden layers. Its input layer parameters include the rotation speed, welding speed, pressure, and temperature difference of the welding area of ​​the friction stir welding machine. The number of neurons in the input layer of the neural network prediction model is 4. The output layer parameter of the neural network prediction model is the pressure adjustment amount, and the number of neurons in the output layer is 1. The selected activation function is the tanh function, and the loss function is the cross-entropy loss function.

[0014] The performance prediction and evaluation module employs a multimodal fusion learning method, which integrates information from different modalities. Specifically, for the surface texture image of the weld joint acquired by the vision module, a convolutional neural network is used for feature extraction to capture the image's texture information. Temperature information acquired by the temperature measurement module, welding process parameters, and image feature information are combined using a multimodal feature fusion method based on splicing and fusion to form the input vector of a fully connected neural network. The output layer uses a sigmoid activation function, limiting the output to between 0 and 1. A stochastic gradient descent algorithm is used to optimize the loss function, and a GA optimization algorithm is used to optimize the initial weights and thresholds of the convolutional neural network. Classification accuracy is used as the fitness function. Through selection, crossover, and compilation operations, the optimized initial weights and thresholds of the convolutional neural network are obtained. To construct the weld strength evaluation model dataset, numerous welding experiments are conducted, covering different rotation speeds, welding rates, and pressure amounts.

[0015] A method for online monitoring and parameter control of the friction stir welding process in an online monitoring system is adopted, which specifically includes the following steps:

[0016] Step 1: Assemble the online monitoring system for the stir welding process. Place the workpiece on the welding machine workbench, adjust the vision module to ensure that the CCD industrial camera can acquire weld seam image information without obstruction, and adjust the position of the fiber optic infrared thermometer to ensure the accuracy of its distance measurement target. The vision module collects weld seam image information, the temperature measurement module collects real-time temperature information, the pressure sensor collects stirring head pressure information, and the welding equipment collects welding process information. Transmit the obtained weld seam image information, real-time temperature information, stirring head pressure information, and welding process information to the process auxiliary computer.

[0017] Step 2: Set the resolution, frame rate, and exposure time in the camera software on the computer, perform system debugging to ensure that the images captured by the CCD industrial camera meet expectations, check the quality of the captured images, test the triggering and synchronization functions of the CCD industrial camera to ensure reliability in the automated control process, set up the fiber optic infrared thermometer, including adjusting the temperature display unit, and use the aiming laser on the fiber optic infrared thermometer to ensure alignment with the temperature measurement target, realize the automatic triggering mechanism for temperature measurement, and select the welding process parameters according to the type and thickness of the aluminum alloy sheet, the parameters of the welding tools, and the process requirements to perform friction stir welding.

[0018] Step 3: The temperature measurement module transmits the real-time temperature to the process auxiliary computer. The real-time temperature is compared with the peak and minimum temperatures of the temperature cycle curve in the process auxiliary computer's database. If the real-time temperature is within the range of the peak and minimum temperatures, welding continues. If it is outside the allowable range, if the measured temperature is greater than the peak temperature, the peak temperature is subtracted from the measured temperature to obtain the temperature difference; if the measured temperature is less than the minimum temperature, the minimum temperature is subtracted from the measured temperature to obtain the temperature difference. The obtained temperature difference is transmitted to the temperature intelligent control decision model to obtain the downward pressure adjustment amount. The welding process auxiliary computer issues instructions to the parameter adjustment actuator based on the downward pressure adjustment amount. When the temperature difference is within the allowable temperature range, the current shoulder downward pressure parameter remains unchanged. If the temperature difference is positive, it is input into the temperature intelligent control decision model to obtain the adjustment amount to reduce the shoulder downward pressure; when the temperature difference is negative, it is input into the temperature intelligent control decision model to obtain the adjustment amount to increase the shoulder downward pressure.

[0019] Step 4: The parameter adjustment actuator receives the adjustment instructions from the welding process auxiliary computer, corrects the process parameters that have been input into the welding machine CNC system in the friction stir welding machine, and controls the pressure plate on the worktable of the friction stir welding machine to move along the vertical sliding guide rail in the direction of the main shaft through the servo fine-tuning motor, thereby controlling the change of the pressure parameter, forming a cyclic control to achieve the purpose of continuous steady-state welding.

[0020] Step 5: Input the real-time temperature of the welded joint, the process parameters during the welding process, and the weld surface image information processed by the camera software into the performance prediction and evaluation module to obtain the welded joint strength prediction and evaluation results. If the output result is marked as 1, it means that it meets the strength design standard and can be accepted by the subsequent structure; if the output result is 0, it means that the strength is unqualified.

[0021] The beneficial effects of this invention are:

[0022] (1) By monitoring the real-time temperature data of the weld and comparing and controlling the peak temperature of the temperature field, the temperature fluctuation along the weld direction is kept within a controllable range, which greatly enhances the strength and stability of the welded structure.

[0023] (2) Based on the dual-wavelength temperature detection method, the temperature of aluminum alloy stir friction welding with variable emissivity material can be accurately measured under non-contact conditions, which improves the temperature measurement efficiency compared with the contact temperature measurement method. Based on the measured temperature fluctuation results, a temperature intelligent control decision model is introduced, which can realize the precise control of welding parameters based on temperature stability during the welding process.

[0024] (3) Based on the powerful image processing capabilities of convolutional neural networks and combined with prior knowledge of the process, a welding joint strength prediction and evaluation model is constructed through a multimodal fusion learning method, which realizes the online accurate prediction of weld strength and avoids non-destructive testing after welding.

[0025] (4) Through the collaborative work of the vision module, temperature measurement module and welding process auxiliary computer, the image information acquisition, temperature measurement and process parameter recording of the entire welding process can be realized, providing data support for the research and development of intelligent friction stir welding technology. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the overall structure of an online monitoring system for the stir-welding process provided by the present invention;

[0027] Figure 2 This is a bottom view of the pressure plate on the worktable of the friction stir welding machine provided by the present invention;

[0028] Figure 3 This is a flowchart of the online monitoring and parameter control method for the stir-welding process provided by the present invention;

[0029] Figure 4 This is a flowchart of the decision-making adjustment pressure model provided by the present invention;

[0030] Figure 5 This is a flowchart of the performance prediction and evaluation model provided by the present invention;

[0031] Figure 6 This is a flowchart of the training process for the performance prediction and evaluation model provided by this invention.

[0032] The reference numerals in the accompanying drawings include:

[0033] 01-Friction stir welding machine, 08-Spindle, 09-Parameter adjustment actuator, 10-Pressure plate on the worktable of friction stir welding machine, 11-Sliding vertical guide rail, 14-LCD display, 15-Stirring head, 16-Welding process auxiliary computer, 17-Workpiece, 23-Guide rod, 24-Bolt, 26-Spherical clamp, 28-Universal bamboo joint tube, 31-Electric air pump, 32-Fiber optic infrared thermometer, 34-PVC conduit, 36-CCD industrial camera, 38-Pressure sensor, 41-Scale dial one, 42-Scale dial two, 44-Slider, 45-Bolt, 46-Slide rail device, 48-Angle scale. Detailed Implementation

[0034] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0035] like Figures 1 to 6 As shown, an online monitoring system for the stir welding process includes a vision module, a temperature measurement module, a feedback execution module, and a performance prediction and evaluation module. The vision module is used for image information acquisition and transmitting the acquired image information to the feedback execution module. The temperature measurement module is used for real-time temperature acquisition and transmitting the real-time temperature to the feedback execution module. The feedback execution module is used for material welding and process parameter adjustment and integrates all the acquired information and transmits it to the performance prediction and evaluation module. The performance prediction and evaluation module analyzes the obtained information to obtain the weld joint strength prediction and evaluation results.

[0036] The vision module includes a CCD industrial camera 36 and a spherical clamp 26. The CCD industrial camera 36 is mounted below the pressure plate 10 on the worktable of the friction stir welding machine via the spherical clamp 26. The spherical clamp 26 and the CCD industrial camera 36 are coaxially mounted with a first scale 41 and a second scale 42. The second scale 42 is closer to the CCD industrial camera 36. The spherical clamp includes a ball head, a guide rod, and a bolt. The guide rod passes through the center of the ball head, and the ball head and the guide rod are fixed by bolts. To prevent the weld seam image from being obscured and information loss, two CCD industrial cameras 36 are used to acquire image information. The CCD industrial camera 36 is mounted on the lower right side of the pressure plate 10 on the worktable of the friction stir welding machine. The ball head is fastened to the guide rod 23 via bolt 24 to adjust the vertical height and plane angle of the camera. The scale 42 at the connection between the CCD industrial camera 36 and the ball head is used to fine-tune the camera angle. By adjusting the guide rod 23 and the scale 41, it is ensured that the CCD industrial camera 36 can acquire weld seam image information without obstruction.

[0037] The temperature measurement module includes a fiber optic infrared thermometer 32, a universal bamboo-joint tube 28, an electric air pump 31, a slide rail device 46, and a dual-wavelength infrared temperature measurement model set in a computer. The slide rail device 46 is installed below the pressure plate 10 on the friction stir welding machine workbench. An angle scale 48 is engraved on the slide rail device 46. Two sliders 44 are bolted into the slide rail device 46. Two fiber optic infrared thermometers 32 are respectively installed below the sliders 44 through the universal bamboo-joint tube 28. The slide rail device 46 adopts a U-shaped guide rail design and is connected to the pressure plate 10 on the friction stir welding machine workbench by bolts. The angle scale 48 on the slide rail device 46 is adjusted... The dial 48 ensures that the laser of the fiber optic infrared thermometer is aligned with the forward or backward side of the stirring head 15, 5mm away from the outer side of the shoulder. The electric air pump 31 is placed on one side of the workbench of the friction stir welding machine. One end of the soft PVC conduit 34 is connected to the air outlet of the electric air pump 31, and the other end of the soft PVC conduit 34 is placed next to the fiber optic infrared thermometer 32. The soft PVC conduit 34 guides the compressed gas to the optical lens of the fiber optic infrared thermometer to clean the lens dust and improve the accuracy of the measured radiation intensity. The fiber optic infrared thermometer 32 inputs the measured two wavelength radiation intensities into the dual-wavelength infrared temperature measurement model to obtain the measured temperature.

[0038] The dual-wavelength infrared thermometry model, taking into account the change in emissivity of aluminum alloy with temperature, employs a dual-wavelength thermometry method when measuring the temperature of the weld joint. This method, combined with the selection of the optimal welding peak temperature T, is based on Wien's displacement law. b is the Wien displacement constant. Peak wavelengths λ1 and λ2 are selected. In this process, in order to ensure that the two selected peak wavelengths are similar and to avoid the infrared absorption bands of water vapor and carbon dioxide, so as to improve the accuracy and reliability of the measurement, this method uses two infrared bandpass filters with center transmission wavelengths of λ1 and λ2, respectively, and FWHM bandwidth of μ.

[0039]

[0040] In the formula, DR(T) is the measured radiation intensity of the target, ε(T) is the temperature-dependent emissivity of the aluminum alloy, λ is the central projection wavelength, and τ λ Let C1 be the spectral transmittance of the infrared bandpass filter, C2 be the first and second radiation constants, and F be the detector offset.

[0041]

[0042] In the formula, ΔDR(T) represents the differential radiation intensity under two different temperature measurement targets, ε(T) is the temperature-dependent emissivity of the aluminum alloy, λ is the central projection wavelength, and τ... λ C1 and C2 are the spectral transmittance of the infrared bandpass filter, and C1 and C2 are the first and second radiation constants, respectively. ambient For ambient temperature,

[0043]

[0044] In the formula, D(T) is the measured target temperature, and r ΔDR (T) represents the ratio of the average radiation intensity of the two bands, C f As the correction factor is a constant, r ε (T) represents the ratio of the average emissivity of the two bands, λ1 and λ2 are the center transmission wavelengths of the two infrared bandpass filters, μ is the FWHM bandwidth, and τ λ1 τ λ2 These represent the spectral transmittances of two infrared bandpass filters, C1 and C2 are the first and second radiation constants, and T is the spectral transmittance of the two filters. ambient The ambient temperature.

[0045] The fiber optic infrared thermometer uses formula (1) to calculate the difference between the radiation intensity measured during welding and the radiation intensity measured at ambient temperature, thus obtaining formula (2). This process corrects the influence of detector drift and ensures the accuracy of temperature measurement. Next, the ratio of the two radiation intensities after center transmission wavelength correction is calculated to eliminate the influence of non-wavelength factors, thus obtaining the dual-wavelength temperature measurement formula (3).

[0046] As can be seen from the thermometric equation (3), the temperature is only related to the radiation intensity of two adjacent wavelengths. The corresponding radiation intensity can be measured on the right side of the equation using a light infrared thermometer. To obtain the numerical solution of the thermometric equation, an aluminum alloy plate was placed in a resistance heating device for heating and temperature measurement. The temperature range was 100–500℃, and the temperature interval was T. step =2~20K. The temperature of the aluminum alloy sheet was measured using thermocouples, while two fiber optic infrared thermometers simultaneously measured the sheet's radiation intensity, obtaining a series of sheet temperature and corresponding radiation intensity data. Subsequently, a polynomial numerical model relating the sheet's radiation intensity and temperature was established using the least squares method, and the cost function was minimized using calculus to solve for the model parameters, resulting in the final infrared thermometry model T=α3×(r ΔDR (T) / C f r ε (T)) 3 +α2×(r ΔDR (T) / C f r ε (T)) 2 +α1×(r ΔDR (T) / C f r ε (T))+α0 where r ΔDR (T) represents the ratio of the average radiation intensity of the two bands, C f As the correction factor is a constant, r ε(T) represents the ratio of the average emissivity of the two bands, and α3, α2, α1, and α0 are the model parameters. α3, α2, α1, and α0 are the parameters of the polynomial numerical model. A series of temperature and radiation intensity data are obtained through temperature measurement. The parameters of the numerical model are then calculated using the least squares method. The radiation intensity at the unknown temperature is then input into the polynomial numerical model to calculate the unknown temperature value. During the welding process, the measured radiation intensity of the two wavelengths is input into the infrared thermometry model to achieve accurate measurement of the weld joint temperature.

[0047] The feedback execution module includes a friction stir welding machine 01, a pressure sensor 38, a friction stir welding machine worktable, a process auxiliary computer 16, and an LCD display 14. The temperature measurement module transmits the real-time temperature of the stirring head 15 at a distance of 5mm from the shoulder to the process auxiliary computer 16. The process auxiliary computer 16 contains a temperature intelligent control decision model, which compares the real-time temperature with the peak and minimum temperatures of the temperature cycle curve in the process auxiliary computer's database. The temperature cycle curve represents the temperature fluctuation range of the welded joint when obtaining an ideal weld surface. The peak and minimum temperatures are the boundaries of this temperature cycle curve. If the real-time temperature is within the range of the peak and minimum temperatures, welding continues. If the real-time temperature is not within the range of the peak and minimum temperatures, the temperature difference is calculated between the real-time temperature and the closest value between the peak and minimum temperatures. This temperature difference is then transmitted to the temperature intelligent control decision model to obtain the downward pressure adjustment amount. The parameter adjustment actuator 09 includes a servo fine-tuning motor and a sliding vertical guide rail 11. The sliding vertical guide rail 11 is located above the worktable of the friction stir welding machine. One side of the pressure plate 10 on the worktable of the friction stir welding machine is installed in the sliding vertical guide rail 11. A servo fine-tuning motor is provided on the side where the pressure plate 10 on the worktable of the friction stir welding machine connects to the sliding vertical guide rail 11 to control the pressure plate on the worktable of the friction stir welding machine to move up and down in the sliding vertical guide rail 11. The parameter adjustment actuator 09 is electrically connected to the friction stir welding machine 01 and the process auxiliary computer 16. The process auxiliary computer 16 issues instructions to the parameter adjustment actuator according to the downward pressure adjustment amount: when the temperature difference of the welding area is within the allowable temperature range, the current shoulder downward pressure parameter remains unchanged; if the temperature difference of the welding area is positive, the temperature difference value is input into the temperature intelligent control decision model to obtain the adjustment amount to reduce the shoulder downward pressure; when the temperature difference of the welding area is negative, the temperature difference value is input into the temperature intelligent control decision model to obtain the adjustment amount to increase the shoulder downward pressure. Pressure sensor 38 is installed at the connection point between the welding machine spindle 08 and the stirring head 15 fixture. It is connected to the process auxiliary computer 16 via a pressure data signal line to measure the stirring head pressure information and transmit this information to the process auxiliary computer 16. LCD display 14 is installed on the friction stir welding machine 01 and electrically connected to the process auxiliary computer 16. If the pressure value exceeds a critical threshold, LCD display 14 will display an alarm signal. LCD display 14 is used to display information such as the current axial pressure of the stirring head, real-time temperature, and the adjustment amount of the welding shoulder pressure.

[0048] The temperature intelligent control decision model employs a backpropagation (BP) neural network structure with two hidden layers. Its input layer parameters include the rotation speed, welding speed, and pressure adjustment of the friction stir welding machine, as well as the temperature difference in the welding area. The input layer has 4 neurons. The output layer parameter is the pressure adjustment amount, and the output layer has 1 neuron. The selected activation function is the tanh function, and the loss function is the cross-entropy loss function. To construct the temperature intelligent control decision model dataset, numerous welding experiments were conducted on aluminum alloys using different welding process parameters. During the welding process, the difference between the theoretical pressure adjustment and the empirically experienced pressure adjustment was recorded as the labels for the training samples.

[0049] The performance prediction and evaluation module employs a multimodal fusion learning method to integrate information from different modalities, aiming to solve the problem of weld joint strength evaluation. Specifically, for the surface texture information images of the weld joint acquired by the vision module, a convolutional neural network is used for feature extraction to capture the texture information of the image. Temperature information, welding process parameters, and image feature information acquired by the temperature measurement module are used to form the input vector of a fully connected neural network based on a multimodal feature fusion method. The output layer uses a sigmoid activation function to restrict the output to between 0 and 1. A stochastic gradient descent algorithm is used to optimize the loss function, thereby achieving intelligent evaluation of weld joint strength. To construct the weld strength evaluation model dataset, numerous welding experiments were conducted, covering different rotation speeds, welding rates, and pressure. During the experiments, weld images, weld joint temperatures, and tensile strength of the samples were collected under different process parameters. Samples meeting the corresponding strength requirements were labeled as 1, and samples not meeting the requirements were labeled as 0.

[0050] A method for online monitoring and parameter control of the friction stir welding process in an online monitoring system is adopted, which specifically includes the following steps:

[0051] Step 1: Assemble the online monitoring system for the stir welding process. Place the workpiece 17 on the welding machine worktable, adjust the vision module to ensure that the CCD industrial camera 36 can acquire weld seam image information without obstruction, and adjust the position of the fiber optic infrared thermometer 32 to ensure the accuracy of its distance measurement target. The vision module collects weld seam image information, the temperature measurement module collects real-time temperature information, the pressure sensor collects stirring head pressure information, and the welding equipment collects welding process information. Transmit the obtained weld seam image information, real-time temperature information, stirring head pressure information, and welding process information to the process auxiliary computer.

[0052] Step 2: Set parameters such as resolution, frame rate, and exposure time in the camera software on the computer to perform system debugging and ensure that the images captured by the CCD industrial camera 36 meet expectations, and check the image quality. Test the triggering and synchronization functions of the CCD industrial camera 36 to ensure reliability in the automated control process. Configure the fiber optic infrared thermometer, including adjusting the temperature display unit, and use the aiming laser on the fiber optic infrared thermometer 32 to ensure alignment with the temperature measurement target, realizing the automatic triggering mechanism for temperature measurement. Select welding process parameters according to the type and thickness of the aluminum alloy sheet, the parameters of the welding tools, and the process requirements to perform friction stir welding.

[0053] Step 3: The temperature measurement module transmits the real-time temperature to the process auxiliary computer 16. The real-time temperature is compared with the peak and minimum temperatures of the temperature cycle curve in the database of the process auxiliary computer 16. If the real-time temperature is within the range of the peak and minimum temperatures, welding continues. If it is not within the allowable range, if the measured temperature is greater than the peak temperature, the peak temperature is subtracted from the measured temperature to obtain the temperature difference; if the measured temperature is less than the minimum temperature, the minimum temperature is subtracted from the measured temperature to obtain the temperature difference. The obtained temperature difference is transmitted to the temperature intelligent control decision model to obtain the downward pressure adjustment amount. The welding process auxiliary computer issues instructions to the parameter adjustment actuator 09 according to the downward pressure adjustment amount. When the temperature difference is within the allowable temperature range, the current shoulder downward pressure parameter remains unchanged. If the temperature difference is positive, it is input into the temperature intelligent control decision model to obtain the adjustment amount to reduce the shoulder downward pressure. When the temperature difference is negative, it is input into the temperature intelligent control decision model to obtain the adjustment amount to increase the shoulder downward pressure.

[0054] Step 4: The parameter adjustment actuator 09 receives the adjustment instructions from the welding process auxiliary computer 16, corrects the process parameters already input into the welding machine CNC system in the friction stir welding machine 01, and controls the pressure plate on the worktable of the friction stir welding machine to move along the vertical sliding guide rail 11 in the direction of the main shaft 08 through the servo fine-tuning motor, thereby controlling the change of the pressure parameter, forming a cyclic control to achieve the purpose of continuous steady-state welding.

[0055] Step 5: Input the acquired temperature data, process parameters during welding, and weld surface image information into the performance prediction and evaluation module to obtain the weld joint strength prediction and evaluation results. If the output result is marked as 1, it indicates that the strength design standard is met, and subsequent structural acceptance can proceed; if the output result is 0, it indicates that the strength is unqualified. This provides data support for subsequent structural inspection and also provides a basis for whether repair welding is needed.

Claims

1. An online monitoring system for friction stir welding processes, characterized in that, It includes a vision module, a temperature measurement module, a feedback execution module, and a performance prediction and evaluation module. The vision module is used for image information acquisition and transmitting the acquired image information to the feedback execution module. The temperature measurement module is used for real-time temperature acquisition and transmitting the real-time temperature to the feedback execution module. The feedback execution module is used for material welding and process parameter adjustment and integrates all the acquired information and transmits it to the performance prediction and evaluation module. The performance prediction and evaluation module analyzes the obtained information to obtain the weld joint strength prediction and evaluation results. The feedback execution module includes a friction stir welding machine, a friction stir welding machine worktable, a pressure sensor, a process auxiliary computer, a parameter adjustment actuator, and an LCD display. The temperature measurement module transmits the measured real-time temperature to the process auxiliary computer, which contains a temperature intelligent control decision model. The real-time temperature is compared with the peak and minimum temperatures of the temperature cycle curve in the process auxiliary computer's database. The temperature cycle curve represents the temperature fluctuation range of the welded joint when obtaining an ideal weld surface. The peak and minimum temperatures are the boundaries of this temperature cycle curve. If the real-time temperature is within the range of the peak and minimum temperatures, welding continues. If the real-time temperature is outside the range, if the real-time temperature is greater than the peak temperature, the temperature difference is obtained by subtracting the peak temperature from the real-time temperature; if the real-time temperature is less than the minimum temperature, the difference is obtained by subtracting the peak temperature from the real-time temperature. The temperature difference is obtained by subtracting the minimum temperature from the real-time temperature. This temperature difference is then transmitted to the temperature intelligent control decision model to obtain the downward pressure adjustment amount. The parameter adjustment actuator includes a servo fine-tuning motor and a sliding vertical guide rail. The sliding vertical guide rail is located above the worktable of the friction stir welding machine. One side of the pressure plate on the worktable of the friction stir welding machine is installed in the sliding vertical guide rail. A servo fine-tuning motor is located on the side of the pressure plate on the worktable of the friction stir welding machine that connects to the sliding vertical guide rail. The servo fine-tuning motor controls the pressure plate on the worktable of the friction stir welding machine to move up and down in the sliding vertical guide rail. The parameter adjustment actuator is electrically connected to the friction stir welding machine and the process auxiliary computer. The process auxiliary computer issues instructions to the parameter adjustment actuator based on the downward pressure adjustment amount: when the real-time temperature of the welding area is within the range of the peak temperature and the minimum temperature, the current shoulder downward pressure parameter remains unchanged. If the temperature difference in the welding area is positive, the temperature difference is input into the intelligent temperature control decision model to obtain the adjustment amount to reduce the shoulder pressure. When the temperature difference in the welding area is negative, the temperature difference is input into the intelligent temperature control decision model to obtain the adjustment amount to increase the shoulder pressure. The pressure sensor is installed at the connection between the welding machine spindle and the stirring head fixture to measure the stirring head pressure information. It is also connected to the process auxiliary computer via a pressure data signal line to transmit the stirring head pressure information to the process auxiliary computer. The LCD is installed on the friction stir welding machine and electrically connected to the process auxiliary computer. If the pressure value exceeds the theoretical maximum critical pressure of heat input in the process auxiliary computer, the LCD will display an alarm signal. The LCD is used to display the current stirring head axial pressure, real-time temperature, and welding shoulder pressure adjustment information.

2. The online monitoring system for friction stir welding process according to claim 1, characterized in that, The vision module includes a CCD industrial camera and a spherical clamp. The CCD industrial camera is mounted under the pressure plate on the worktable of the friction stir welding machine via the spherical clamp. A first scale and a second scale are coaxially mounted at the connection position between the spherical clamp and the CCD industrial camera. The second scale is closer to the side of the CCD industrial camera. The spherical clamp includes a ball head, a guide rod, and a bolt. The guide rod passes through the center of the ball head, and the ball head and the guide rod are fixed together by the bolt. There are two CCD industrial cameras, which are mounted on the lower right side of the pressure plate on the worktable of the friction stir welding machine.

3. The online monitoring system for friction stir welding process according to claim 2, characterized in that, The temperature measurement module includes a fiber optic infrared thermometer, a universal bamboo-joint tube, an electric air pump, a slide rail device, and a dual-wavelength infrared temperature measurement model installed in the process auxiliary computer. The slide rail device is installed below the pressure plate on the friction stir welding machine's worktable. An angle scale is engraved on the slide rail device, and two sliders are bolted into it. Two fiber optic infrared thermometers are installed below the sliders, passing through the universal bamboo-joint tube. The slide rail device adopts a U-shaped guide rail design and is connected to the pressure plate on the welding machine's worktable by bolts. The electric air pump is placed on one side of the friction stir welding machine's worktable. One end of a soft PVC conduit is connected to the air outlet of the electric air pump, and the other end of the soft PVC conduit is placed next to the fiber optic infrared thermometer. The fiber optic infrared thermometer inputs the measured two-wavelength radiation intensity into the dual-wavelength infrared temperature measurement model to obtain the measured temperature.

4. The online monitoring system for friction stir welding process according to claim 3, characterized in that, The dual-wavelength infrared thermometry model is a formula model obtained by minimizing the cost function using calculus and solving for the model parameters. The formula is as follows: In the formula The ratio of the average radiation intensity of the two bands. The correction factor takes the value of a constant. The ratio of the average emissivity of the two bands. Set parameters for the model, where T is the measured temperature.

5. The online monitoring system for friction stir welding process according to claim 4, characterized in that, The intelligent temperature control decision model adopts a neural network prediction model, which contains a BP neural network structure with two hidden layers. Its input layer parameters include the rotation speed, welding speed, pressure, and temperature difference of the welding area of ​​the friction stir welding machine. The number of neurons in the input layer of the neural network prediction model is 4. The output layer parameter of the neural network prediction model is the pressure adjustment amount, and the number of neurons in the output layer is 1. The selected activation function is the tanh function, and the loss function is the cross-entropy loss function.

6. The online monitoring system for friction stir welding process according to claim 5, characterized in that, The performance prediction and evaluation module employs a multimodal fusion learning method, which integrates information from different modalities. Specifically, for the surface texture image of the weld joint acquired by the vision module, a convolutional neural network is used for feature extraction to capture the image's texture information. Temperature information acquired by the temperature measurement module, welding process parameters, and image feature information are combined using a multimodal feature fusion method based on splicing and fusion to form the input vector of a fully connected neural network. The output layer uses a sigmoid activation function, limiting the output to between 0 and 1. A stochastic gradient descent algorithm is used to optimize the loss function, and a GA optimization algorithm is used to optimize the initial weights and thresholds of the convolutional neural network. Classification accuracy is used as the fitness function. Through selection, crossover, and compilation operations, the optimized initial weights and thresholds of the convolutional neural network are obtained. To construct the weld strength evaluation model dataset, numerous welding experiments are conducted, covering different rotation speeds, welding rates, and pressure amounts.

7. The method for online monitoring and parameter control of friction stir welding process in an online monitoring system for friction stir welding process as described in claim 6, characterized in that, Includes the following steps: Step 1: Assemble the online monitoring system for friction stir welding process. Place the workpiece on the welding machine workbench, adjust the vision module to ensure that the CCD industrial camera can acquire weld seam image information without obstruction, and adjust the position of the fiber optic infrared thermometer to ensure the accuracy of its distance measurement target. The vision module collects weld seam image information, the temperature measurement module collects real-time temperature information, the pressure sensor collects stirring head pressure information, and the welding equipment collects welding process information. Transmit the obtained weld seam image information, real-time temperature information, stirring head pressure information, and welding process information to the process auxiliary computer. Step 2: Set the resolution, frame rate, and exposure time in the camera software of the process-aided computer, perform system debugging to ensure that the image captured by the CCD industrial camera meets expectations, check the quality of the captured image, test the triggering and synchronization functions of the CCD industrial camera to ensure reliability in the automated control process, set up the fiber optic infrared thermometer, including adjusting the temperature display unit, and use the aiming laser on the fiber optic infrared thermometer to ensure alignment with the temperature measurement target, realize the automatic triggering mechanism for temperature measurement, and select welding process parameters according to the type and thickness of the aluminum alloy sheet, the parameters of the welding tools, and the process requirements for friction stir welding. Step 3: The temperature measurement module transmits the real-time temperature to the process auxiliary computer. The real-time temperature is compared with the peak temperature and minimum temperature of the temperature cycle curve in the process auxiliary computer database. If the real-time temperature is within the range of the peak temperature and minimum temperature, welding continues. If it is not within the allowable range, if the measured temperature is greater than the peak temperature, the peak temperature is subtracted from the measured temperature to obtain the temperature difference. If the measured temperature is less than the minimum temperature, the minimum temperature is subtracted from the measured temperature to obtain the temperature difference. The obtained temperature difference is transmitted to the temperature intelligent control decision model to obtain the downward pressure adjustment amount. The process auxiliary computer issues instructions to the parameter adjustment actuator according to the downward pressure adjustment amount. When the temperature difference is within the allowable temperature range, the current shoulder downward pressure parameter remains unchanged. If the temperature difference is positive, the temperature difference is input into the temperature intelligent control decision model to obtain the adjustment amount to reduce the shoulder pressure. When the temperature difference is negative, the temperature difference is input into the temperature intelligent control decision model to obtain the adjustment amount to increase the shoulder pressure. Step 4: The parameter adjustment actuator receives the adjustment instructions from the process auxiliary computer, corrects the process parameters that have been input into the welding machine CNC system in the friction stir welding machine, and controls the pressure plate on the worktable of the friction stir welding machine to move along the vertical sliding guide rail in the direction of the main shaft through the servo fine-tuning motor, thereby controlling the change of the pressure parameter, forming a cyclic control to achieve the purpose of continuous steady-state welding. Step 5: Input the real-time temperature of the welded joint, the process parameters during the welding process, and the weld surface image information processed by the camera software into the performance prediction and evaluation module to obtain the welded joint strength prediction and evaluation results. If the output result is marked as 1, it means that it meets the strength design standard and can be accepted by the subsequent structure; if the output result is 0, it means that the strength is unqualified.

Images