A friction welding machine slide table control method and device

Abstract

The present application relates to the technical field of friction welding machine, and discloses a friction welding machine slide table control method and device, the method comprises the following steps: real-time acquisition of the movement data of the friction welding machine stirrer head, the establishment of the stirrer head dynamics model, based on the movement data and the dynamics model, the output of the preliminary control strategy through the optimal linear quadratic regulation control algorithm, based on the movement data, the prediction of the future movement data of the stirrer head through the prediction control algorithm, the optimization of the preliminary control strategy, the control signal to the servo motor of the slide table controls the servo motor, the difference between the real-time stirrer head movement data and the prediction result is taken as feedback data, and the control strategy parameters are adjusted, the present application takes the difference between the real-time movement data of the stirrer head and the prediction result as feedback, forms a closed-loop control mechanism, continuously monitors and optimizes the control parameters, improves the position control precision of the stirrer head, effectively solves the response lag problem in high-speed welding, ensures the weld seam trajectory consistency, and avoids position deviation.

Description

Technical Field

[0001] The present invention relates to the technical field of friction welding machines, and in particular to a method and device for controlling a sliding table of a friction welding machine. Background Art

[0002] Friction welding technology is gaining widespread attention in the welding industry due to its unique advantages, particularly for industrial applications requiring high-precision and high-strength connections. However, existing friction welding machine control systems lack precise displacement control, resulting in inaccurate movement of the stirring head during welding, which in turn affects weld quality.

[0003] Existing friction welding machines typically use open-loop or simple closed-loop control systems to control the displacement of the slide, driving the stirring head for welding. Due to the lack of precise feedback and control mechanisms, these existing solutions struggle to ensure the stability and accuracy of the stirring head during welding. This can lead to problems such as uneven welds and unstable weld strength, especially during high-speed welding and welding complex structures. Summary of the Invention

[0004] The purpose of the present invention is to provide a friction welding machine slide control method and device, which solves the problem that in the prior art, friction welding machines usually adopt open-loop or simple closed-loop control systems, lack accurate feedback and control mechanisms, and are prone to uneven welds and unstable welding strength.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: A friction welding machine slide control method, comprising the following steps:

[0006] Real-time collection of movement data of the friction welding machine stirring head, including movement direction and movement speed, and establishment of a stirring head dynamics model;

[0007] Based on the motion data and dynamic model, the optimal control input is calculated through the optimal linear quadratic control algorithm, and the preliminary control strategy is output;

[0008] Based on the movement data, the future movement data of the mixing head is predicted through the predictive control algorithm, and the control input is adjusted based on the prediction results to optimize the preliminary control strategy;

[0009] Generate a control signal based on the optimized control strategy, and send the control signal to the servo motor of the slide to control the servo motor;

[0010] The difference between the real-time stirring head movement data and the predicted results is used as feedback data to adjust the control strategy parameters to adapt to different welding conditions.

[0011] Preferably, the stirring head dynamic model includes:

[0012] ;

[0013] in, is the mass of the slide; is the acceleration of the slide; is the damping coefficient; is the speed of the slide; is the elastic coefficient; is the position of the slide; Feedback force is a control force signal generated based on the difference between movement data and predicted results.

[0014] Preferably, the cost function of the optimal linear quadratic regulation control algorithm is used to minimize the weighted sum of position error and velocity error, and the cost function is:

[0015] ;

[0016] in, is the position vector of the slide; is the velocity vector of the slide; and is a positive definite matrix, which represents the degree of penalty for controlling position error and velocity error; is the cost function, and the goal is to minimize the cost function to obtain the optimal control input.

[0017] Preferably, the secondary regulation control algorithm generates an optimal control input based on a dynamic model, and adjusts the speed and position of the slide by a servo motor to control the movement of the stirring head.

[0018] Preferably, the objective function of the predictive control algorithm is:

[0019] ;

[0020] in, The future time step predicted for the system The position of the mixing head; is the time step target trajectory; is the time step Control input; and is the weight matrix, which represents the relative importance of state and control input; is the transpose operator; is the predicted number of steps.

[0021] Preferably, the predictive control algorithm is based on the maximum velocity and acceleration of the slide when optimizing the control input, and adjusts the preliminary control strategy in real time to minimize the control error at future moments.

[0022] Preferably, the movement data also includes real-time detection of the material and thickness of the workpiece, and real-time updating of the stirring head dynamics model according to the detection results.

[0023] Preferably, the collecting of movement data of the friction welding machine stirring head further includes, after collecting the movement data, preprocessing the movement data, wherein the preprocessing includes filtering the movement data to eliminate sensor noise, remove outliers, and align the movement data with the coordinate system and time.

[0024] Preferably, adjusting the control strategy parameters to adapt to different welding conditions specifically includes:

[0025] Calculate the error between the real-time stirring head movement data and the predicted results,

[0026] Based on the error, the weight matrices in the optimal linear quadratic regulation control algorithm and the predictive control algorithm are dynamically adjusted.

[0027] A friction welding machine slide control device, comprising:

[0028] The acquisition module collects the movement data of the friction welding machine stirring head in real time based on sensors and establishes a dynamic model of the stirring head;

[0029] The optimal control calculation module calculates the optimal control input based on the motion data and dynamic model using the optimal linear quadratic regulation control algorithm and outputs the preliminary control strategy;

[0030] A predictive control module is used to predict the future movement data of the mixing head through a predictive control algorithm based on the movement data, and adjust the control input based on the prediction result to optimize the preliminary control strategy;

[0031] A control signal generating module is used to generate a control signal based on the optimized control strategy and send the control signal to the servo motor of the slide to drive the servo motor to control the movement of the slide;

[0032] The feedback regulation module is used to compare the difference between the real-time movement data of the stirring head and the predicted result, and adjust the control strategy parameters according to the difference to adapt to different welding conditions.

[0033] In summary, the present invention includes at least one of the following beneficial technical effects:

[0034] 1. This invention uses the difference between the real-time stirring head movement data and the predicted results as feedback data to adjust the control strategy of the slide's servo motor, thereby forming a closed-loop control mechanism. This can continuously monitor the actual operating status of the stirring head and dynamically adjust the control parameters to form a continuous, high-speed feedback loop. This technical solution significantly improves the precision of stirring head position control, ultimately effectively improving weld uniformity and overall quality. Compared with traditional open-loop or low-sensitivity closed-loop systems, this solution overcomes the problem of control system response lag, especially in high-speed welding tasks, and can better ensure high trajectory consistency and solve the problem of weld position offset.

[0035] 2. The present invention introduces a predictive control algorithm to dynamically estimate the position of the stirring head at future moments, which can greatly improve the stability and control accuracy of the system and effectively suppress trajectory drift. Therefore, it solves the problem in the prior art that the control system can only passively respond to the current state, which is prone to control errors when the welding speed changes or the turning points of the path.

[0036] 3. The present invention establishes a dynamic model of the stirring head based on the material and thickness of the workpiece, and the control strategy weight can self-adjust according to the error between the actual and predicted results, dynamically correcting the dynamic model. This technical solution greatly enhances the system's adaptability to complex welding conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 Schematic diagram of the method flow of the present invention;

[0038] Figure 2 Schematic diagram of the device system architecture of the present invention. DETAILED DESCRIPTION

[0039] The following is combined with Figure 1 , the present invention is described in further detail.

[0040] The present invention provides a friction welding machine slide control method, comprising the following steps:

[0041] Real-time collection of movement data of the friction welding machine stirring head, including movement direction and movement speed, and establishment of a stirring head dynamics model;

[0042] In this embodiment, to ensure the accuracy and dynamic adaptability of the subsequent control strategy generation, it is necessary to accurately perceive the real-time motion state of the stirring head and establish an accurate dynamic model based on this. To this end, this embodiment introduces a preprocessing module centered around data acquisition and modeling during the initial phase of method execution. This module not only provides the basic physical modeling support for the control algorithm but also significantly improves control accuracy and the robustness of system response through data cleaning and structural optimization.

[0043] Specifically, a sensor system mounted on the friction welding machine's slide collects real-time data on the movement of the stirrer. This data includes at least two dimensions: the direction of the stirrer's movement during the welding process, and its speed. The sensor can be a laser displacement meter, photoelectric encoder, or inertial measurement unit (IMU) to ensure high resolution and low latency of the collected data.

[0044] Generally, raw sensor data is affected by factors such as environmental noise, welding heat interference, and equipment vibration. Therefore, the present invention introduces a preprocessing mechanism. Specifically:

[0045] First, multiple high-precision sensors installed on the slide or stirring head collect movement data of the stirring head during the welding process. This data includes two basic dimensions: movement direction and movement speed. The movement direction can be measured by a three-axis accelerometer and an angular velocity sensor, while the movement speed can be obtained by a laser interferometer or linear grating ruler.

[0046] In one possible implementation, the collected raw data needs to undergo a series of preprocessing operations before being input into the system model to enhance its stability and credibility. Generally, the preprocessing process includes but is not limited to the following steps:

[0047] Filter the sensor data, such as using bandpass filters, Kalman filters, and other technologies, to effectively eliminate high-frequency fluctuations introduced by sensor noise or environmental interference;

[0048] Implement outlier removal mechanisms, such as using sliding mean and standard deviation analysis within a certain time window to identify and remove data points that deviate from statistical characteristics;

[0049] Perform coordinate system and time alignment operations to synchronize multi-source data under a unified spatial and temporal reference, ensuring the consistency and temporality of modeling data.

[0050] Specifically, after the data has been pre-processed, it will enter the stage of dynamic modeling of the stirring head. In the present invention, the following dynamic model is used to model the motion process of the sliding table.

[0051] As an option, some embodiments further incorporate a material sensing module. This module detects the material parameters and thickness of the workpiece to be welded in real time, dynamically updating the constructed stirring head dynamics model. In one possible implementation, ultrasonic sensors or electromagnetic induction sensors are used to identify the material type (e.g., aluminum alloy, stainless steel) and thickness (e.g., 3mm, 5mm), and the detection results are fed back to the model parameter calculation module in real time.

[0052] Specifically, based on the above pre-processed movement data, the present invention establishes the following stirring head dynamic model:

[0053] ;

[0054] in, is the mass of the slide; is the acceleration of the slide; is the damping coefficient; is the speed of the slide; is the elastic coefficient; is the position of the slide; Feedback force is a control force signal generated based on the difference between movement data and predicted results.

[0055] In some embodiments, this dynamic model provides a theoretical basis for the subsequent calculation of optimal control inputs. This model can be used to derive a control strategy based on control algorithms (such as linear quadratic control or model predictive control) to precisely control the slide servo motor drive and achieve coordinated regulation of the stirrer's position and speed.

[0056] It should be noted that in practical applications, the parameters of the dynamic model can be obtained through experimental identification or online adaptive updating to adapt to the influence of different welding conditions, equipment status or material changes.

[0057] In summary, this embodiment not only builds a systematic data support mechanism from the perspective of mobile data collection and processing, but also combines physical modeling methods to form a complete slide control modeling link, laying a solid technical foundation for the subsequent generation of optimal control strategies.

[0058] Based on the motion data and dynamic model, the optimal control input is calculated through the optimal linear quadratic control algorithm, and the preliminary control strategy is output;

[0059] In this embodiment, after completing the data collection and preprocessing of the friction welding machine's stirring head movement and constructing a dynamic model that conforms to actual operating conditions, the present invention further introduces an optimal linear quadratic regulator (LQR) algorithm to effectively regulate the slide motion and generate a preliminary control strategy for the current system state. This control algorithm plays a central role in connecting physical modeling and control execution in this method. The control inputs it calculates are not only stable and real-time, but also possess the fundamental functionality to be further iteratively optimized by the predictive control algorithm.

[0060] Specifically, first, based on the slide dynamics model obtained in the previous step, in order to facilitate optimal control design, the above second-order differential equation is converted into a standard state space form. In general, the system state vector is defined as:

[0061] ;

[0062] in, is the position of the slide; is the speed of the slide;

[0063] The control input is defined as the servo motor control signal , represents the actual force applied to the slide (or motor command). Thus, the standard linear system model is obtained:

[0064] ;

[0065] Among them, the system matrix and the input matrix According to the kinetic model 、 、 Its specific structure can be obtained by engineering identification or automatic calculation of simulation system.

[0066] To achieve optimal control of the slide motion state, this embodiment minimizes the weighted sum of position error and velocity error based on the following cost function:

[0067] ;

[0068] in, is the position vector of the slide; is the velocity vector of the slide; and is a positive definite matrix, which represents the degree of penalty for controlling position error and velocity error; is the cost function, and the goal is to minimize the cost function to obtain the optimal control input.

[0069] The cost function aims to minimize the weighted sum of position and velocity errors over the entire time domain while balancing the energy consumption of the control input;

[0070] Specifically, if:

[0071] ;

[0072] ;

[0073] in, Represents the control weight of the position error; Represents the control weight of the speed error; Indicates the intensity of suppression of control signal energy. These parameters can be adjusted according to the system response characteristics.

[0074] Then, by solving the Algebraic-Riccati-Equation, the optimal feedback gain matrix can be obtained: , and then construct the linear feedback control law:

[0075] ;

[0076] In this embodiment, the calculated control input It is output as a preliminary control strategy and sent to the servo motor via the control signal generation module to drive the slide to move, thereby achieving indirect control of the stirring head.

[0077] As an option, under some specific working conditions, the control strategy can integrate hardware constraints such as the maximum speed and maximum acceleration of the slide, and adjust 、 The weight parameters of the matrix ensure that the control input not only meets the performance index requirements but also takes into account the system execution capability.

[0078] Specifically, in one possible implementation, the system can automatically adjust the weight coefficient according to dynamic working conditions such as workpiece thickness and welding speed requirements, thereby improving the adaptability of the control strategy. For example, when welding thick workpieces, appropriately increase the , in order to strengthen the suppression of position error; and in the high-speed welding of thin workpieces, it can improve , to strengthen the control of speed response.

[0079] Therefore, this control strategy not only has good robustness and engineering feasibility, but also provides high-quality control input initial values ​​for subsequent predictive control and feedback regulation modules, thereby ensuring the coordination and high-precision operation of the entire slide control system.

[0080] Based on the movement data, the future movement data of the mixing head is predicted through the predictive control algorithm, and the control input is adjusted based on the prediction results to optimize the preliminary control strategy;

[0081] In this embodiment, after the initial control strategy is generated, considering the uncertainty and dynamic changes in the working conditions during welding, such as material thermal expansion, friction coefficient fluctuations, and structural transient response, relying solely on feedback control input calculated based on the current state may not be able to meet the requirements of higher-precision trajectory control. In order to improve the control system's ability to predict future states and further optimize control performance, the present invention introduces a predictive control algorithm into the control system. Based on the currently collected motion data and the established dynamic model, this algorithm estimates the future motion state of the stirring head in advance and modifies the initial control strategy accordingly, thereby significantly improving the continuity and stability of the welding trajectory and the forward-looking control response.

[0082] In this embodiment, predictive control adopts the Model-Predictive-Control (MPC) framework. Its core lies in: using a sliding time window to predict the state evolution trend of the system over a period of time in the future, and based on the prediction, determining the current control input by solving the optimization problem, so that the future state is as close to the target trajectory as possible.

[0083] In general, the predictive control module uses the current movement data as input, including the current position of the stirrer, the current speed, the welding path direction, and the target trajectory point set. The future state is deduced through the state space model, and the following objective function (cost-function) is used as the optimization criterion:

[0084] ;

[0085] in, The future time step predicted for the system The position of the mixing head; is the time step target trajectory; is the time step Control input; and is the weight matrix, which represents the relative importance of state and control input; is the transpose operator; is the predicted number of steps.

[0086] In the specific implementation, the predictive control adopts a recursive method to calculate the state variables The prediction chain is constructed by performing step-by-step estimation and using each state update as the initial condition for the next step. All control inputs within the prediction window are uniformly solved by an optimization solver (such as a QP solver).

[0087] Specifically, predictive control not only considers trajectory errors but also introduces the physical operating limitations of the slide. For example, in some embodiments, to prevent the slide from accelerating or decelerating too quickly or exceeding the limit, the system adds the following constraints:

[0088] Maximum slide speed constraint:

[0089] ;

[0090] Maximum acceleration constraint of the slide:

[0091] ;

[0092] Maximum control input Amplitude constraints;

[0093] ;

[0094] The above constraints are introduced into the optimization model in the form of linear inequalities to ensure the engineering feasibility and equipment safety of the optimization results.

[0095] In one possible implementation, the predictive control module applies only the first control input and output to the current moment in each control cycle, discarding the remaining prediction results. This strategy, known as receding-horizon-control, ensures that the system optimizes based on the latest state at every moment, improving its dynamic response capabilities.

[0096] As an option, in some embodiments, if there is a significant deviation between the predicted result and the current preliminary control strategy, the system can make a weighted adjustment based on the preset fusion weight. For example, the final control input can be expressed as:

[0097] ;

[0098] in, is the preliminary control strategy previously obtained by the LQR algorithm; The optimized control input for the predictive control module;

[0099] is the weighting coefficient, and its value range is , which can be adjusted dynamically according to the size of the error.

[0100] Through this control input fusion mechanism, the present invention realizes dual regulation based on feedback and prediction, improves the robustness and anti-interference performance of the system, and is particularly suitable for high-quality welding control tasks under complex or dynamically changing working conditions.

[0101] In addition, in order to improve the accuracy of the prediction model, the present invention also supports online correction of model parameters during the control process, for example, by continuously adjusting the system matrix or gain term in the state space model through the deviation between the actual feedback data and the predicted value, thereby realizing model adaptive adjustment.

[0102] In summary, the predictive control method proposed in the present invention not only makes full use of the current motion data and dynamic model, but also integrates the target trajectory, hardware constraints and multi-strategy fusion mechanism, forming a control optimization module with complete structure, multiple strategies and accurate prediction, which provides the friction welding machine slide system with accurate, controllable and future-oriented control path generation capability.

[0103] Generate a control signal based on the optimized control strategy, and send the control signal to the servo motor of the slide to control the servo motor;

[0104] In this embodiment, after optimizing and adjusting the control input based on the predictive control algorithm, to ensure the mixing head accurately tracks the target trajectory, the present invention further introduces a control signal generation and distribution mechanism. This mechanism converts the optimized control input into an electrical control signal executable by the servo drive system and sends it to the servo motor in the slide control system, thereby driving the slide to achieve the corresponding displacement response. This process is the final execution link in the system control chain and is closely logically and functionally coupled with previous modules such as preliminary control strategy generation, predictive control correction, and real-time system status feedback, forming a complete closed-loop control process from state perception, control calculation, and execution feedback.

[0105] Specifically, the optimized control input is typically represented by a digital signal, whose physical meaning typically corresponds to a velocity setpoint, acceleration adjustment, or displacement increment command in the direction of motion of the slide. This input is processed by a digital controller (such as a DSP or embedded master control unit) and converted into a corresponding analog control voltage signal or pulse-width modulation (PWM) signal for the servo drive.

[0106] Typically, a controller includes a control signal scheduling module that outputs updated control commands within each control cycle, based on the real-time sampling period. Within the controller, this module receives the control input from the predictive control algorithm and generates an analog signal via a digital-to-analog converter (DAC) interface. This signal is filtered, buffered, and amplitude-adjusted before being input to the servo drive.

[0107] Specifically, after receiving the control signal, the servo driver applies the corresponding current to the motor according to its internally set control mode (such as speed control, position control, or current control), achieving closed-loop control of torque or speed, thereby driving the mechanical slide. In this embodiment, the control mode is preferably a position-speed composite control mode, in which:

[0108] The upper master control system sets the target displacement or trajectory;

[0109] The control input is provided by the MPC optimization module;

[0110] The underlying servo controller implements real-time response adjustment of the actual motor position and speed.

[0111] In one possible implementation, to ensure the response consistency and time synchronization of control instructions, the system will implement a fixed-frequency interrupt scheduling mechanism in each control cycle to ensure that the control signal can complete the entire process of collection, calculation, conversion and issuance within the sampling period.

[0112] As an option, in some embodiments, in order to prevent motor excitation instability or overload caused by sudden changes in the control signal, the controller is also equipped with a signal smoothing module, which performs slow-changing processing on the control signal through interpolation filtering, sliding average or hysteresis compensation, thereby avoiding mechanical shock to the actuator.

[0113] Specifically, in the signal generation module, the sliding filter algorithm can control the amount of each cycle Compared with the value of the previous period Perform weighted averaging, such as setting weight coefficients , so that the actual signal is sent for:

[0114] ;

[0115] in, , to adjust the degree of dependence of the current instruction on the previous state, thereby controlling the smoothness of signal changes.

[0116] At the system execution level, the control signal is amplified by the driver and applied to the servo motor windings in the form of a constant voltage or pulsed current, generating electromagnetic excitation and causing the rotor to move. The rotor converts rotation into linear motion via a lead screw, slide rails, or linear module, ultimately controlling the two- or three-dimensional movement of the mixing head. This motion is monitored in real time by a position encoder, whose feedback signal is sampled and fed into the upper-level control system to establish a closed-loop regulation pathway.

[0117] In some embodiments, to improve the control system's anti-interference capabilities and stability, the control signal generation module also incorporates mechanisms such as voltage amplitude limiting, slope limiting, and signal discontinuity protection. For example, in high-speed trajectory sections, to prevent oscillation caused by frequent and significant changes in the input control signal, the system can set a maximum rate-of-change threshold to constrain the amplitude of signal changes per cycle.

[0118] Furthermore, to adapt to different types of slide actuators, the present invention also supports a control signal reconstruction mechanism based on device parameters. During the initialization phase, the system reads characteristic data such as the servo motor model, electrical parameters, and torque constant, automatically calculates the mapping between the control signal and the actual displacement, and dynamically matches the signal range, offset, and calibration factor to ensure the accuracy of the control signal output and device compatibility.

[0119] In summary, this invention achieves seamless integration from optimized control input to physical drive execution by building a complete control signal generation and execution mechanism. This not only improves the execution consistency and dynamic response performance of the system, but also provides stable and reliable final-level control guarantees for high-precision trajectory control. This link, as the terminal output interface of the entire control process, plays a key role in achieving accurate movement of the stirring head and precise tracking of the welding path.

[0120] The difference between the real-time stirring head movement data and the predicted results is used as feedback data to adjust the control strategy parameters to adapt to different welding conditions.

[0121] In this embodiment, to achieve adaptive adjustment for different welding conditions, the system uses the error between real-time stirring head movement data and predicted results as a feedback signal to adjust the control strategy parameters. This adjustment process dynamically adjusts key parameters in the control algorithm based on the real-time error calculation and the change in error, specifically the weight matrix in the optimal linear quadratic regulation (LQR) control algorithm and the predictive control algorithm, to achieve optimal control and weld quality.

[0122] Specifically, first, the real-time stirring head movement data is collected and compared with the ideal result predicted by the system, and the error between the two is calculated. This error reflects the deviation between the actual situation in the current welding process and the theoretical expectation. The error is usually calculated using the following formula:

[0123] ;

[0124] in, Indicates error, is the actual position data of the real-time stirring head, is the ideal position data calculated by the prediction model.

[0125] Based on this error, the system will dynamically adjust the weight matrices in the LQR and predictive control algorithms to optimize the control effect. Specifically, the control matrix in the LQR algorithm and the optimization problem in the predictive control algorithm will be updated according to the real-time error feedback. The weight matrix of the LQR controller usually includes the state weight matrix and the control input weight matrix ,These matrices will be dynamically adjusted according to the error feedback.,The optimization goal of the control strategy is to minimize the system error while ensuring the stability and,efficiency of the welding process.

[0126] In one possible implementation, the specific process of dynamically adjusting the weight matrix of the LQR control algorithm can be expressed as follows:

[0127] ;

[0128] in, is the optimized control gain matrix, is the error, is the control input, and is a positive definite matrix, and at every moment , control matrix Will be adjusted based on the current error. It is the upper limit of the optimization time interval, representing the time range of control strategy optimization.

[0129] In addition, the weight matrix of the predictive control algorithm will also be dynamically adjusted based on the size of the error. For example, the system may update the weight matrix in the predictive control using the following formula:

[0130] ;

[0131] in, is the adjusted predictive control weight matrix, is the original predictive control weight matrix, is the adjustment factor, Indicates error. Adjustment factor The selection will be optimized according to the changes in actual welding conditions in order to better adapt to the welding requirements under different working conditions.

[0132] Therefore, through this adjustment process, the system can adapt to different welding conditions in real time, ensuring optimal control throughout the welding process, thereby improving welding quality and efficiency. Generally, as the welding environment or parameters change, the system can adjust control parameters based on real-time feedback to ensure process stability and consistent welding results. Optionally, the feedback control system can also incorporate other sensor data to further enhance its adaptability to complex welding conditions.

[0133] Moreover, the process of adjusting the control strategy not only relies on error calculation, but can also be combined with other auxiliary information, such as welding current, voltage, temperature and other parameters, to further optimize the control strategy and ensure that the welding quality meets the expected standards.

[0134] Furthermore, to further enhance the system's efficiency and intelligence in repetitive welding tasks, a data storage and reuse mechanism has been designed. A wide range of data generated during each welding process, including information on the stirrer's position, welding temperature, current, voltage, welding speed, welding depth, and weld quality evaluation indicators, is collected and stored in a database in real time. This data encompasses not only workpiece geometry and welding parameter settings, but also the system's dynamic response to the welding process, creating a comprehensive dataset of welding conditions.

[0135] When the system detects that the workpiece currently being welded has similar or identical geometric features and material properties to a certain type of workpiece in historical records, it can quickly retrieve and extract the corresponding process parameters and movement path data based on this historical welding data. This path data includes the three-dimensional trajectory of the stirring head during the welding process, the speed change curves of each path segment, and posture change information; welding parameters include key control parameters such as rotational speed, thrust speed, axial pressure, and preheating temperature. This parameter setting method based on historical experience significantly reduces the test verification time required before repeated workpiece welding, improving the system's work efficiency and welding quality consistency.

[0136] Furthermore, to predict the motion trends of the stirrer during the welding process, this embodiment incorporates a machine learning algorithm, specifically a support vector machine (SVM) model. This model uses historical welding process data from similar workpieces as training samples to learn the dynamic behavior patterns of the stirrer under specific working conditions. By constructing an optimal hyperplane in a high-dimensional feature space, the SVM effectively identifies the mapping relationship between input features (such as the current welding process, position information, speed, and historical errors) and the stirrer's motion state.

[0137] In practice, the system first matches the current weld workpiece type and working condition information with a historical database. It then selects the historical data most closely matching the current task as a training set. This data is then used to train a support vector machine model to predict the future motion trends of the stirrer, including its displacement, velocity changes, and posture adjustment direction. The predictions serve as an important reference input for the control system, working in conjunction with a real-time error feedback mechanism to further enhance the control algorithm's responsiveness and stability to the dynamic welding process.

[0138] Through this mechanism, the system not only effectively leverages historical welding experience but also incorporates intelligent prediction capabilities, providing enhanced adaptability and process optimization support for complex welding environments. This approach significantly reduces the need for initial process testing and the operator's reliance on parameter settings, effectively improving the standardization and replicability of welding processes in mass-produced, standardized production scenarios.

[0139] Combined with attachment Figure 2 This embodiment also provides a friction welding machine slide control method, including:

[0140] The acquisition module collects the movement data of the friction welding machine stirring head in real time based on sensors and establishes a dynamic model of the stirring head;

[0141] The optimal control calculation module calculates the optimal control input based on the motion data and dynamic model using the optimal linear quadratic regulation control algorithm and outputs the preliminary control strategy;

[0142] A predictive control module is used to predict the future movement data of the mixing head through a predictive control algorithm based on the movement data, and adjust the control input based on the prediction result to optimize the preliminary control strategy;

[0143] A control signal generation module is used to generate a control signal based on the optimized control strategy and send the control signal to the servo motor of the slide to drive the servo motor to control the movement of the slide;

[0144] The feedback regulation module is used to compare the difference between the real-time stirring head movement data and the predicted results, and adjust the control strategy parameters according to the difference to adapt to different welding conditions.

[0145] In this embodiment, the acquisition module is responsible for collecting real-time motion data of the stirring head during the welding process, mainly including parameter information such as movement direction, speed, and acceleration. To achieve high-precision data acquisition, the system can be equipped with various types of sensors, such as laser rangefinders, photoelectric encoders, and inertial measurement units (IMUs). Before entering the control system, the collected raw data will undergo preprocessing steps such as filtering, outlier removal, and time alignment to improve data quality and stability. In addition, the module also has a material recognition function, which can automatically adjust model parameters according to changes in workpiece material or thickness, providing a more realistic modeling foundation for subsequent control algorithms.

[0146] Based on the real-time data provided by the acquisition module and the dynamic modeling results of the mixing head, the optimal control calculation module uses an optimal linear quadratic regulation (LQR) control algorithm to calculate the optimal control input at the current moment. This algorithm aims to minimize the weighted sum of the mixing head position error and velocity error, ensuring that the control strategy strikes a good balance between accuracy and energy efficiency. The output of this preliminary control strategy is not only responsive but also highly robust, serving as the basic input for the predictive control algorithm and providing a reference for further optimization.

[0147] The predictive control module plays a proactive regulatory role in the control logic. Based on the current system state and modeling results, this module uses a model predictive control (MPC) algorithm to simulate and predict the motion state of the mixing head for several future time steps. It then optimizes the control inputs by combining the system's target trajectory with physical constraints (such as maximum velocity and acceleration). In each control cycle, the system only applies the first input value in the optimal sequence; the remaining values ​​are replaced with each new cycle, achieving rolling optimization. This module significantly enhances the system's stability and accuracy under complex paths or dynamic conditions.

[0148] The control signal generation module converts the optimized control input from the previous step into an electrical control signal to control the servo motor's corresponding action. Signal conversion can be performed using methods such as digital-to-analog conversion, PWM pulse-width modulation, or voltage-current mapping to ensure high compatibility and low-latency response between the control signal and the servo drive. To prevent sudden changes in excitation from impacting the equipment, the system also incorporates a signal smoothing mechanism to limit the rate of change of the input signal, thereby improving control stability and the service life of the mechanical structure.

[0149] The feedback control module provides closed-loop feedback for the entire control process. By comparing the error between the actual motion of the stirrer and the predicted value, the system can evaluate the effectiveness of the current control strategy in real time and fine-tune control parameters, specifically adaptively optimizing the weight matrices in the LQR and MPC algorithms. This mechanism allows the system to automatically adjust control sensitivity and response strength based on different welding conditions (such as material hardness variations and weld structure complexity), improving the overall system's adaptability and control stability.

[0150] In summary, the method proposed in this embodiment achieves a complete closed-loop control process from data perception and intelligent decision-making to precise execution. Its multi-module collaboration not only improves the control accuracy and path consistency of the stirring head during the welding process, but also enhances the system's real-time adaptability to changing working conditions, significantly promoting the development of intelligent and high-performance friction welding equipment.

[0151] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A friction welding machine slide control method, characterized in that: The following steps are involved: Real-time collection of movement data of the friction welding machine stirring head, including movement direction and movement speed, and establishment of a stirring head dynamics model; Based on the motion data and dynamic model, the optimal control input is calculated through the optimal linear quadratic control algorithm, and the preliminary control strategy is output; The cost function of the optimal linear quadratic regulation control algorithm is used to minimize the weighted sum of position error and velocity error. The cost function is: ； in, is the position vector of the slide; is the velocity vector of the slide; and is a positive definite matrix, which represents the degree of penalty for controlling position error and velocity error; is the cost function, and the goal is to minimize the cost function to obtain the optimal control input; Based on the movement data, the future movement data of the mixing head is predicted by a predictive control algorithm, and the control input is adjusted based on the prediction result to obtain an optimized control input; the optimized control strategy is generated by weighted fusion of the preliminary control strategy and the optimized control input; the predictive control algorithm is based on the maximum speed and acceleration of the slide when optimizing the control input; Generate a control signal based on the optimized control strategy, and send the control signal to the servo motor of the slide to control the servo motor; The difference between the real-time stirring head movement data and the predicted results is used as feedback data to adjust the control strategy parameters to adapt to different welding conditions; The adjustment of control strategy parameters to adapt to different welding conditions specifically includes: Calculate the error between the real-time stirring head movement data and the predicted results, Based on the error, the weight matrices in the optimal linear quadratic regulation control algorithm and the predictive control algorithm are dynamically adjusted.

2. A friction welding machine slide control method according to claim 1, characterized in that: The stirring head dynamic model includes: ； in, is the mass of the slide; is the acceleration of the slide; is the damping coefficient; is the speed of the slide; is the elastic coefficient; is the position of the slide; Feedback force is a control force signal generated based on the difference between movement data and predicted results.

3. A friction welding machine slide control method according to claim 1, characterized in that: The quadratic control algorithm generates optimal control inputs based on the dynamic model and adjusts the speed and position of the slide via the servo motor to control the movement of the stirring head.

4. A friction welding machine slide control method according to claim 1, characterized in that: The objective function of the predictive control algorithm is: ； in, The future time steps predicted for the system The position of the mixing head; is the time step target trajectory; is the time step Control input; and is the weight matrix, which represents the relative importance of state and control input; is the transpose operator; is the predicted number of steps.

5. The method for controlling a friction welding machine slide according to claim 1, wherein: The movement data also includes real-time detection of the material and thickness of the workpiece, and real-time updating of the stirring head dynamics model based on the detection results.

6. A friction welding machine slide control method according to claim 1, characterized in that: The collecting of movement data of the friction welding machine stirring head further includes, after collecting the movement data, preprocessing the movement data, wherein the preprocessing includes filtering the movement data to eliminate sensor noise, remove abnormal values, and align the movement data with the coordinate system and time.

7. A friction welding machine slide control device, based on a friction welding machine slide control method according to any one of claims 1 to 6, characterized in that: include: The acquisition module collects the movement data of the friction welding machine stirring head in real time based on sensors and establishes a dynamic model of the stirring head; The optimal control calculation module calculates the optimal control input based on the motion data and dynamic model using the optimal linear quadratic regulation control algorithm and outputs the preliminary control strategy; A predictive control module is used to predict the future movement data of the mixing head through a predictive control algorithm based on the movement data, and adjust the control input based on the prediction result to optimize the preliminary control strategy; A control signal generating module is used to generate a control signal based on the optimized control strategy and send the control signal to the servo motor of the slide to drive the servo motor to control the movement of the slide; The feedback regulation module is used to compare the difference between the real-time movement data of the stirring head and the predicted result, and adjust the control strategy parameters according to the difference to adapt to different welding conditions.

Images