Friction stir solid phase additive manufacturing intelligent control method based on data driving

Abstract

The invention relates to the technical field of friction stir solid-phase additive manufacturing, and discloses a friction stir solid-phase additive manufacturing intelligent control method based on data driving, and the method comprises the steps: obtaining key data parameters in an additive manufacturing process in real time through a multi-source data collection system, and carrying out the comprehensive analysis of collected data through a data fusion algorithm, and high-precision monitoring data is generated, so that the data accuracy and reliability are remarkably improved in the process, and a solid foundation is provided for subsequent intelligent prediction. The machine learning model is utilized to intelligently predict high-precision monitoring data, and abnormal conditions are found in advance, so that accurate optimization and adjustment of process control parameters are ensured. And meanwhile, process control parameters are dynamically adjusted according to monitoring data and prediction results through a closed-loop multi-level feedback control mechanism, changes in the manufacturing process are quickly responded, defects are effectively restrained, and finally high-quality forming of friction-stir solid-phase additive manufacturing is achieved through stable control over the friction-stir solid-phase additive manufacturing process.

Description

Technical Field

[0001] The present invention belongs to the technical field of friction stir solid phase additive manufacturing, and in particular relates to a data-driven intelligent control method for friction stir solid phase additive manufacturing. Background Art

[0002] As a key area of ​​advanced manufacturing, additive manufacturing (AM) has been widely adopted in high-end manufacturing sectors such as aerospace, automotive, and medical. Friction stir solid-phase additive manufacturing (FSAM), with its advantages of high material utilization, environmentally friendly process, and stable performance, has gradually become a hot topic in research and application. This technology enables the efficient forming of large structural parts, significantly improving manufacturing efficiency and the consistency of material properties. However, with the increasing demand for applications, the demand for precise control and intelligent capabilities of the AM process is increasing.

[0003] However, existing control methods have limited ability to monitor dynamic changes in the additive manufacturing process. They primarily rely on local data from a single fiber optic sensor, making it difficult to achieve comprehensive, real-time perception of the temperature field, mechanical field, and motion state. This makes it difficult to precisely control heat input and material plastic deformation. Furthermore, traditional process control parameter adjustments often rely on manual experience and lack the ability to deeply mine and optimize multi-source data in real time, making it difficult to dynamically adapt to complex process conditions. Feedback control is often static or single-step, unable to dynamically adjust based on real-time data. This can easily lead to problems such as uneven forming quality and difficulty repairing internal defects. Summary of the Invention

[0004] The purpose of the present invention is to overcome the above problems and provide a data-driven intelligent control method for stir friction solid-phase additive manufacturing, aiming to solve the problems of insufficient real-time monitoring capabilities, low level of intelligent data analysis and imperfect feedback control mechanism in the existing technology.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions: The present invention provides a data-driven intelligent control method for friction stir solid-phase additive manufacturing, comprising the following steps: Real-time data collection of additive manufacturing process parameters through multi-source data acquisition system; Inputting the collected data parameters into a data processing module through a real-time data stream for preprocessing to obtain preprocessed data; Use multi-source data fusion algorithm to conduct comprehensive analysis on pre-processed data to generate high-precision monitoring data; Using a machine learning model, support vector machine (SVM), to perform intelligent prediction on the high-precision monitoring data, predict potential abnormal situations, and obtain prediction results; Based on the high-precision monitoring data and prediction results, process control parameters are dynamically adjusted through closed-loop multi-level feedback control.

[0006] A further improvement of the present invention is that the multi-source data acquisition system includes a distributed optical fiber sensor, an infrared thermal imager, a high-speed camera and a laser vision sensor, wherein the sampling frequency of the multi-source data acquisition system is 10-100 Hz.

[0007] A further improvement of the present invention is that the data parameters in the additive manufacturing process are divided into process control parameters and forming quality parameters; the process control parameters include travel speed, feed speed, water cooling temperature, upset force, pressure and torque; and the forming quality parameters include additive body temperature and additive body width and height.

[0008] A further improvement of the present invention is that the preprocessing includes removing outliers, signal smoothing, standardization processing and multi-source data time synchronization.

[0009] A further improvement of the present invention is that the multi-source data fusion algorithm utilizes weighted least squares method and Kalman filtering algorithm to optimize the fusion result and obtain high-precision monitoring data.

[0010] A further improvement of the present invention is that the method of optimizing the fusion result by using the weighted least squares method and the Kalman filter algorithm to obtain high-precision monitoring data is: A linear regression analysis model was constructed using the weighted least squares method to analyze the correlation between rotation speed, travel speed, feed speed, water cooling temperature, upset force, pressure, torque, and the temperature, width, and height of the additive body. n influencing factors (n ≥ 3) were selected, and the influencing parameters related to the influencing factors were calculated and selected as the important influencing parameters. Kalman filtering is used to divide the preprocessed data into process control parameters for state prediction and forming quality parameters for observation and correction. Then, self-supervised high-precision reconstruction is completed through a continuous prediction-update mechanism within the discrete-time state space framework, and dynamically optimized high-precision monitoring data is generated in real time, providing accurate input for closed-loop multi-level feedback control.

[0011] A further improvement of the present invention is that the method for intelligently predicting high-precision monitoring data using a machine learning model support vector machine (SVM) is: Support vector machine (SVM) is used to train high-precision monitoring data, and a prediction model is established to predict potential abnormal situations. The control threshold of the prediction model is set during the experiment to determine whether the current process control parameters need to be adjusted.

[0012] A further improvement of the present invention is that when the support vector machine (SVM) predicts that a process control parameter needs to be adjusted, a PID controller is used to adjust the influencing parameter in real time, and the PID controller automatically adjusts the control signal according to the error signal; The error signal includes a difference between an actual width and a target width or a difference between an actual additive body temperature and a target additive body temperature, ensuring that the process control parameter remains within a range of 10% of the target parameter.

[0013] A further improvement of the present invention is that the closed-loop multi-level feedback control includes real-time feedback, process feedback and global feedback.

[0014] A further improvement of the present invention is that the real-time feedback adjusts the water cooling temperature, rotation speed, rotation speed, feed speed and upset force based on real-time data; the process feedback optimizes heating and pressure control based on the distribution of thermal field and mechanical field; and the global feedback is combined with quality assessment to dynamically optimize the control strategy.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a data-driven intelligent control method for stir friction solid-phase additive manufacturing. The method acquires key data parameters in the additive manufacturing process in real time through a multi-source data acquisition system, and performs comprehensive analysis on these data in combination with a data fusion algorithm to generate high-precision monitoring data. This process significantly improves the accuracy and reliability of the data, providing a solid foundation for subsequent intelligent prediction. By using a machine learning model to intelligently predict high-precision monitoring data, abnormal situations can be discovered in advance, thereby ensuring the accurate optimization and adjustment of process control parameters. At the same time, through a closed-loop multi-level feedback control mechanism, process control parameters are dynamically adjusted according to monitoring data and prediction results, and changes in the manufacturing process are quickly responded to, effectively reducing the generation of defects. Ultimately, through the stable control of the stir friction solid-phase additive manufacturing process, high-quality forming of stir friction solid-phase additive manufacturing is achieved; not only manufacturing efficiency is improved, but also the quality and reliability of the product are significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are for illustrative purposes only and are not intended to limit the scope of the present invention in any way. In addition, the shapes and proportional dimensions of the various components in the drawings are merely schematic and are used to help understand the present invention, and are not intended to specifically limit the shapes and proportional dimensions of the various components of the present invention.

[0017] Figure 1 Schematic diagram of the flow of the data-driven intelligent control method for friction stir solid-phase additive manufacturing of the present invention; Figure 2 This is a diagram of the data-driven intelligent control device for friction stir solid-phase additive manufacturing of the present invention; Figure 3 A schematic flow chart of a data-driven intelligent control method for friction stir solid-phase additive manufacturing provided by an embodiment of the present invention; Figure 4 A temperature-time curve comparison diagram of the additive process with and without the intelligent control method in an embodiment of the present invention; Figure 5 A comparison of the macroscopic morphologies of the additive process using and not using the intelligent control method according to an embodiment of the present invention; Figure 6 This is a comparison chart of mechanical properties when the intelligent control method is used in the additive process of an embodiment of the present invention.

[0018] Among them: 1. Processing platform; 2. Additive body; 3. Water cooling device; 4. Infrared thermal imager; 5. Laser vision sensor; 6. Machine head; 7. Upsetting force fiber optic sensor and speed fiber optic sensor; 8. Push rod; 9. Torque fiber optic sensor; 10. Temperature fiber optic sensor; 11. High-speed camera; 12. Data collector; 13. Computer; 14. Laser vision sensor generates image; 15. Displacement distance fiber optic sensor; 16. Substrate; 17. Pressure fiber optic sensor. DETAILED DESCRIPTION

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0020] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort shall fall within the scope of protection of the present invention.

[0021] The present invention is described in further detail below with reference to the accompanying drawings: like Figure 1 As shown, the present invention provides a data-driven intelligent control method for stir friction solid-phase additive manufacturing, comprising the following steps: S1, real-time collection of data parameters during the additive manufacturing process through a multi-source data acquisition system; S2, inputting the collected data parameters into a data processing module through a real-time data stream for preprocessing to obtain preprocessed data; S3, uses multi-source data fusion algorithm to conduct comprehensive analysis on the pre-processed data to generate high-precision monitoring data; S4, using a machine learning model support vector machine (SVM) to perform intelligent prediction on the high-precision monitoring data, predict potential abnormal conditions, and obtain a prediction result; S5, dynamically adjusting process control parameters through closed-loop multi-level feedback control based on the high-precision monitoring data and prediction results.

[0022] The multi-source data acquisition system described in S1 includes distributed fiber optic sensors, infrared thermal imagers, high-speed cameras and laser vision sensors. The sampling frequency of the multi-source data acquisition system is 10-100 Hz, which is used to collect process control parameters and forming quality parameters in the additive manufacturing process in real time. The process control parameters include travel speed, feed speed, water cooling temperature, upset force, pressure and torque; the forming quality parameters include additive body temperature and additive body width and height, so as to dynamically monitor the thermal field, mechanical response, motion state and quality characteristics.

[0023] Specifically, the data parameters collected in real time by the multi-source data acquisition system during the additive manufacturing process are: Temperature parameters: dynamic temperature distribution along the center axis of the square bar and the dynamic temperature distribution of the additive area; Mechanical parameters: upsetting force and torque of the machine head; Motion state parameters: rotation speed, feed speed and travel speed of the machine head; Forming quality parameters: the temperature, width and height of the additive body obtained by the visual sensor.

[0024] In the step S2, the collected data parameter set is input into the data processing module through the real-time data stream for preprocessing to obtain preprocessed data, wherein the preprocessing includes removing outliers, signal smoothing, normalization processing and multi-source data time synchronization. The outlier refers to the data that exceeds 100% of the previous and next data, and the signal smoothing refers to interpolating the adjacent mean values ​​of the outlier data.

[0025] It should be noted that based on the temperature fiber optic sensor data, the data processing module trains and calibrates the parameters of the infrared thermal imager, and dynamically adjusts the weights of the temperature fiber optic sensor and the infrared thermal imager in the model. The position close to the temperature fiber optic sensor continuously adjusts the infrared thermal imager image target parameters in the additive process according to the data obtained by the temperature fiber optic sensor, including emissivity, reflection temperature, atmospheric temperature, relative humidity, target distance, external optical temperature and external optical transmittance to improve the monitoring accuracy and adaptability of the infrared thermal imager, ensure the consistency and fusion effect of multi-source data, and provide high-quality support for real-time analysis and feedback control in the subsequent additive manufacturing process.

[0026] In the step S3, a multi-source data fusion algorithm is used to perform a comprehensive analysis on the pre-processed data, and the weighted least squares method and Kalman filter algorithm are used to optimize the fusion results to generate high-precision monitoring data.

[0027] The method for obtaining high-precision monitoring data is: A linear regression analysis model was constructed using the weighted least squares method to analyze the correlation between rotation speed, travel speed, feed speed, water cooling temperature, upset force, pressure, torque, and the temperature, width, and height of the additive body. n influencing factors (n ≥ 3) were selected, and the influencing parameters related to the influencing factors were calculated and selected as the important influencing parameters. Kalman filtering is used to divide the preprocessed data into process control parameters for state prediction and forming quality parameters for observation and correction. Then, self-supervised high-precision reconstruction is completed through a continuous prediction-update mechanism within the discrete-time state space framework, and dynamically optimized high-precision monitoring data is generated in real time, providing accurate input for closed-loop multi-level feedback control.

[0028] Specifically, the weighted least squares method is used to construct a linear regression analysis model to analyze the correlation between process control parameters and forming quality parameters: Assume that the process control parameter matrix is , the forming quality parameter vector is , the weight matrix is : in Formula 1 Where, For the The first feature data , among which The value range is ≥10, The value range is ≥1000; Formula 2 Where, For the forming quality parameter characteristics, The value range of is ≥1; Formula 3 Where, For the The weight corresponding to each forming quality parameter feature, The value range is ≥10; The solution of the weighted least squares method is: Formula 4 Where, is the regression coefficient vector, is the process control parameter matrix, is the forming quality parameter matrix, is the weight matrix, is transposed; Then calculate the residual and update the weights , iterate to weight convergence; set up For the centered process control parameter matrix, calculate the covariance matrix : Formula 5 Where, is the process control parameter matrix The number of feature data in Then the correlation coefficient matrix for: Formula 6 Where, For the Features and The covariance of the features, For the The variance of the parameters, For the The variance of the parameters; Then select 3 influencing factors and eliminate the covariance The redundant influencing factors are calculated again, and the influencing parameters related to the influencing factors are selected as the important influencing parameters.

[0029] A high-precision data optimization model is established using Kalman filtering, and the initialization model is: Formula 7 Where, is the Kalman gain matrix, is the covariance matrix, is the observation matrix, is the transpose, is the noise covariance matrix of the observed data; Formula 8 Where, is the optimized estimate after one observation update, that is, the process control parameter value matrix after one observation update, is the initial state estimate, that is, the initial value matrix of the process control parameters, is the Kalman gain matrix, is the actual observation value for the first time, is the observation matrix; in, Formula 9 Formula 10 According to the preprocessed data and actual observation values, the estimation error is reduced, and finally a high-precision parameter state is obtained, and a dynamically optimized .

[0030] In step S4, the method for intelligently predicting high-precision monitoring data using the machine learning model support vector machine (SVM) is as follows: Use support vector machines (SVMs) to train high-precision monitoring data, establish a prediction model to predict potential abnormal conditions, set the control threshold of the prediction model during the experiment, and determine whether the current process control parameters need to be adjusted; The support vector machine SVM deduces the decision function classification formula: Formula 11 Where, is the new sample data, is the Lagrange multiplier, is the Gaussian kernel function, is the bias term; when When , the parameters need to be adjusted; when No parameters need to be adjusted.

[0031] When the SVM predicts that process control parameters need to be adjusted, the PID controller is used to adjust the key influencing parameters in real time. The PID controller automatically adjusts the control signal based on the error signal, such as the difference between the actual width and the target width or the difference between the actual and target additive body temperatures, to ensure that the process control parameters fluctuate within 10% of the target parameters. The control formula of the PID is: Formula 12 Where, is the control signal, is the error, that is, the difference between the target value and the current value, is the proportionality coefficient, is the integration coefficient, is the differential coefficient, It's an error.

[0032] Potential abnormal conditions include temperature runaway, dimensional deformation, and equipment capacity overload.

[0033] In the S5 step, the rotation speed, feed speed and water cooling temperature are dynamically adjusted through closed-loop multi-level feedback control based on high-precision monitoring data and prediction results to ensure the dynamic stability and performance consistency of the additive manufacturing process. Specifically, the closed-loop multi-level feedback control includes real-time feedback, process feedback and global feedback. The real-time feedback adjusts the rotation speed and feed speed based on real-time data; the process feedback optimizes the water cooling temperature and feed speed based on the distribution of thermal field and mechanical field; and the global feedback is combined with quality assessment to dynamically optimize the rotation speed, water cooling temperature and feed speed control strategies.

[0034] like Figure 2 The stir friction solid phase additive manufacturing intelligent control device shown in the figure includes a processing platform 1, an additive body 2, a water cooling device 3, an infrared thermal imager 4, a laser vision sensor 5, a machine head 6, an upsetting force fiber optic sensor and a speed fiber optic sensor 7, a push rod 8, a torque fiber optic sensor 9, a temperature fiber optic sensor 10, a high-speed camera 11, a data collector 12, a computer 13, a laser vision sensor to generate an image 14, a displacement distance fiber optic sensor 15, a substrate 16 and a pressure fiber optic sensor 17; wherein, below the processing platform 1 is a pressure fiber optic sensor platform composed of multiple pressure fiber optic sensors 17, and the substrate 16 is installed on the upper surface of the processing platform 1, and the substrate 16 has a vibration fiber optic sensor and an array distribution Temperature fiber optic sensor, the additive body 2 is produced above the substrate 16, the additive body 2 is continuously rotated by the head 6, and then the material inside the head 6 is squeezed out under the drive of the push rod 8. The head 6 is equipped with a water cooling device 3, a torque fiber optic sensor 9, a temperature fiber optic sensor 10 and a displacement distance fiber optic sensor 15, and the push rod 8 is equipped with an upsetting force fiber optic sensor and a speed fiber optic sensor 7. An infrared thermal imager 4, a laser vision sensor 5 and a high-speed camera 11 are installed around the additive body 2. The infrared thermal imager 4 and the laser vision sensor generate images 14 and the high-speed camera 11 transmits real-time data to the data collector 12, and the data collector 12 transmits the data directly to the computer 13 for processing, analysis and command issuance.

[0035] Example 1 like Figure 3 As shown, this embodiment provides a data-driven intelligent control method for stir friction solid-phase additive manufacturing, which mainly includes the following steps: Step 1: Set initial parameters by inputting the macroscopic characteristic length, width, and height of the additive body 2, as well as the expected placement position of the additive body 2. Based on the macroscopic position and expected placement position of the additive body 2, the starting position coordinates, the rotation speed and travel speed of the machine head 6, the feed speed of the push rod 8, and the spatial displacement of the entire processing path are automatically generated; Step 2: The spindle executes the instruction and sends the initial parameters to the equipment spindle. The head 6 and the processing platform 1 determine the movement direction according to the processing path. The head 6, the processing platform 1 and the push rod 8 complete the instruction according to the set rotation speed, travel speed and feed speed; Step 3: Multi-source data acquisition. The collectors at each position collect corresponding data, including cooling water temperature monitoring data from the water cooling device 3, temperature monitoring data of the additive body 2 from the infrared thermal imager 4, path of the machine head 6 and shape of the additive body 2 displayed by the image 14 generated by the laser vision sensor, upsetting force and feed speed data of the push rod 8 from the upsetting force fiber optic sensor and the speed fiber optic sensor 7, torque data of the machine head 6 from the torque fiber optic sensor 9, temperature data of the machine head 6 and its internal material from the temperature fiber optic sensor 10, macroscopic morphology data of the additive body 2 from the high-speed camera 11, height data of the additive body from the displacement distance sensor 15, vibration data of the processing platform 1 and bottom temperature data of the additive body 2 from the substrate 16 equipped with a vibration fiber optic sensor and an array distributed temperature fiber optic sensor, and pressure data on the additive body 2 from the pressure fiber optic sensor 17. Step 4: Data fusion and feature extraction. Through a multimodal algorithm that fuses spatiotemporal features, combined with weighted least squares and Kalman filtering algorithms for optimization, the main influencing factors are found to be the additive body temperature and upset force. The main influencing parameters with the greatest correlation with the main influencing factors are the water cooling temperature, rotation speed, and feed speed. High-precision monitoring data is then obtained. Step 5, machine learning prediction, input the extracted features into the machine learning prediction module, and judge whether the experimental parameters need to be adjusted according to the current processing state of the additive body temperature, upsetting force and forming quality parameters. When , the parameters need to be adjusted; when When the need for process control parameter adjustment is predicted, the PID controller is used to adjust the main influencing parameters in real time. The PID controller automatically adjusts the control signal based on the error signal, such as the difference between the actual width and the target width or the difference between the actual and target additive body temperatures, to ensure that the process control parameters operate within the optimal range.

[0036] Step 6: Exception handling and dynamic adjustment. During the multi-source data acquisition process, if an exception detection is triggered, the system immediately starts real-time exception alarm and adaptive adjustment. The system first monitors the data for abnormalities and performs mean interpolation on false error data. For true errors, the algorithm generates target-oriented parameter adjustments. The user then reviews and confirms the parameters, and then issues new execution parameters. The updated parameters drive the device action through the spindle execution instruction, and re-enters the multi-source data acquisition stage to form a closed loop.

[0037] Example 2 In this embodiment, a 6061 aluminum alloy metal plate with a size of 300 mm × 150 mm × 10 mm is used as the substrate, and a 6061 aluminum alloy square bar of the same material is used as the raw material. The feed bar size is 20 mm × 20 mm × 20 mm. The head is made of H13 alloy steel. The experiment is carried out using a stir friction solid phase additive manufacturing intelligent control device combined with an intelligent control method.

[0038] In the intelligent control method, through real-time data acquisition and closed-loop feedback regulation, the process control parameters are dynamically adjusted to set the rotation speed to 600 rpm (dynamic adjustment range of ±200 rpm), the feed speed to 30 mm / min (dynamic adjustment range of ±10 mm / min), the travel speed to 300 mm / min (dynamic adjustment range of ±100 mm / min), and the upset force to 10 kN (dynamic adjustment range of ±3 kN), with a single-direction travel of 200 mm.

[0039] like Figure 4 As shown in the figure, the additive manufacturing process is divided into three stages: preheating, additive manufacturing, and cooling. During the preheating stage, the intelligent control method uses an infrared thermal imager to work in conjunction with a temperature fiber optic sensor arranged on the substrate to monitor the temperature distribution in real time. Based on the data from the temperature fiber optic sensor, the data processing module trains and calibrates the infrared thermal imager parameters to improve the monitoring accuracy and adaptability of the infrared thermal imager and ensure the consistency and fusion effect of multi-source data. By adjusting the rotation speed and feed speed in real time, the forging force is quickly and evenly increased. Then, as the temperature rises, the material softens, the forging force begins to decrease, and the temperature rises to the target temperature, and then the additive manufacturing stage begins.

[0040] like Figure 4 As shown, during the additive manufacturing stage, the intelligent control method controls the maximum temperature fluctuation range within ±20°C by dynamically adjusting each process control parameter, that is, the real-time temperature of the additive body at the head, ensuring the consistency of the thermal influence of the material. During the additive manufacturing stage, the intelligent control method uses data fusion and prediction algorithms to dynamically adjust the multi-dimensional data collected in real time (rotation speed, upset force, feed speed, travel speed, temperature field distribution, upset force change and path deviation). The optimized rotation speed, upset force, feed speed and travel speed ensure the stability of the processing process, and the temperature fluctuation range is precisely controlled within ±20°C, effectively avoiding overheating or underheating. Thanks to the uniformity of the thermal field and force field, the plastic flow of the material is more stable and the microstructure is denser.

[0041] During the cooling stage, the intelligent control method optimizes the heat dissipation path and dynamically adjusts the cooling rate to quickly and evenly reduce the temperature of the additive part, thus avoiding the local area of ​​the substrate staying in a high temperature state for a long time, thereby effectively suppressing the grain coarsening phenomenon. Figure 5As shown in (a), the macroscopic surface of the cooled additive part is flat and smooth, without ripples, cracks or peeling, and the roughness is significantly reduced. Figure 6 As can be seen from the results, the tensile strength and elongation of the additively manufactured body under the intelligent control method are significantly improved, and the overall mechanical properties are excellent. This not only proves that there are no obvious internal defects, but also highlights the outstanding advantages of intelligent control methods in the field of additive manufacturing.

[0042] Comparative Example 1 In this comparative example, a 6061 aluminum alloy metal plate with a size of 300 mm × 150 mm × 10 mm is used as the substrate, a 6061 aluminum alloy square bar of the same material is used as the raw material, the feed bar size is 20 mm × 20 mm × 20 mm, the machine head is made of H13 alloy steel, and the experiment is carried out using a stir friction solid phase additive manufacturing intelligent control device combined with an uncontrolled method.

[0043] In the uncontrolled method, the process control parameters are fixed, and the rotation speed is set to 600 rpm, the feed speed is set to 30 mm / min, the travel speed is set to 300 mm / min, and the total travel is 200 mm. Due to the fixed nature of the process control parameters, the additive manufacturing process lacks the ability to adjust the thermal field and force field in real time, such as Figure 4 Temperature-time curve, when there is no temperature control method, the thermal field fluctuation range is as high as ±100℃, the heating rate in the preheating stage is unstable, and overheating or underheating occurs in local areas during the additive stage, which directly affects the plastic flow of the material and leads to uneven microstructure. Figure 5 As shown in (b), the macroscopic morphology of the additive body produces significant defects. In addition, the fixed rotation speed, feed speed and travel speed fail to adapt to the dynamically changing process conditions, further exacerbating the stress concentration and crack propagation in the processing path. Figure 5 As shown in (b), the surface quality of the additive part is poor, with obvious cracks, peeling and rough ripples, which significantly reduces the forming quality. Figure 6 As shown in Figure 3, the mechanical properties of the additive body without control method are relatively poor, which further illustrates the existence of internal defects and uneven microstructure distribution.

[0044] Many embodiments and applications beyond the examples provided will be apparent to those skilled in the art upon reading the foregoing description. Therefore, the scope of the present teachings should be determined not with reference to the foregoing description, but rather with reference to the preceding claims, along with the full scope of equivalents to which such claims are entitled. For the purpose of completeness, all articles and references, including the disclosures of patent applications and publications, are incorporated herein by reference. The omission of any aspect of the subject matter disclosed herein from the preceding claims is not a disclaimer of such subject matter, nor should it be interpreted that the applicants did not consider such subject matter to be part of the disclosed inventive subject matter.

[0045] The above content is a further detailed description of the present invention, and it cannot be considered that the specific implementation methods of the present invention are limited to these. For ordinary technicians in the technical field to which the present invention belongs, they can make several simple deductions or substitutions without departing from the concept of the present invention, which should be regarded as belonging to the scope of protection of the present invention determined by the submitted claims.

Claims

1. A data-driven intelligent control method for friction stir solid-phase additive manufacturing, characterized in that: The following steps are involved: Real-time data collection of additive manufacturing process parameters through multi-source data acquisition system; Inputting the collected data parameters into a data processing module through a real-time data stream for preprocessing to obtain preprocessed data; Use multi-source data fusion algorithm to conduct comprehensive analysis on pre-processed data to generate high-precision monitoring data; Using a machine learning model, support vector machine (SVM), to perform intelligent prediction on the high-precision monitoring data, predict potential abnormal situations, and obtain prediction results; Based on the high-precision monitoring data and prediction results, process control parameters are dynamically adjusted through closed-loop multi-level feedback control.

2. The data-driven intelligent control method for friction stir solid-phase additive manufacturing according to claim 1, characterized in that: The multi-source data acquisition system includes a distributed optical fiber sensor, an infrared thermal imager, a high-speed camera and a laser vision sensor, wherein the sampling frequency of the multi-source data acquisition system is 10-100 Hz.

3. The data-driven intelligent control method for friction stir solid-phase additive manufacturing according to claim 1, characterized in that: The data parameters in the additive manufacturing process are divided into process control parameters and forming quality parameters; the process control parameters include travel speed, feed speed, water cooling temperature, upset force, pressure and torque; the forming quality parameters include additive body temperature and additive body width and height.

4. The data-driven intelligent control method for friction stir solid-phase additive manufacturing according to claim 1, characterized in that: The preprocessing includes outlier removal, signal smoothing, standardization and multi-source data time synchronization.

5. The data-driven intelligent control method for friction stir solid-phase additive manufacturing according to claim 1, characterized in that: The multi-source data fusion algorithm utilizes weighted least squares method and Kalman filtering algorithm to optimize the fusion result and obtain high-precision monitoring data.

6. The data-driven intelligent control method for friction stir solid-phase additive manufacturing according to claim 5, characterized in that: The method of using the weighted least squares method and the Kalman filter algorithm to optimize the fusion result and obtain high-precision monitoring data is as follows: A linear regression analysis model was constructed using the weighted least squares method to analyze the correlation between rotation speed, travel speed, feed speed, water cooling temperature, upset force, pressure, torque, and the temperature, width, and height of the additive body. n influencing factors (n ≥ 3) were selected, and the influencing parameters related to the influencing factors were calculated and selected as the important influencing parameters. Kalman filtering is used to divide the important influencing parameters after preprocessing into process control parameters for state prediction and forming quality parameters for observation and correction. Then, self-supervised high-precision reconstruction is completed through a continuous prediction-update mechanism within the discrete time state space framework, and dynamically optimized high-precision monitoring data is generated in real time, providing accurate input for closed-loop multi-level feedback control.

7. The data-driven intelligent control method for friction stir solid-phase additive manufacturing according to claim 1, characterized in that: The method for intelligently predicting high-precision monitoring data using a machine learning model support vector machine (SVM) is as follows: Support vector machine (SVM) is used to train high-precision monitoring data, and a prediction model is established to predict potential abnormal situations. The control threshold of the prediction model is set during the experiment to determine whether the current process control parameters need to be adjusted.

8. The data-driven intelligent control method for friction stir solid-phase additive manufacturing according to claim 7, characterized in that: When the support vector machine (SVM) predicts that the process control parameters need to be adjusted, the PID controller is used to adjust the influencing parameters in real time. The PID controller automatically adjusts the control signal according to the error signal. The error signal includes a difference between an actual width and a target width or a difference between an actual additive body temperature and a target additive body temperature, ensuring that the process control parameter remains within a range of 10% of the target parameter.

9. The data-driven intelligent control method for friction stir solid-phase additive manufacturing according to claim 1, characterized in that: The closed-loop multi-level feedback control includes real-time feedback, process feedback and global feedback.

10. The data-driven intelligent control method for friction stir solid-phase additive manufacturing according to claim 9, characterized in that: The real-time feedback adjusts the water cooling temperature, rotation speed, rotation speed, feed speed and upset force based on real-time data; the process feedback optimizes heating and pressure control based on the distribution of thermal and mechanical fields; and the global feedback dynamically optimizes the control strategy in combination with quality assessment.

Images