Bimodal fusion mirror image additive manufacturing surface online monitoring system and method

Abstract

The bimodal fusion mirror image additive manufacturing surface online monitoring system comprises a first monitoring module and a second monitoring module which are arranged in a mirror symmetry mode and are the same in structure. The monitoring module comprises a main body plate, and a pressure part, a thermal infrared imager, an industrial camera, a shearing part, a re-conveying part, a tension part, a heat source part, a connecting piece, a fiber placement tail end movement mechanism, a pressure part, a shearing part, a re-conveying part and a tension part which are sequentially arranged on the main body plate. A surface thermal image and a high-definition photo in the mirror image additive manufacturing process are obtained in real time through the thermal infrared imager and the industrial camera, dislocation and wrinkle defects on the surface of the strip are monitored on line through a visual recognition algorithm and a closed-loop control system, process parameters are adjusted in real time, and an infrared method and a visible light method are carried out at the same time and mutually verified.

Description

Technical Field

[0001] The present invention relates to a technology in the field of material processing, specifically to a dual-mode fusion mirror image additive manufacturing surface online monitoring system and method. Background Art

[0002] In the precision process of composite material manufacturing, wire laying technology is crucial to the manufacture of complex structural parts such as aircraft wings. Although thermoplastic wire laying technology provides a solution for high-efficiency additive manufacturing of continuous fibers and has been used in the manufacture of large structural parts, it often relies on laying on the surface of large molds, which makes the manufacturing cost high and the cycle long because large molds need to be designed and manufactured. At the same time, the use of molds also greatly reduces manufacturing flexibility and limits its application in special environments such as space, remote areas, disaster areas, etc. where large molds cannot be manufactured or are difficult to carry. In addition, when the rigid mold and the flexible wire laying head contact each other and move at high speed, it is easy to cause wire laying accuracy errors. Summary of the invention

[0003] In view of the above-mentioned deficiencies in the prior art, the present invention proposes a dual-modal fusion mirror additive manufacturing surface online monitoring system and method, which uses an infrared thermal imager and an industrial camera to obtain real-time surface thermal images and high-definition photos of the mirror additive manufacturing process, and uses a visual recognition algorithm and a closed-loop control system to monitor the misalignment and wrinkle defects on the strip surface online, and adjust the process parameters in real time. The infrared and visible light methods are carried out simultaneously and verify each other.

[0004] The present invention is achieved through the following technical solutions:

[0005] The present invention relates to an online monitoring system for a dual-modal fusion mirror-image additive manufacturing surface, comprising: a first monitoring module and a second monitoring module which are arranged in mirror symmetry and have the same structure, the monitoring module comprising: a main plate and a pressure component, a thermal infrared imager and an industrial camera, a shearing component, a re-feeding component, a tension component, a heat source component, a connecting piece, and a wire laying terminal motion mechanism which are arranged thereon in sequence.

[0006] The industrial camera is an improved part of the device of the present invention. It is installed at the end of the printing using a bracket, and the lens observation direction is perpendicular to the material surface. It is used to capture high-definition images of the printed material surface in real time as a data set for the visual recognition model.

[0007] The pressure component includes: a pressure roller, a supporting device and a first cylinder, wherein: the central axis of the pressure roller is parallel to the front-to-back direction, the supporting device is located on the right side of the pressure roller and connected to it, and the first cylinder is arranged on the main plate and is used to push the pressure roller to the left, cooperate with the pressure roller on the opposite side to squeeze and generate compaction pressure.

[0008] The thermal infrared imager is installed at an angle of 67.5° to the pressure roller, and the industrial camera is installed side by side with the end of the first cylinder and is located above the pressure roller.

[0009] The shearing component comprises: a second cylinder and a blade, wherein: the second cylinder is used to drive the blade to shear the ribbon, and the second cylinder is arranged on the main body plate.

[0010] The heavy-feeding component includes: a driving wheel, a driven wheel, a motor and a third cylinder. The driving wheel and the driven wheel are arranged vertically and parallel to the left and right directions. The motor is used to drive the driving wheel to rotate, and the third cylinder is used to press the driven wheel onto the driving wheel. The motor and the third cylinder are arranged on the main body plate; before the silk ribbon is installed and the silk laying begins, the third cylinder is not started, the driving wheel and the driven wheel are separated, and when the silk laying begins, the silk ribbon is installed in the channel and passes between the driving wheel and the driven wheel. The third cylinder is started, and the driven wheel presses the silk ribbon onto the driving wheel to achieve the pressing of the silk ribbon; the motor drives the driving wheel to rotate during the silk laying process, driving the driven wheel to squeeze out the silk ribbon together.

[0011] The tension component includes: a material disc and a brake. The material disc is used to wind the ribbon. During the wire laying process, the material roll is forced to rotate under the drive of the heavy feeding component to realize the feeding of the ribbon. The brake is used to provide torque for the material disc to maintain a stable tension force for wire feeding to prevent the ribbon from being unstable fed or even falling off caused by the high-speed rotation of the material roll. The material disc and the brake are both arranged on the main plate.

[0012] The heat source component comprises: a laser heater, which is used to heat the ribbon and is arranged on the main body plate.

[0013] The connecting piece is arranged on the main body plate and connected to the end of the wire placement end movement mechanism.

[0014] The present invention relates to a closed-loop control method for online monitoring of a manufacturing surface based on the above system, wherein a path file containing system processing instructions is generated according to process parameters and paths, and ribbons are laid in pairs in a mirror-symmetrical manner for each line of instructions in the path file, and after the laying process, based on the collected thermal infrared image and visible light image, it is judged whether there is a misalignment according to the thermal infrared image, and defect detection is performed through a visual recognition model according to the visible light image to judge whether there is a misalignment and / or wrinkle, and then the path file and / or laser parameters are updated according to the misalignment judgment result, and then the downlink instruction is processed. Technical Effects

[0015] The present invention mainly monitors the misalignment defects generated during mirror image additive manufacturing through a thermal infrared imager, and mainly monitors the wrinkle defects generated during mirror image additive manufacturing through an industrial camera. The two are integrated to form a dual-modal monitoring system, which improves the accuracy of defect monitoring and the manufacturing precision of mirror image additive manufacturing. It is not easily affected by ambient light sources and ambient temperatures, and can efficiently and accurately monitor the image results collected during mirror image additive manufacturing in real time, and then analyze the defects generated during manufacturing. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of the structure of the present invention;

[0017] Figure 2 It is a schematic diagram of the structure of the visual detection model;

[0018] Figure 3 Schematic diagram of the visual detection model training process;

[0019] Figure 4 A schematic diagram of manual labeling during the visual detection model training process;

[0020] Figure 5 It is a schematic diagram of the strip installation process;

[0021] Figure 6 The schematic diagram of the process of online monitoring and closed-loop control of the surface quality of mirror additive manufacturing based on infrared and visible light dual-modal fusion;

[0022] Figure 7 It is a schematic diagram of the process of infrared monitoring and closed-loop control;

[0023] Figure 8 It is a schematic diagram of the arrangement of the measurement frame and the measurement results in the field of view of the infrared thermal imager during the infrared monitoring and closed-loop control process;

[0024] Fig. 9 for Figure 8 Physical photos of the corresponding monitoring locations;

[0025] Fig.10 It is a schematic diagram of the process of online monitoring and closed-loop control of visible light surface quality;

[0026] Fig.11 High-definition photos and corresponding labels of surfaces that are judged as normal wrinkles by the visual recognition model during online monitoring and closed-loop control of visible light surface quality;

[0027] Fig.12 High-definition photos and corresponding labels of surfaces that are judged to have too many wrinkles by the visual recognition model during online monitoring and closed-loop control of visible light surface quality;

[0028] Fig.13This is a picture of the effectiveness verification result of the visual inspection model in the process of online monitoring and closed-loop control of visible light surface quality;

[0029] Fig.14 The end effector center point distance-printing distance curve, misalignment disturbance position schematic diagram and misalignment-printing distance curve of experimental group 1;

[0030] Fig.15 The end effector center point distance-printing distance curve, misalignment disturbance position schematic diagram and misalignment-printing distance curve of control group 1;

[0031] Fig.16 The laser power-printing distance curve, the strip surface wrinkle detection schematic diagram, and the wrinkle-printing distance curve of experimental group 2;

[0032] Fig.17 The laser power-printing distance curve of the control group 2, the schematic diagram of the strip surface wrinkle detection, and the wrinkle-printing distance curve.

[0033] In the figure: 1 pressure sensor, 2 first cylinder, 3 connecting piece, 4 material roll, 5 support plate, 6 heater, 7 infrared thermal imager, 8 third cylinder, 9 pressure roller, 10 driving wheel, 11 second cylinder, 12 driven wheel, 13 motor, 14 blade, 15 brake, 16 heat insulation asbestos board, 17 main board, 20 industrial camera. DETAILED DESCRIPTION

[0034] like Figure 1 As shown, an online monitoring and closed-loop control system for the surface quality of mirror-image additive manufacturing based on dual-modal fusion of infrared and visible light includes: a first monitoring module and a second monitoring module that are arranged in mirror symmetry and have the same structure.

[0035] The monitoring module includes: a main plate 17 and a pressure component, a thermal infrared imager and an industrial camera, a shearing component, a re-feeding component, a tension component, a heat source component, a connecting part, a wire laying end motion mechanism, a pressure component, a shearing component, a re-feeding component and a tension component arranged thereon in sequence.

[0036] The pressure component includes: a pressure roller 9, a supporting device 16 and a first cylinder 2, wherein: the central axis of the pressure roller is parallel to the front-to-back direction, the supporting device is located on the right side of the pressure roller and connected to it, and the first cylinder 2 is arranged on the main plate 17 and is used to push the pressure roller to the left, cooperate with the pressure roller on the opposite side to squeeze and generate compaction pressure.

[0037] The shearing component includes: a second cylinder 8 and a blade, wherein: the second cylinder 8 is used to drive the blade to shear the ribbon, and the second cylinder 8 is arranged on the main body plate 17.

[0038] The heavy-feeding component includes: a driving wheel, a driven wheel 12, a motor 10 and a third cylinder 11, wherein: the driving wheel and the driven wheel 12 are arranged vertically and parallel to the left and right directions, the motor 10 is used to drive the driving wheel to rotate, the third cylinder 11 is used to press the driven wheel 12 onto the driving wheel, and the motor 10 and the third cylinder 11 are arranged on the main plate 17.

[0039] The tension component includes: a material disc 4 and a brake 15, wherein: the material disc 4 is used to wind the ribbon, and the brake 15 is used to provide torque for the material disc 4 to maintain a stable tension force for wire feeding, and the material disc 4 and the brake 15 are both arranged on the main plate 17.

[0040] The heat source component includes: a laser heater 6 , which is used to heat the ribbon, and the laser heater 6 is arranged on the main body plate 17 .

[0041] The thermal infrared imager 7 is installed at an angle of 67.5° to the pressure roller, and the industrial camera 20 is installed side by side with the end of the first cylinder 2 and is located above the pressure roller 9.

[0042] The connecting member 3 is arranged on the main plate 10 and connected to the end of the wire placement end movement mechanism.

[0043] This embodiment is based on the closed-loop control method for online monitoring of the manufacturing surface of the above-mentioned system. A path file containing system processing instructions is generated according to process parameters and paths. For each line of instructions in the path file, the ribbons are laid in pairs in a mirror-symmetrical manner. After the laying process, based on the collected thermal infrared image and visible light image, it is determined whether there is misalignment according to the thermal infrared image, and defect detection is performed through a visual recognition model based on the visible light image to determine whether there is misalignment and / or wrinkles. After updating the path file and / or laser parameters according to the misalignment judgment result, the downlink instruction is processed.

[0044] like Figure 2 As shown, the visual recognition model includes: a backbone neural network unit, a multi-scale feature enhancement module unit and a detection head unit, wherein: the backbone neural network unit extracts feature information of the image based on the image acquired by the industrial camera to obtain a multi-scale feature map of the model; the multi-scale feature enhancement module unit performs data refinement and enhancement processing based on the multi-scale feature map obtained by the backbone neural network unit to obtain an enhanced feature map; the detection head unit performs target classification, position regression and confidence estimation processing based on the enhanced feature map to obtain the final defect type and position information.

[0045] like Figure 3 As shown in the visual recognition model training method, in the experiment, an industrial camera was used to capture about 700 images of printed samples during the printing process, and the real defect labels of the samples were manually annotated (such as Figure 4 As shown), including: slight misalignment, severe misalignment, wrinkles and other defects. Wrinkles are mainly judged by the white reflective part of the surface image. Other defects refer to holes that have no obvious relationship with the process parameters in the additive manufacturing process. Each label contains the type of misalignment and the coordinates of the misaligned area, forming a data set for defect detection; the data set images are input into the built detection network for training, and the defect detection results obtained by the detection network are output, and the loss is calculated with the corresponding defect real label, and then the loss function is derived to update the model parameters. Then continue to input images and continuously update the model parameters to improve the detection accuracy of the model;

[0046] The mirror-symmetrical manner of laying the ribbons in pairs specifically includes:

[0047] 1) Ribbon installation; Figure 5 As shown, after the ribbon 21 is wound on the material roll and loaded into the wire laying channel, it leaves the material roll, passes between the driving wheel a and the driven wheel, and the blade reaches the pressure roller in sequence, and is then set on the top bracket, which is a vertical frame or a truss set on the ground. After that, the connecting piece is connected to the end of the wire laying end movement mechanism (mechanical arm); when the ribbon 21 passes between the driving wheel and the driven wheel, the third cylinder is started, and the driven wheel presses the ribbon 21 onto the driving wheel to achieve the compression of the ribbon 21; the motor is started, driving the driving wheel to rotate, driving the driven wheel to extrude the ribbon 21 together; when the ribbon 21 passes through the pressure roller, the first cylinder is started, so that the pressure rollers on both sides are pushed out, and cooperate with the pressure rollers on the opposite side to squeeze, so as to further achieve the compression of the ribbon 21;

[0048] 2) The first wire laying: After the wire 21 is set on the bracket, the wire laying end motion mechanism starts to move along the wire laying path. At the same time, the laser heater 4 is started to heat the wire 21. When the first wire laying is about to be completed, the second cylinder is started to drive the blade to shear the wire 21 and cut off the excess part. After shearing, the second cylinder is reset, and the reserved wire 21 between the blade and the pressure roller continues to be laid until the first laying is completed;

[0049] 3) Second wire laying: After the first wire laying is completed, the wire laying end motion mechanism moves to the initial position, and then starts to move along the wire laying path (the two wires 21 laid for the second time are located on the two sides of the two wires 21 laid in the first wire laying process that are away from each other). At the same time, the laser heater 4 is started to heat the wire 21. When the second wire laying is about to be completed, the second cylinder is started to drive the blade to shear the wire 21 and cut off the excess part. After shearing, the second cylinder is reset, and the reserved wire 21 between the blade and the pressure roller continues to be laid. After the second laying is completed, the third cylinder, the motor, the first cylinder, the third cylinder, and the laser heater 6 are turned off, and the connection between the connector and the end of the wire laying end motion mechanism is disconnected;

[0050] The process of using this dual-modality mirror-based additive manufacturing surface quality real-time monitoring and adaptive optimization method to monitor and optimize surface defects in the additive manufacturing process is as follows: Figure 6 As shown, the infrared recognition method and the visual recognition method are performed simultaneously;

[0051] During the printing process, the robot reads the path file line by line and adjusts to the corresponding process parameters according to the instructions, moves from the current position to the target coordinates, and completes the printing of the current step. The infrared monitoring method and the visual monitoring method are performed simultaneously to modify the path coordinate file and the laser parameters. After completing this step, the robot continues to read the next line of path coordinate instructions in the path coordinate file. If the line is a termination instruction, it is terminated and the printing of the line is completed. If the line is not a termination instruction, the robot enters the next step, continues to read the target coordinates and adjusts the corresponding process parameters to complete the next step of printing;

[0052] Infrared monitoring is performed in the current step, the process is as follows Figure 7 As shown in FIG. 1 , two measurement frames are arranged in the infrared thermal image according to the position of the strip, and the average temperatures t1 and t2 of the left and right edges of the strip in the field of view are measured respectively. Figure 8 As shown;

[0053] The infrared thermal imager installed in the second monitoring module measures t1 and t2. If t1-t2>2℃, Figure 8 As shown in a, the monitoring result is: the strip installed by the first monitoring module is positively offset, and the y coordinate modification value step is 0.1. In order to verify whether the judgment result is accurate, the following is intercepted Fig. 9 a shows the actual image of the monitoring position, where it can be seen that the strip is offset to the left;

[0054] If t2-t1>2℃, Figure 8 b, the monitoring result is: the strip installed by the first monitoring module is negatively offset, and the y coordinate modification value step is -0.1. In order to verify whether the judgment result is accurate, the following is intercepted Fig. 9 b shows the actual image of the monitoring position, where it can be seen that the strip is offset to the right;

[0055] If |t2-t1|≤2℃, Figure 8 c, the monitoring result is: the strip has no misalignment, the y coordinate modification value step is 0, to verify whether the judgment result is accurate, the following is intercepted Fig. 9 c shows the actual picture of the monitoring position, and it can be seen that the strip has no deviation;

[0056] Compare the modified value step with the suggested modified value Δy obtained by visual monitoring. If the signs are the same, modify the y coordinate value in the path coordinate file by step. If the signs are opposite, it proves that the conclusions obtained by the two monitoring and adaptive optimization methods conflict. In order to prevent the wrong correction caused by wrong judgment and affect the printing quality, the path coordinate correction is not performed in this step.

[0057] like Fig.10 As shown, judging whether there is a misalignment according to the thermal infrared image specifically includes:

[0058] a) Use a high-definition industrial camera to record video at the beginning of the wire laying process and capture surface images in real time;

[0059] b) Input the image into the trained visual recognition model, and the algorithm marks the types of corresponding surface defects in the image, including: misalignment, wrinkles and damage, and marks the regional coordinates and area of ​​each defect in the image;

[0060] c) For the detected surface dislocation defect, the model determines the direction of the dislocation according to the coordinate position and area of ​​the dislocation defect, determines the positive or negative value of the y-coordinate modification value Δy, and gives the specific value of Δy according to the severity of slight wrinkles and severe wrinkles indicated by the defect label;

[0061] d) The y coordinate modification value Δy given by the model is used to verify the misalignment defect monitoring result obtained by the infrared method. If the judgment results are the same, the adaptive optimization of the infrared method is not interfered. If the judgment results are opposite, the step is reset to 0, and the path coordinate y is not modified at this moment;

[0062] e) The model determines the severity of the misalignment based on the number and area of ​​the detected surface wrinkle defects, such as Fig.11 As shown, the surface wrinkle condition is judged as normal and the laser power is not modified; Fig.12 As shown, the surface wrinkle condition is judged to be too much, and a reduced value of the laser power is given to reduce the degradation of the surface quality caused by excessive laser power;

[0063] f) Correct the laser power in real time according to the solution given by the model;

[0064] After specific experiments, the model was verified using test set images. The results are as follows Fig.13 As shown in the figure, it shows that the visual inspection model can identify various surface defects in the test set images, including slight misalignment, severe misalignment and wrinkles, and the recognition results are consistent with the actual results, proving the effectiveness of the model;

[0065] In order to verify the effectiveness of the visual recognition model in the present invention, the model is used to identify the test set images, and the recognition effect of each defect type in the result is evaluated. The precision P, recall rate R, mean average precision mAP50 when the intersection-over-union ratio threshold is 0.5, and mean average precision mAP50-90 when the intersection-over-union ratio threshold is from 0.5 to 0.95 are calculated respectively, as shown in Table 1. Among them, mAP50 is an important indicator for comprehensively evaluating the performance of the model in the target detection task. The recognition results of each defect type of the model have a mAP50 of about 0.9, which proves that its comprehensive performance is good.

[0066] Table 1 Defect Type P R mAP50 mAP50-95 all 0.815 0.913 0.911 0.653 Slight misalignment 0.706 0.857 0.841 0.629 Serious dislocation 0.929 0.969 0.971 0.712 Wrinkles 0.81 0.913 0.921 0.62

[0067] In order to prove that the online monitoring and closed-loop control method for the surface quality of mirror-image additive manufacturing of the present invention can effectively correct the surface defects of the printed strip in real time, the following experiments are conducted to verify:

[0068] Experimental group 1: The online monitoring and closed-loop control method for surface quality of mirror additive manufacturing in the present invention was used. The initial output power of the laser heater was set to 90 W, the initial speed of the robot arm was set to 10 mm / s, and the printing distance was set to 200 mm. When the printing distance reached 50 mm, the distance between the center points of the two end effectors (TCP distance) was set to 1 mm as a disturbance.

[0069] Control group 1: basically the same as experimental group 1, except that the mirror additive manufacturing surface quality online monitoring and closed-loop control method was not used;

[0070] Experimental group 2: The online monitoring and closed-loop control method for surface quality of mirror additive manufacturing in the present invention was used, and the initial output power of the laser heater was set to 90 W, the initial speed of the robot arm was set to 10 mm / s, and the printing distance was set to 200 mm. When the printing distance reached 50 mm, the laser power was set to 150 W as a disturbance;

[0071] Control group 2: basically the same as experimental group 2, except that the mirror additive manufacturing surface quality online monitoring and closed-loop control method was not used;

[0072] The end effector center point distance-printing distance curve, dislocation disturbance position diagram and dislocation-printing distance curve of experimental group 1 and control group 1 in the mirror additive manufacturing process are shown in Figure 2. Fig.14 and Fig.15 As shown in the figure, it can be seen from the comparison that after the distance between the center points of the two end effectors was modified, the strip produced obvious misalignment. Compared with the gradual increase in the misalignment in the control group, the use of the mirror additive manufacturing surface quality online monitoring and closed-loop control method to quickly restore the misalignment value to 0 can effectively control the misalignment defects on the strip surface.

[0073] Laser power-printing distance curves of experimental group 2 and control group 2 during mirror additive manufacturing, schematic diagram of strip surface wrinkle detection, and wrinkle-printing distance curves Fig.16 and Fig.17 As shown, by comparison, it can be seen that after the laser power increases, the surface wrinkles of the strips in both groups of experiments are more serious. Compared with the obvious surface wrinkles in the control group, the surface wrinkles can be controlled within a smaller range by using the mirror additive manufacturing surface quality online monitoring and closed-loop control method, which can effectively control the wrinkle defects on the strip surface.

[0074] The above-mentioned specific implementation can be partially adjusted in different ways by those skilled in the art without departing from the principle and purpose of the present invention. The protection scope of the present invention shall be based on the claims and shall not be limited by the above-mentioned specific implementation. Each implementation scheme within its scope shall be subject to the constraints of the present invention.

Claims

1. A dual-mode fusion mirror additive manufacturing surface online monitoring system, characterized in that: include: A first monitoring module and a second monitoring module are arranged in mirror symmetry and have the same structure, and the monitoring module includes: a main plate and a pressure component, a thermal infrared imager and an industrial camera, a shearing component, a re-feeding component, a tension component, a heat source component, a connecting part, a wire laying end motion mechanism, a pressure component, a shearing component, a re-feeding component and a tension component arranged thereon in sequence.

2. The dual-modal fusion mirror additive manufacturing surface online monitoring system according to claim 1 is characterized in that: The pressure component includes: a pressure roller, a supporting device and a first cylinder, wherein: the central axis of the pressure roller is parallel to the front-to-back direction, the supporting device is located on the right side of the pressure roller and connected to it, and the first cylinder is arranged on the main plate and is used to push the pressure roller to the left, cooperate with the pressure roller on the opposite side to squeeze and generate compaction pressure.

3. The dual-modal fusion mirror additive manufacturing surface online monitoring system according to claim 1 is characterized in that: The thermal infrared imager is installed at an angle of 67.5° to the pressure roller, and the industrial camera is installed side by side with the end of the first cylinder and is located above the pressure roller.

4. The dual-modal fusion mirror additive manufacturing surface online monitoring system according to claim 1 is characterized in that: The shearing component comprises: a second cylinder and a blade, wherein: the second cylinder is used to drive the blade to shear the ribbon, and the second cylinder is arranged on the main body plate.

5. The dual-modal fusion mirror additive manufacturing surface online monitoring system according to claim 1 is characterized in that: The heavy-feeding component includes: a driving wheel, a driven wheel, a motor and a third cylinder. The driving wheel and the driven wheel are arranged vertically and parallel to the left and right directions. The motor is used to drive the driving wheel to rotate, and the third cylinder is used to press the driven wheel onto the driving wheel. The motor and the third cylinder are arranged on the main body plate; before the silk ribbon is installed and the silk laying begins, the third cylinder is not started, the driving wheel and the driven wheel are separated, and when the silk laying begins, the silk ribbon is installed in the channel and passes between the driving wheel and the driven wheel. The third cylinder is started, and the driven wheel presses the silk ribbon onto the driving wheel to achieve the pressing of the silk ribbon; the motor drives the driving wheel to rotate during the silk laying process, driving the driven wheel to squeeze out the silk ribbon together.

6. The dual-modal fusion mirror additive manufacturing surface online monitoring system according to claim 1 is characterized in that: The tension component includes: a material disc and a brake. The material disc is used to wind the ribbon. During the wire laying process, the material roll is forced to rotate under the drive of the heavy feeding component to realize the feeding of the ribbon. The brake is used to provide torque for the material disc to maintain a stable tension force for wire feeding to prevent the ribbon from being unstable fed or even falling off caused by the high-speed rotation of the material roll. The material disc and the brake are both arranged on the main plate.

7. The dual-modal fusion mirror additive manufacturing surface online monitoring system according to claim 1 is characterized in that: The heat source component comprises: a laser heater, which is used to heat the ribbon and is arranged on the main body plate.

8. The dual-modal fusion mirror additive manufacturing surface online monitoring system according to claim 1 is characterized in that: The connecting piece is arranged on the main body plate and connected to the end of the wire placement end movement mechanism.

9. A closed-loop control method for online monitoring of a manufacturing surface based on the system described in any one of claims 1 to 8, characterized in that: A path file containing system processing instructions is generated according to the process parameters and path. For each line of instructions in the path file, the ribbons are laid in pairs in a mirror-symmetrical manner. After the laying process, based on the collected thermal infrared image and visible light image, it is determined whether there is misalignment based on the thermal infrared image, and defect detection is performed through the visual recognition model based on the visible light image to determine whether there is misalignment and / or wrinkles. After updating the path file and / or laser parameters based on the misalignment judgment result, the downlink instructions are processed.

Images