Multi-stage regulation ultrasonic vibration assisted electric arc additive device and method based on deep learning

Abstract

A kind of multi-stage regulation ultrasonic vibration auxiliary electric arc additive device and method based on deep learning, including additive robot and workbench, and the workbench on the clamping frame of vibration pickup connected with additive robot, ultrasonic vibration device is connected below workbench, flange is connected outside ultrasonic vibration device, upper cover is connected on the upper part of flange, upper cover side and slide rail are connected, slide rail is connected on clamping frame, upper cover and lower hydraulic device are connected, hydraulic device is connected on clamping frame, lower cavity is connected in the lower part of flange, clamping device is connected below lower cavity;Clamping frame is installed on the lower guide rail, and clamping frame is controlled to move on the lower guide rail by stepper motor;Ultrasonic vibration device is controlled by control power supply;Method is based on the multi-stage self-adaptive regulation based on deep learning of ultrasonic vibration power of ultrasonic vibration device based on the distance measuring effect of vibration pickup, to realize the consistency of actual power effect to the same layer;The present application realizes the uniformity of performance enhancement of additive component.

Description

Technical Field

[0001] This invention relates to the field of electric arc additive manufacturing technology, specifically to a deep learning-based multi-level adjustable ultrasonic vibration-assisted electric arc additive manufacturing device and method. Background Technology

[0002] Arc-wire additive manufacturing is a technology that uses an electric arc as a heat source to rapidly melt and deposit metal wire along a predetermined path to achieve near-net-shape forming of metal parts. Inspired by ultrasonic-assisted welding in traditional welding, ultrasonic vibration-assisted arc additive manufacturing has developed rapidly in recent years. The cavitation and acoustic flow effects generated by ultrasound in the molten pool can promote grain refinement and molten pool flow, thereby improving the mechanical properties and forming quality of additively manufactured components.

[0003] In application, the ultrasonic generator is difficult to move with the shape during printing and the ultrasonic vibration power is difficult to adjust with the number of printing layers, resulting in different ultrasonic vibration intensities at different locations of the printed component, which in turn leads to differences in component performance. Patent (application number CN202211243409.7, entitled "A Method for Ultrasonic-Assisted Laser Welding Additive Manufacturing of Titanium Alloy") discloses an ultrasonic-assisted additive manufacturing method for titanium alloys, in which an ultrasonic vibration device is set under the worktable to promote grain refinement. However, this method uses fixed-point ultrasonic vibration, and the ultrasonic vibration intensity inside the part varies with the additive trajectory and additive height, resulting in non-uniformity of the final formed part's microstructure and properties. The patent (application number CN202310026944.5, entitled "A Plasma-Coordinated Ultrasonic-Assisted Additive Manufacturing Device and Method") discloses an additive manufacturing device that combines plasma treatment and ultrasonic treatment. The ultrasonic device is located behind the printing path and moves with the shape to apply ultrasonic treatment to the printed layer. However, there is a certain distance between the ultrasonic device and the welding gun, making it difficult to apply ultrasonic treatment. Moreover, it can only process simple planar parts, limiting its application scope. Summary of the Invention

[0004] In order to overcome the shortcomings of the prior art, the present invention aims to provide a multi-level adjustable ultrasonic vibration assisted arc additive manufacturing device and method based on deep learning, which can adjust the position and intensity of ultrasonic vibration according to the shape during the printing process, and achieve uniform performance enhancement of additive components by adjusting the ultrasonic vibration power as the additive deposition height increases.

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

[0006] A deep learning-based multi-level adjustable ultrasonic vibration-assisted arc additive manufacturing device includes an additive manufacturing robot 1 and a worktable 3. A vibration meter 2 connected to the additive manufacturing robot 1 is connected to the worktable 3. The worktable 3 is mounted on a clamping frame 6. An ultrasonic vibration device 4 is connected below the worktable 3. A flange 5 is connected to the outside of the ultrasonic vibration device 4. An upper pressure cover 7 is connected to the upper part of the flange 5. The side of the upper pressure cover 7 is connected to a slide rail 8. The slide rail 8 is connected to the clamping frame 6. The upper pressure cover 7 is connected to a hydraulic device 10 below it. The hydraulic device 10 is connected to the clamping frame 6. A lower pressure chamber 9 is connected to the lower part of the flange 5. A clamping device 11 is connected below the lower pressure chamber 9. The clamping frame 6 is mounted on a lower guide rail 12 and is controlled by a stepper motor 14 to move left and right on the lower guide rail 12. The ultrasonic vibration device 4 is controlled by a control power supply 13.

[0007] The additive robot 1 includes a six-axis robot and a CMT welding machine.

[0008] The ultrasonic vibration device 4 has a vibration frequency of 15kHz, and the ultrasonic vibration power is adjusted by the clamping force.

[0009] The control power supply 13 can display the instantaneous power of the ultrasonic vibration device 4 during the ultrasonic vibration process.

[0010] A method for using a deep learning-based, multi-level adjustable ultrasonic vibration-assisted arc additive manufacturing device includes the following steps:

[0011] Step 1: Initialize the device and set the process parameters for additive robot 1;

[0012] Step 2: Move the clamping frame 6 to the position where the additive manufacturing path begins;

[0013] Step 3: Turn on the control power supply 13, start the clamping device 11, turn on the hydraulic device 10, and adjust the ultrasonic vibration power of the ultrasonic vibration device 4 to the required level.

[0014] Step 4: Set the stepper motor 14 to move at the same speed as the additive robot 1 welding speed;

[0015] Step 5: Simultaneously turn on the stepper motor 14 and the additive robot 1 to begin ultrasonic-assisted additive manufacturing;

[0016] Step Six: Adaptively adjust the ultrasonic vibration power of the ultrasonic vibration device 4 based on the following distance measurement effect of the vibration meter 2 to achieve consistency in the actual power effect for the same layer; train a deep learning-based neural network with deposition height (i.e., deposition layer) as the label and expected average grain size as the input, and ultrasonic vibration power as the output; optimize the neural network and deposition process parameters through deposition of a single wall, and optimize the process parameters through orthogonal experiments with the stability of deposition morphology as the standard; collect data based on the optimized process parameters, and determine the maximum number of layers that can achieve grain refinement by using the maximum ultrasonic vibration power, and build the dataset by changing the applied power of the bottom layer and increasing the amplitude layer by layer, and perform more than one destructive experiment to sample EBSD grain size statistics for each pass; sequentially screen the dataset and train and validate the deep learning model;

[0017] Step 7: Repeat steps 1 to 5. When the deposition height increases, adjust the ultrasonic vibration power of ultrasonic vibration device 4 according to the deep learning prediction results in step 6 until all layers are deposited.

[0018] In step six, data is collected based on the optimized process parameters. The dataset collection method involves determining the maximum number of layers capable of grain refinement using the maximum ultrasonic vibration power, establishing the dataset by changing the applied ultrasonic vibration power at the bottom layer and gradually increasing the amplitude layer by layer, and performing at least one destructive experiment to sample and statistically analyze the EBSD grain size in each pass. The dataset is then sequentially selected, and the deep learning model is trained and validated. The specific steps are as follows:

[0019] 6.1) Deposit layer by layer using the maximum ultrasonic vibration power as the standard to determine the maximum number of layers n that can achieve grain refinement;

[0020] 6.2) Conduct an initial ultrasonic vibration power test and obtain EBS D data of the sample through a destructive experiment to obtain the average grain size;

[0021] 6.3) Using the grain refinement effect of the first layer as a standard, an attempt was made to increase the ultrasonic vibration power of the second layer;

[0022] 6.4) Increase the number of layers one by one, and determine the power W that can achieve the same grain refinement effect as the previous layer in the i-th layer. i When W i Deposition was stopped when the maximum ultrasonic vibration power was approached; a co-i-layer deposition single wall was obtained to verify the overall grain refinement effect;

[0023] 6.5) Divide the dataset obtained in step 6.4) into training set, validation set and test set; use the average grain size and the height of the deposited layer from the substrate as input and the ultrasonic vibration power as output to train the neural network model. The evaluation index used is the uniformity of grain refinement along the deposition height direction.

[0024] Compared with the prior art, the beneficial effects of this invention are as follows:

[0025] (A) This invention uses a neural network model to predict the ultrasonic vibration power layer by layer, which can effectively ensure the stability of the ultrasonic vibration intensity at the molten pool and can stably exert influence on the molten pool. Through the acoustic flow effect and acoustic cavitation effect, it promotes the flow of the molten pool and grain refinement, improves the mechanical properties and forming quality of the additive manufacturing components, and at the same time ensures the uniformity of the microstructure and mechanical properties of the formed parts.

[0026] (B) The present invention can stably adjust the position and intensity of ultrasonic vibration, and the application method is simple and convenient, with considerable economic benefits. It can be applied to near-net-shape manufacturing of components of most shapes and sizes.

[0027] (C) This invention can be applied to different heat source forms, including plasma arc welding additive manufacturing, electron beam welding additive manufacturing, laser welding additive manufacturing, etc., and can be applied to different metal materials, including magnesium alloys, nickel-titanium shape memory alloys, Invar alloys, etc. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the device structure according to an embodiment of the present invention.

[0029] Figure 2 This is a flowchart illustrating the data set selection and the training and validation process of the deep learning model in an embodiment of the present invention.

[0030] Figure 3 This is a schematic diagram of the dataset collection method in an embodiment of the present invention. Detailed Implementation

[0031] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0032] Reference Figure 1A deep learning-based multi-level adjustable ultrasonic vibration-assisted arc additive manufacturing device includes an additive manufacturing robot 1 and a worktable 3. A vibration meter 2 connected to the additive manufacturing robot 1 is connected to the worktable 3. The worktable 3 is mounted on a clamping frame 6. An ultrasonic vibration device 4 is connected below the worktable 3. A flange 5 is connected to the outside of the ultrasonic vibration device 4. An upper cover 7 is bolted to the upper part of the flange 5. The side of the upper cover 7 is connected to a slide rail 8, which can slide up and down on the slide rail 8. The slide rail 8 is connected to the clamping frame 6. The upper cover 7 is connected to a hydraulic device 10 below it. The hydraulic device 10 is connected to the clamping frame 6 and acts as a buffer for the sliding of the upper cover 7. A lower pressure chamber 9 is bolted to the lower part of the flange 5. A clamping device 11 is connected below the lower pressure chamber 9. The clamping frame 6 is mounted on a lower guide rail 12 and is controlled by a stepper motor 14 to move left and right on the lower guide rail 12. The ultrasonic vibration device 4 is controlled by a control power supply 13.

[0033] The additive robot 1 includes a six-axis robot and a CMT welding machine.

[0034] The ultrasonic vibration device 4 has a vibration frequency of 15kHz, and its power is adjusted by the clamping force.

[0035] The control power supply 13 can display the instantaneous power of the ultrasonic vibration device 4 during the ultrasonic vibration process.

[0036] A method for using a deep learning-based, multi-level adjustable ultrasonic vibration-assisted arc additive manufacturing device includes the following steps:

[0037] Step 1: Initialize the device and set the process parameters for additive robot 1;

[0038] Step 2: Move the clamping frame 6 to the position where the additive manufacturing path begins;

[0039] Step 3: Turn on the control power supply 13, start the clamping device 11, turn on the hydraulic device 10, and adjust the ultrasonic vibration power of the ultrasonic vibration device 4 to the required level.

[0040] Step 4: Set the stepper motor 14 to move at the same speed as the additive robot 1 welding speed;

[0041] Step 5: Simultaneously turn on the stepper motor 14 and the additive robot 1 to begin ultrasonic-assisted additive manufacturing;

[0042] Step Six: Based on the ranging effect of the vibration meter 2, the ultrasonic vibration power of the ultrasonic vibration device 4 is adaptively adjusted to achieve consistency in the actual power effect for the same layer; a deep learning-based neural network is trained with the deposition height (i.e., the deposition layer) as the label and the expected average grain size as the input, and the ultrasonic vibration power as the output. The neural network training and the optimization of deposition process parameters are achieved through deposition of a single wall; the process parameters are optimized through orthogonal experiments with the standard of stable deposition morphology (avoiding obvious defects such as humps, melt collapse, etc.).

[0043] Reference Figure 2 , Figure 3 Data is collected based on the optimized process parameters, through... Figure 2 The dataset was selected and the deep learning model was trained and validated in the following order: A dataset was established by changing the applied ultrasonic vibration power at the bottom layer and gradually increasing the amplitude layer by layer. Five destructive experiments were performed in each pass to sample and statistically analyze the EBSD grain size. The specific steps included are as follows:

[0044] 6.1) Using the maximum ultrasonic vibration power of 1200W as the standard, layer-by-layer deposition was performed to determine the maximum number of layers n that could achieve grain refinement;

[0045] 6.2) Conduct initial ultrasonic vibration power tests and obtain EBS D data of the samples through destructive experiments to obtain the average grain size (e.g., Figure 3 As shown in the first layer, the power in this embodiment is attempted to start without ultrasound, with an increase of 200W and an upper limit of 1200W.

[0046] 6.3) Using the grain refinement effect of the first layer as a standard, the ultrasonic vibration power amplification of the second layer was tested. The power was tested at 50W intervals in the second layer. When the power was W2 (>200W), the grain refinement effect was consistent with that caused by 200W in the first layer (corresponding to...). Figure 3 Second layer);

[0047] 6.4) Increase the number of layers one by one, and determine the power W that can achieve the same grain refinement effect as the previous layer in the i-th layer. i When W i Deposition was stopped when the temperature approached 1200W; a single wall of co-i layers was obtained, and the overall grain refinement effect was verified (compared to...). Figure 3 (Corresponding to the intermediate layer);

[0048] 6.5) Divide the dataset obtained by the above process into training set, validation set and test set in a ratio of 6:2:2; train the neural network model with average grain size and deposition layer height from substrate as input and ultrasonic vibration power as output, and use the uniformity of grain refinement along the deposition height direction as the evaluation index.

[0049] Step 7: Repeat steps 1 to 5. When the deposition height increases, adjust the ultrasonic vibration power of ultrasonic vibration device 4 according to the deep learning prediction results in step 6 until all layers are deposited.

[0050] In this embodiment, TC4 titanium alloy wire with a diameter of 1.2 mm is used. The welding speed of the additive manufacturing equipment 1 is set to 0.3 m / min, the wire feeding speed is 7 m / min, the welding heat input is 315.9 J / mm, and the initial ultrasonic vibration power is 100 W. For every 10 mm increase in deposition height, the ultrasonic vibration power is increased by 20 W.

[0051] Deposited tensile specimens in the horizontal direction were processed by wire electrical discharge machining with a gauge length of 10 mm. The surface oxide layer was removed by grinding, and room temperature tensile tests were conducted. The results of the tensile tests are shown in Table 1. It can be seen that the strength of the specimens is greatly improved while maintaining plasticity, which demonstrates the effectiveness of the method of the present invention.

[0052] Table 1

[0053] Yield strength tensile strength elongation control group 824.67 994 12.96 Ultrasonic vibration specimen 978 1050 13.12

[0054] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention shall fall within the scope of the technical solution of the present invention.

Claims

1. A deep learning-based multi-level adjustable ultrasonic vibration-assisted arc additive manufacturing device, comprising an additive manufacturing robot (1) and a worktable (3), characterized in that: The vibration meter (2) connected to the additive manufacturing robot (1) is connected to the worktable (3), which is set on the clamping frame (6). An ultrasonic vibration device (4) is connected below the worktable (3). A flange (5) is connected to the outside of the ultrasonic vibration device (4). An upper cover (7) is connected to the upper part of the flange (5). The side of the upper cover (7) is connected to the slide rail (8), which is connected to the clamping frame (6). The upper cover (7) is connected to the hydraulic device (10) below it. The hydraulic device (10) is connected to the clamping frame (6), the lower part of the flange (5) is connected to the pressure chamber (9), and the clamping device (11) is connected below the pressure chamber (9); the clamping frame (6) is installed on the lower guide rail (12), and the clamping frame (6) is controlled by the stepper motor (14) to move left and right on the lower guide rail (12); the ultrasonic vibration device (4) is controlled by the control power supply (13); the ultrasonic power is controlled based on the deposition height and the average grain size according to deep learning.

2. The apparatus according to claim 1, characterized in that: The additive robot (1) includes a six-axis robot and a CMT welding machine.

3. The apparatus according to claim 1, characterized in that: The ultrasonic vibration device (4) has a vibration frequency of 15KHZ and its power is adjusted by the clamping force.

4. The apparatus according to claim 1, characterized in that: The control power supply (13) can display the instantaneous power of the ultrasonic vibration device (4) during ultrasonic vibration.

5. A method for using a deep learning-based multi-level adjustable ultrasonic vibration-assisted arc additive manufacturing device according to any one of claims 1-4, characterized in that, Includes the following steps: Step 1: Initialize the device and set the process parameters of the additive robot (1); Step 2: Move the clamp (6) to the position where the additive manufacturing path begins; Step 3: Turn on the control power (13), start the clamping device (11), turn on the hydraulic device (10), and adjust the ultrasonic vibration power of the ultrasonic vibration device (4) to the required level. Step 4: Set the stepper motor (14) to move at the same speed as the additive robot (1) welding speed; Step 5: Simultaneously turn on the stepper motor (14) and the additive robot (1) to begin ultrasonic-assisted additive manufacturing; Step 6: Based on the following distance measurement effect of the vibration meter (2), the ultrasonic vibration power of the ultrasonic vibration device (4) is adaptively adjusted to stabilize the actual power of the same layer; a neural network based on deep learning is trained with the deposition height (i.e., the deposition layer) as the label and the expected average grain size as the input, and the ultrasonic vibration power as the output. The neural network training and the optimization of deposition process parameters are achieved through deposition of a single wall. The process parameters are optimized through orthogonal experiments with the stability of the deposition morphology as the standard; data is collected based on the optimized process parameters. The dataset collection method is to determine the maximum number of layers that can achieve grain refinement by using the maximum ultrasonic vibration power. The dataset is established by changing the magnitude of the ultrasonic vibration power applied to the bottom layer and increasing the amplitude layer by layer. Each pass is subjected to more than one destructive experiment to sample EBSD grain size statistics. The dataset is then screened and the deep learning model is trained and verified in sequence. Step 7: Repeat steps 1 to 5. When the deposition height increases, adjust the ultrasonic vibration power of the ultrasonic vibration device (4) according to the deep learning prediction results in step 6 until all layers are deposited.

6. The method according to claim 5, characterized in that, In step six, data is collected based on the optimized process parameters. The dataset collection method involves determining the maximum number of layers capable of grain refinement using the maximum ultrasonic vibration power, establishing the dataset by changing the applied ultrasonic vibration power at the bottom layer and gradually increasing the amplitude layer by layer, and performing at least one destructive experiment to sample and statistically analyze the EBSD grain size in each pass. The dataset is then sequentially selected, and the deep learning model is trained and validated. The specific steps are as follows: 6.1) Deposit layer by layer using the maximum ultrasonic vibration power as the standard to determine the maximum number of layers n that can achieve grain refinement; 6.2) Conduct an initial ultrasonic vibration power test and obtain EBSD data of the sample through a destructive experiment to obtain the average grain size; 6.3) Using the grain refinement effect of the first layer as a standard, an attempt was made to increase the ultrasonic vibration power of the second layer; 6.4) Increase the number of layers one by one, and determine the power W that can achieve the same grain refinement effect as the previous layer in the i-th layer. i When W i Deposition was stopped when the maximum ultrasonic vibration power was approached; a co-i-layer deposition single wall was obtained to verify the overall grain refinement effect; 6.5) Divide the dataset obtained in step 6.4) into training set, validation set and test set; use the average grain size and the height of the deposited layer from the substrate as input and the ultrasonic vibration power as output to train the neural network model. The evaluation index used is the uniformity of grain refinement along the deposition height direction.

Images