Ultrasonic vibration wire assisted melting electrode arc additive device

Abstract

The application discloses a kind of ultrasonic vibration wire auxiliary melting electrode arc additive devices, including wire feeding mechanism, welding torch mechanism and ultrasonic vibration mechanism, wire feeding mechanism is used to provide power for welding wire, and welding wire is sent to the direction of welding torch mechanism;Welding torch mechanism is used to add current to welding wire and drive welding wire to work area;Ultrasonic vibration mechanism is arranged between wire feeding mechanism and welding torch mechanism, and ultrasonic vibration is exerted on welding wire by transducer and amplitude bar, so that, during welding, ultrasonic vibration is conducted to molten pool along with welding wire, and molten pool is stirred.In the embodiment, ultrasonic vibration is conducted to molten pool through wire, and the mechanical properties of additive components, such as microhardness and tensile strength, are finally improved by improving droplet transfer, promoting molten pool flow, uniform molten pool composition and breaking coarse grains through cavitation effect, acoustic streaming effect, mechanical effect and thermal effect.

Description

Technical Field

[0001] This invention relates to additive manufacturing, and more particularly to an ultrasonic arc additive manufacturing apparatus. Background Technology

[0002] Wire arc additive manufacturing (WAAM) technology has become a key development direction in the metal manufacturing field due to its advantages such as no need for molds, simple process flow, and high forming efficiency. WAAM technology is based on the concept of welding overlay, using an electric arc as the heat source and wire as the filler material. It utilizes 3D software to slice components and shape them according to a predetermined path. Its advantages include high forming efficiency, low cost, applicability to a wide range of materials, and the ability to add large and complex components. In addition, WAAM technology can also be used for the repair of metal components.

[0003] Arc additive manufacturing has opened up new avenues for component production, but it also faces many unresolved challenges, such as poor molten pool fluidity, inhomogeneous additive component composition, coarse grains, high residual stress, and susceptibility to porosity. These problems not only affect the forming quality of the components but also significantly weaken their mechanical properties and reduce their service life. Among these, the most significant issue in arc additive manufacturing is the tendency for coarse grains. This is because the electric arc, as a heat source, generates a large heat input, and the ambient temperature of arc additive manufacturing causes the molten pool to undergo rapid heating and cooling, resulting in a large temperature gradient within the molten pool and causing the additively formed components to develop coarse columnar grains. According to Hall-Petch's law, the larger the grain size, the lower the yield strength, and the worse the mechanical properties of the component. Therefore, to obtain additive components with good mechanical properties, the grains must be small and uniformly distributed. This makes grain refinement a crucial issue in arc additive manufacturing. In recent years, many experts and scholars at home and abroad have combined other energy fields with arc additive manufacturing to achieve the purpose of refining grains and reducing porosity, such as magnetic fields, ultrasonic energy fields, and laser composites. This invention is a composite of ultrasonic energy fields and arc additive manufacturing, so the following mainly introduces the ultrasonic energy field composite. The main mechanism of ultrasonic energy field acting on arc additive manufacturing is that ultrasonic waves generate cavitation effects, acoustic flow effects, and thermal effects in the molten pool. These effects can reduce grain size, reduce porosity formation, and reduce residual stress.

[0004] Ultrasonic Wire-Assisted Molten Electrode Additive Manufacturing (EAD) is a technology that utilizes the combined effects of ultrasonic and electric arc energy fields to achieve rapid prototyping of metal components. Currently, ultrasonic-assisted EAD mainly falls into two categories: one applies ultrasonic vibration to a substrate or deposited layer, and the other applies ultrasonic vibration to a wire or molten pool. The former has the disadvantage that the range and intensity of ultrasonic vibration are affected by factors such as the shape, size, thickness, and surface roughness of the substrate or deposited layer, and the direction of ultrasonic vibration may not be consistent with the direction of molten pool movement, potentially leading to uneven action on the molten pool. The latter has the disadvantage that the range and intensity of ultrasonic vibration are affected by factors such as the diameter, material, and feed speed of the wire, and ultrasonic vibration may affect wire transport and droplet transition. Applying pressure to the deposited layer using an ultrasonic device can lead to excessive friction and ultrasonic energy loss, as the surface of the deposited layer may not be smooth enough.

[0005] In summary, there are still some technical difficulties that need to be addressed with the existing technology. Summary of the Invention

[0006] The purpose of this invention is to provide an ultrasonic vibration wire-assisted fused electrode arc additive manufacturing device to solve the aforementioned problems existing in the prior art.

[0007] Technical solution: A device for ultrasonically vibrating wire-assisted fused electrode arc additive manufacturing, comprising:

[0008] The wire feeding mechanism is used to provide power to the welding wire and transport it towards the welding gun mechanism;

[0009] The welding torch mechanism is used to apply current to the welding wire and move the welding wire to the working area;

[0010] An ultrasonic vibration mechanism is installed between the wire feeding mechanism and the welding torch mechanism. It applies ultrasonic vibration to the welding wire through a transducer and an amplitude transformer, so that during welding, the ultrasonic vibration is transmitted to the molten pool along with the welding wire and stirs the molten pool.

[0011] According to one aspect of this application, the ultrasonic vibration mechanism includes:

[0012] The transducer is electrically connected to the ultrasonic power supply and converts electrical energy into vibration during operation.

[0013] The amplitude transformer, connected to the transducer, is used to concentrate the energy of the transducer and apply it to the tool head;

[0014] The tool head is fixed at one end to the amplitude transformer, and the other end abuts against the wire guide tube of the welding wire during operation, transmitting vibration energy to the welding wire.

[0015] According to one aspect of this application, the welding torch mechanism includes:

[0016] The welding torch body has a wire passage hole through which the welding wire passes and reaches the welding work area;

[0017] The wire feed roller, located at the wire inlet end of the welding torch body, is used to provide power to the welding wire.

[0018] The gas cylinder is connected to the gas channel on the welding torch body and extends to the welding area. During operation, it delivers gas to the welding work area.

[0019] Welding power source, connected to welding wire and substrate.

[0020] According to one aspect of this application, a speckle structured light component that moves with the welding torch is also included, comprising:

[0021] A structural laser emitter is used to guide the laser to the welding work area and to emit the laser into the molten pool area;

[0022] The structured laser receiver receives the laser reflected from the molten pool region and guides it to the laser imaging unit;

[0023] The multi-channel laser processing module includes a vibration channel unit, a molten pool morphology channel unit, a welding temperature channel unit, and an attention module disposed between the channels;

[0024] The data acquired by the laser imaging unit is preprocessed and then sent to the vibration channel unit, the molten pool morphology channel unit, and the welding temperature channel unit for further processing; thus, vibration information, molten pool morphology information, and welding temperature information are obtained.

[0025] According to one aspect of this application, the vibration channel unit includes:

[0026] The molten pool image extraction unit acquires data from the laser imaging unit, performs denoising and enhancement, acquires an image of the molten pool region, and divides it into several overlapping image sub-regions. Each image sub-region contains at least one calculation point and a predetermined number of pixels. The overall shape of each image sub-region is similar.

[0027] The displacement field calculation unit compares the corresponding image sub-regions in the reference molten pool image and the target molten pool image, solves for the displacement vector of each calculation point, and thus obtains the displacement field of the molten pool region.

[0028] The vibration parameter extraction unit extracts vibration parameters based on the variation characteristics of the displacement field.

[0029] According to one aspect of this application, the displacement field calculation unit is further configured to:

[0030] Find the image sub-region in the target molten pool image that is most similar to the image sub-region in the reference molten pool image, and use the coordinate difference of its center point as the initial displacement value;

[0031] Read and perform Taylor expansion on the relevant function or correlation coefficient, and solve the subpixel displacement using the iterative least squares method. During the iterative solution process, perform grayscale interpolation on the subpixel positions in the target image to obtain their grayscale values.

[0032] The initial values ​​are diffused between neighboring calculation points using a reliability-guided method to calculate the total displacement and obtain the final displacement field.

[0033] The final displacement field is smoothed, and different strains and deformations are calculated using different strain and deformation calculation formulas.

[0034] According to one aspect of this application, the welding temperature channel unit is further comprising:

[0035] The temperature calculation unit reads the intensity and wavelength of the preprocessed laser reflection signal and calculates the temperature and absorptivity using a preset model.

[0036] T=hc / (λ×k b ×ln(2hc) 2 / λ 5 IR+1); α=(aI0-bI) R ) / I0;

[0037] h is Planck's constant, c is the speed of light, and k is the speed of light. b Where λ is the Boltzmann constant, I0 is the laser wavelength, and I is the incident laser intensity. R This represents the intensity of the reflected laser; a and b are correction factors.

[0038] The temperature correction unit measures the actual temperature of the molten pool area using an infrared sensor, compares it with the estimated value of the laser reflection signal, and corrects for the error.

[0039] According to one aspect of this application, the molten pool morphology channel unit is further comprising:

[0040] The edge detection unit acquires data from the laser imaging unit, generates an image sequence and converts it into a grayscale image, performs filtering and enhancement processing on the grayscale image, and uses the Sobel operator to extract the edge of the image, obtaining the region in the image where the grayscale value changes beyond the threshold, thus obtaining the edge image of the molten pool.

[0041] The contour extraction unit extracts the contour of the edge image using the boundary tracking method to obtain the contour information of the molten pool, and represents the edge of the molten pool with a set of ordered points or line segments.

[0042] The morphology generation unit constructs a gray-level co-occurrence matrix to extract features from the image or contour of the molten pool, obtaining molten pool parameters including surface roughness, surface tension, and surface temperature distribution, so as to convert the image information of the molten pool into numerical information for subsequent neural network processing.

[0043] According to one aspect of this application, it also includes a temperature field construction module, which calculates the temperature distribution of the actual temperature of the molten pool based on a fitting method, obtains the temperature distribution map and temperature gradient of the molten pool, and analyzes the spatial distribution and variation law of the temperature field of the molten pool.

[0044] According to one aspect of this application, the vibration parameter extraction unit further comprises:

[0045] The final displacement field is smoothed, interpolated, and fitted, and then Fourier transform and wavelet transform are used to extract the vibration characteristics of the molten pool.

[0046] Beneficial effects

[0047] 1. Ultrasonic vibration is transmitted to the molten pool through the wire, making the ultrasonic energy acting on each part of the molten pool more uniform and consistent.

[0048] 2. Ultrasonic vibration is transmitted to the molten pool through the wire, which can avoid direct contact between the amplitude transformer and the deposited layer, and avoid machine damage caused by high contact friction or overheating.

[0049] 3. The simultaneous operation of the ultrasonic mechanism and welding torch allows for greater design flexibility, enabling the fabrication of more complex components.

[0050] 4. By improving droplet transition, promoting molten pool flow, homogenizing molten pool composition, and breaking up coarse grains through cavitation effect, acoustic flow effect, mechanical effect and thermal effect, the additive components can ultimately improve their microhardness, tensile strength and other mechanical properties.

[0051] 5. By using a multi-channel data processing module to acquire multiple characteristic parameters of the molten pool region, the additive manufacturing process can be better tracked and monitored. Attached Figure Description

[0052] Figure 1 This is a system structure diagram of the present invention. Detailed Implementation

[0053] As mentioned in the background section, arc additive manufacturing also faces many unresolved challenges, such as poor molten pool fluidity, uneven composition of additive components, coarse grains, high residual stress, and susceptibility to porosity. These problems not only affect the forming quality of the components but also significantly weaken their mechanical properties and reduce their service life. Among these, the most significant problem in arc additive manufacturing is the tendency for grains to become coarse. This is because the electric arc, as a heat source, generates a large heat input. Furthermore, the arc additive manufacturing environment is at room temperature, causing the molten pool to undergo rapid heating and cooling, resulting in a large temperature gradient within the molten pool. This leads to the formation of coarse columnar grains in the additively formed components.

[0054] like Figure 1 As shown, the device for ultrasonic vibrating wire-assisted fused electrode arc additive manufacturing includes:

[0055] The wire feeding mechanism is used to provide power to the welding wire and transport it towards the welding gun mechanism;

[0056] The welding torch mechanism is used to apply current to the welding wire and move the welding wire to the working area;

[0057] An ultrasonic vibration mechanism is installed between the wire feeding mechanism and the welding torch mechanism. It applies ultrasonic vibration to the welding wire through a transducer and an amplitude transformer, so that during welding, the ultrasonic vibration is transmitted to the molten pool along with the welding wire and stirs the molten pool.

[0058] like Figure 1 As shown, in one embodiment, the robotic arm 3 is connected to the transducer 5 via the upper connector 4. The transducer 5 is connected to the ultrasonic power supply 7 and the amplitude transformer 6. The amplitude transformer 6 is connected to the welding torch 9 via the lower connector. The conductive nozzle 10 is connected to the welding torch 9. The gas cylinder 15 is introduced into the welding torch 9 to provide protective gas for the additive manufacturing process. The positive and negative terminals of the welding power supply 11 are connected to the welding torch 9 and the substrate 12, respectively. 13 is the molten pool, and 14 is the deposition layer.

[0059] Under the pushing action of the wire feeder 1, the wire 2 passes sequentially through the robotic arm 3, the upper connector 4, the ultrasonic vibration mechanism, the lower connector 8, and the welding torch 9, finally reaching the conductive nozzle 10. The ultrasonic action generated by the ultrasonic vibration mechanism is sequentially transmitted to the conductive nozzle 10. During additive manufacturing, the conductive nozzle will have frictional contact with the wire, so the ultrasonic energy will be introduced into the additive molten pool through the wire. The molten pool is subjected to ultrasonic vibration, resulting in a more uniform composition, finer grains, and more uniform heat distribution, thus obtaining a component with good forming quality and excellent mechanical properties.

[0060] In this embodiment, the transducer 5 and the amplitude transformer 6 will need to be redesigned because the filament needs to pass through the center. The through holes of the transducer 5 and the amplitude transformer 6 must be isolated from the filament 2, but cannot be too large to affect the ultrasonic frequency.

[0061] In this embodiment, the upper connector 4 and the lower connector 8 should be made of insulating materials to avoid the welding current affecting the ultrasonic mechanism.

[0062] According to one aspect of this application, the ultrasonic vibration mechanism includes:

[0063] The transducer is electrically connected to the ultrasonic power supply and converts electrical energy into vibration during operation.

[0064] The amplitude transformer, connected to the transducer, is used to concentrate the energy of the transducer and apply it to the tool head;

[0065] The tool head is fixed at one end to the amplitude transformer, and the other end abuts against the wire guide tube of the welding wire during operation, transmitting vibration energy to the welding wire.

[0066] To avoid fatigue fracture caused by ultrasonic vibration, high-strength and high-toughness welding wire is used, and the frequency and amplitude of ultrasonic vibration are controlled within a reasonable range to prevent excessive vibration from damaging the welding wire. Simultaneously, the temperature and stress of the welding wire, as well as the strength and quality of the weld joint, are monitored during the welding process.

[0067] According to one aspect of this application, the welding torch mechanism includes:

[0068] The welding torch body has a wire passage hole through which the welding wire passes and reaches the welding work area;

[0069] The wire feed roller, located at the wire inlet end of the welding torch body, is used to provide power to the welding wire.

[0070] The gas cylinder is connected to the gas channel on the welding torch body and extends to the welding area. During operation, it delivers gas to the welding work area.

[0071] Welding power source, connected to welding wire and substrate.

[0072] In one embodiment, the specific details are as follows: an apparatus for ultrasonic vibration wire-assisted fused electrode arc additive manufacturing includes a wire feeding mechanism, an ultrasonic vibration mechanism, an additive manufacturing mechanism, and a substrate;

[0073] The wire feeding mechanism includes a wire feeder 1, wire 2, wire guide tube 3, and wire pressing roller 8;

[0074] The ultrasonic vibration mechanism includes a transducer 4, an amplitude transformer 6, an ultrasonic power supply 5, and a tool head 7;

[0075] The additive manufacturing mechanism includes a welding torch 9, a welding power source 11, and a gas cylinder 10.

[0076] Wire 2 is fed into wire guide tube 3 by wire feeder 1, and then fed into welding torch 9 through wire pressure roller after exiting wire guide tube 3. Transducer 4 is connected to ultrasonic power supply 7 and amplitude transformer 6, and amplitude transformer 6 is connected to tool head 7. Gas cylinder 10 provides protective gas to welding torch 9 for additive manufacturing process. The positive and negative terminals of welding power supply 11 are connected to welding torch 9 and substrate 12 through wires, respectively. 13 is the molten pool and 14 is the deposition layer.

[0077] The wire passes sequentially through the wire guide tube, drawing wheel, and welding gun under the action of the wire feeder. The ultrasonic vibration generated by the ultrasonic vibration mechanism is transmitted through the following path: ultrasonic power supply - transducer - amplitude transformer - tool head - wire guide tube - wire - molten pool. The ultrasonic energy is introduced into the additive molten pool through the wire.

[0078] The tool head 7 in the ultrasonic vibration mechanism has a through hole, and the diameter of the through hole matches the diameter of the guide wire tube.

[0079] The ultrasonic vibration mechanism transmits ultrasonic vibration to the tool head 7, and the tool head 7 and the guide tube 3 are tightly connected, so the ultrasonic vibration will be transmitted to the wire through the guide tube.

[0080] The transducer 4 and the amplitude transformer 6 in the ultrasonic vibration mechanism vibrate in an up-and-down direction.

[0081] According to one aspect of this application, a speckle structured light component that moves with the welding torch is also included, comprising:

[0082] A structural laser emitter is used to guide the laser to the welding work area and to emit the laser into the molten pool area;

[0083] The structured laser receiver receives the laser reflected from the molten pool region and guides it to the laser imaging unit;

[0084] The multi-channel laser processing module includes a vibration channel unit, a molten pool morphology channel unit, a welding temperature channel unit, and an attention module disposed between the channels;

[0085] The data acquired by the laser imaging unit is preprocessed and then sent to the vibration channel unit, the molten pool morphology channel unit, and the welding temperature channel unit for further processing; thus, vibration information, molten pool morphology information, and welding temperature information are obtained.

[0086] In this embodiment, the synergistic effect of ultrasonic vibration and structured light is effectively utilized to improve welding quality and efficiency, while simultaneously achieving high-precision laser imaging and image processing. Through an attention mechanism, weights are allocated and optimized based on different channel information, thereby improving the accuracy and stability of laser imaging, as well as enhancing the effectiveness and performance of image analysis and processing.

[0087] According to one aspect of this application, the vibration channel unit includes:

[0088] The molten pool image extraction unit acquires data from the laser imaging unit, performs denoising and enhancement, acquires an image of the molten pool region, and divides it into several overlapping image sub-regions. Each image sub-region contains at least one calculation point and a predetermined number of pixels. The overall shape of each image sub-region is similar.

[0089] The purpose of this step is to extract the image of the molten pool region, along with its geometric features and computational points, from the original speckle structured light image. An algorithm based on wavelet transform and thresholding is used to denoise and enhance the original image, eliminating background noise and speckle fringes, and improving the contrast and clarity of the molten pool. An algorithm based on morphology and edge detection is used to extract the molten pool region from the denoised and enhanced image, obtaining its contour and area, as well as its geometric center and centroid. An algorithm based on mesh generation and overlapping segmentation is used to segment the molten pool region image, resulting in several overlapping image sub-regions. Each sub-region contains at least one computational point and a predetermined number of pixels, and the overall shapes of the sub-regions are similar. The geometric center or centroid of each sub-region is used as a computational point for subsequent displacement field calculations.

[0090] The displacement field calculation unit compares the corresponding image sub-regions in the reference molten pool image and the target molten pool image, solves for the displacement vector of each calculation point, and thus obtains the displacement field of the molten pool region.

[0091] The vibration parameter extraction unit extracts vibration parameters based on the variation characteristics of the displacement field.

[0092] Based on the algorithm of Fourier transform and power spectral density estimation, frequency domain analysis is performed on the displacement vector of each calculation point to extract the vibration parameters of the molten pool, including vibration frequency, vibration amplitude, vibration phase, etc.

[0093] Perform a Fourier transform on the displacement vector of each calculation point to obtain its frequency domain representation, i.e., the spectrum of complex values.

[0094] The power spectrum, i.e. the signal power at each frequency component, is obtained by taking the square of the modulus of the spectrum at each calculation point.

[0095] The power spectrum of each calculation point is normalized and divided by the sampling frequency and signal length to obtain its power spectral density, which is the signal power per unit frequency band.

[0096] The power spectral density at each calculation point is analyzed to extract its vibration parameters, such as vibration frequency, vibration amplitude, and vibration phase. These parameters can be determined by peak detection and interpolation of the power spectral density function, or estimated by other frequency domain analysis methods, such as the periodogram method and the Welch method.

[0097] The power spectral density function is estimated using the periodogram method or the Welch method. The effects of energy leakage and spectral leakage are reduced using the Hanning window or the Hamming window. Peak detection and interpolation methods are used to determine the values ​​of vibration parameters.

[0098] The periodogram method treats N sampling points of a random signal as a single signal with finite energy. After performing a discrete Fourier transform, the square of the amplitude is divided by N to estimate the true power spectrum. Its advantage is high computational efficiency, as it does not require calculating the autocorrelation function.

[0099] The Welch method involves segmenting the signal, windowing it, calculating its power spectral density, and then averaging the results. Its advantages include reduced random fluctuations, good convergence, smooth curves, and low variance in the estimated result. However, its disadvantages include a wide main lobe in the power spectrum and low resolution. If the signal sequence is not long enough, the ability of the power spectral density estimate to distinguish different frequency components decreases as each segment becomes shorter.

[0100] The Hanning window or Hamming window multiplies the two ends of a signal by an attenuation coefficient, making the two ends of the signal approach zero, thereby reducing the effects of energy leakage and spectral leakage. Energy leakage refers to the leakage of signal energy from the dominant frequency component to other frequency components, resulting in a wider main lobe and higher side lobes in the power spectrum, leading to increased distortion. Spectral leakage refers to the leakage of a signal's frequency components from one frequency point to an adjacent frequency point, causing a shift in the peak position of the power spectrum and reducing accuracy.

[0101] Peak detection and interpolation: The function is to find the largest peak in the power spectral density function, and then use interpolation to determine the precise location and value of the peak, thereby obtaining vibration parameters such as vibration frequency, vibration amplitude, and vibration phase.

[0102] In this embodiment, the morphology and size of the molten pool, as well as welding quality and defects, can be monitored in real time, providing data support for the control and optimization of the welding process. Vibration parameters of the molten pool can be accurately extracted, and the influence of ultrasonic vibration on the molten pool can be analyzed, providing a basis for the adjustment and matching of ultrasonic vibration. The vibration parameters of the molten pool, and their relationship with physical parameters such as temperature, stress, strength, and mass, can be used to provide a reference for the study of the welding process mechanism and model.

[0103] In another embodiment of this application, two adjacent molten pool region images are obtained from the molten pool image extraction unit, serving as a reference molten pool image and a target molten pool image, respectively. The time interval between these two images is ∆t, determined by the sampling frequency of the laser imaging unit. An algorithm based on digital image correlation technology compares corresponding image sub-regions in the reference molten pool image and the target molten pool image, and solves for the displacement vector at each calculation point, thereby obtaining the displacement field of the molten pool region. A cross-correlation function is used to measure the similarity between the two image sub-regions, the least squares method is used to optimize the solution of the displacement vector, and a multi-resolution and multi-step strategy is used to improve the accuracy and stability of the displacement field.

[0104] According to one aspect of this application, the displacement field calculation unit is further configured to:

[0105] Find the image sub-region in the target molten pool image that is most similar to the image sub-region in the reference molten pool image, and use the coordinate difference of its center point as the initial displacement value;

[0106] To determine the approximate location of an image sub-region within two images, thus providing initial values ​​for subsequent subpixel displacement calculation, the Normalized Correlation Coefficient (NCC) can be used as a similarity metric. It is defined as the product of the covariance of the grayscale values ​​of two image sub-regions divided by their standard deviations. The NCC value ranges from -1 to 1: 1 when the two image sub-regions are identical, -1 when they are completely opposite, and 0 when they are unrelated. Therefore, a sliding window search can be performed in the target melt pool image, using an image sub-region in the reference melt pool image as a template, to find the location that maximizes the NCC, which is then used as the matching position for the image sub-region. Then, the coordinate difference between the center points of the two image sub-regions can be calculated as the initial displacement value.

[0107] Read and perform Taylor expansion on the relevant function or correlation coefficient, and solve the subpixel displacement using the iterative least squares method. During the iterative solution process, perform grayscale interpolation on the subpixel positions in the target image to obtain their grayscale values.

[0108] To improve the accuracy of the displacement, it is necessary to move from the pixel level to the sub-pixel level. The sub-pixel DIC method can be used, which is based on the assumption that within a small displacement range, the grayscale value changes of an image sub-region can be approximated by a Taylor series expansion.

[0109] ,

[0110] Where f(x,y) and g(x,y) represent the gray values ​​in the reference image and the target image, respectively; u and v represent the horizontal and vertical displacements, respectively; ∂f / ∂x and ∂f / ∂y represent the gray-level gradients in the horizontal and vertical directions of the reference image, respectively; and ∂f / ∂x and ∂f / ∂y represent the gray-level gradients in the horizontal and vertical directions of the reference image, respectively. 2 f / ∂x 2 、∂ 2 f / ∂x∂y and ∂ 2 f / ∂y 2 These represent the gray-level curvatures in the horizontal, vertical, and blending directions of the reference image, respectively. Higher-order terms can be ignored, retaining only the first and second-order terms, resulting in...

[0111] ,

[0112] This equation can be viewed as a nonlinear equation about u and v, and then solved using iterative least squares. Specifically, a correlation function or correlation coefficient can be defined to represent the degree of matching between the gray values ​​of two image sub-regions, for example:

[0113] ,

[0114] The goal is to find u and v that minimize R(u,v), which is to solve the following optimization problem: minR(u,v);

[0115] To solve this problem, the Newton-Raphson method can be used. It is an iterative method based on Taylor expansion. Its basic idea is: in each iteration, the objective function is approximated by a quadratic function, the optimal solution is found on the quadratic function, and this optimal solution is used as the initial value for the next iteration until convergence. Specifically, R(u,v) can be expressed as the initial value of the current iteration (u...). k ,v k Performing a Taylor expansion at position ) yields:

[0116] ;

[0117] The initial values ​​are diffused between neighboring calculation points using a reliability-guided method to calculate the total displacement and obtain the final displacement field.

[0118] To improve the reliability of displacement and avoid local matching errors caused by noise or other factors, a reliability-guided method can be employed. This method, based on neighborhood information, works by averaging the displacement values ​​of each calculation point in each iteration, based on the displacement values ​​and correlation coefficients of its neighboring calculation points, thus obtaining a more stable displacement value. Specifically, a reliability function can be defined to represent the confidence level of the displacement value at each calculation point, for example:

[0119]

[0120] Among them, R i C is the reliability function value at the i-th calculation point. i α is the correlation coefficient value at the i-th calculation point, and α is an adjustment parameter used to control the shape of the reliability function curve. The reliability function ranges from [0,1]. When the correlation coefficient is larger, the reliability function is closer to 1, indicating that the displacement value is more reliable; when the correlation coefficient is smaller, the reliability function is closer to 0, indicating that the displacement value is less reliable.

[0121] Then, based on the reliability function value of each calculation point, a weighted average of the displacement values ​​of its neighboring calculation points can be performed to obtain the corrected displacement value, for example...

[0122] , where u i and v i N represents the horizontal and vertical displacement values ​​of the i-th calculation point. i R is the neighborhood set of the i-th computation point.j It is the reliability function value at the j-th calculation point, u j and v j These are the horizontal and vertical displacement values ​​of the j-th calculation point. In this way, neighborhood information can be used to correct the initial displacement value, resulting in a more reliable displacement value.

[0123] The final displacement field is smoothed, and different strains and deformations are calculated using different strain and deformation calculation formulas.

[0124] To eliminate noise and sawtooth effects in the displacement field and obtain a smoother displacement field for more accurate strain and deformation calculations, methods such as Gaussian filtering or bicubic interpolation can be used to smooth the displacement field. Then, different strains and deformations can be calculated based on the derivatives of the displacement field, for example:

[0125]

[0126] Where εxx and εyy are normal strains, γxy is shear strain, Ex and Ey are elongation, and θ is the shear angle. These strains and deformations can reflect the deformation characteristics of the molten pool, thus allowing the extraction of vibration parameters.

[0127] In this embodiment, the basic idea is to divide the reference image (i.e., the undistorted image) and the target image (i.e., the distorted image) into several image sub-regions, and then find the corresponding position of each image sub-region in the two images, thereby calculating the displacement and strain of each image sub-region. To improve the accuracy and reliability of the calculation, DIC typically adopts a multi-scale and multi-step strategy, that is, starting with a large-size and large-step image sub-region, gradually reducing the size and step size until the preset accuracy requirement is achieved. The advantage of DIC is that it does not require special processing of the object surface; as long as the image clarity and contrast are guaranteed, non-contact full-field displacement and strain measurement can be achieved.

[0128] According to one aspect of this application, the welding temperature channel unit is further comprising:

[0129] The temperature calculation unit reads the intensity and wavelength of the preprocessed laser reflection signal and calculates the temperature and absorptivity using a preset model.

[0130] T=hc / (λ×k b ×ln(2hc) 2 / λ 5 IR+1); α=(aI0-bI) R ) / I0;

[0131] h is Planck's constant, c is the speed of light, and k is the speed of light. bWhere λ is the Boltzmann constant, I0 is the laser wavelength, and I is the incident laser intensity. R This represents the intensity of the reflected laser; a and b are correction factors.

[0132] The temperature correction unit measures the actual temperature of the molten pool area using an infrared sensor, compares it with the estimated value of the laser reflection signal, and corrects for the error.

[0133] According to one aspect of this application, the molten pool morphology channel unit is further comprising:

[0134] The edge detection unit acquires data from the laser imaging unit, generates an image sequence and converts it into a grayscale image, performs filtering and enhancement processing on the grayscale image, and uses the Sobel operator to extract the edge of the image, obtaining the region in the image where the grayscale value changes beyond the threshold, thus obtaining the edge image of the molten pool.

[0135] To extract edge information of the molten pool from the original color image, thus providing a foundation for subsequent contour extraction and feature extraction, the following data processing workflow can be adopted:

[0136] Image sequences of the molten pool are acquired from the laser imaging unit. Each image has a resolution of 640×480 pixels, 10 images are acquired per second, and each image is 921.6 KB in size and BMP format.

[0137] Each image is converted to a grayscale image. A grayscale image is an image that only contains brightness information. The value of each pixel is an integer between 0 and 255, representing the degree of black and white, where 0 is black, 255 is white, and the values ​​in between are gray. The size of the grayscale image is 307.2KB, and the image format is BMP.

[0138] Each grayscale image undergoes filtering and enhancement processing. Filtering removes noise and blur from an image, while enhancement improves contrast and detail. Methods such as Gaussian filtering and histogram equalization can be used to filter and enhance grayscale images, resulting in clearer and brighter images.

[0139] Edge extraction is performed on each filtered and enhanced grayscale image. Edges typically represent the contours or surface features of an object. The Sobel operator can be used for edge extraction. The Sobel operator calculates the horizontal and vertical grayscale gradients of each pixel in the image, and then determines whether the pixel is an edge based on the magnitude and direction of the gradient. A threshold can be set; when the gradient magnitude exceeds the threshold, the pixel is considered an edge; otherwise, it is considered a non-edge. This results in a binary image where edge pixels are white and non-edge pixels are black, representing the edge image of the melt pool.

[0140] The contour extraction unit extracts the contour of the edge image using the boundary tracking method to obtain the contour information of the molten pool, and represents the edge of the molten pool with a set of ordered points or line segments.

[0141] To extract the contour information of the melt pool from the edge image, thus providing a basis for subsequent feature extraction, the following data processing workflow can be adopted:

[0142] Contour extraction is performed on each edge image. The boundary tracking method is a neighborhood search-based contour extraction approach. Its basic idea is to arbitrarily select an edge pixel as a starting point and then search along the neighborhood of that edge pixel in a clockwise or counter-clockwise direction until returning to the starting point, forming a closed contour. This yields a set of ordered points or line segments representing the contour information of the melt pool.

[0143] The morphology generation unit constructs a gray-level co-occurrence matrix to extract features from the image or contour of the molten pool, obtaining molten pool parameters including surface roughness, surface tension, and surface temperature distribution, so as to convert the image information of the molten pool into numerical information for subsequent neural network processing.

[0144] Feature extraction is performed on each image or contour of the molten pool. Features typically represent certain attributes or characteristics of the image, such as color, texture, and shape. A gray-level co-occurrence matrix (GLCM) can be used to extract features from the molten pool image or contour. The GLCM reflects the texture features of the image, such as contrast, homogeneity, energy, and entropy. The following data processing workflow can be used to construct the GLCM and extract features:

[0145] Define a direction and a distance, such as the horizontal direction and a distance of 1 pixel. Then, iterate through each pixel in the image, counting the frequency of each pair of adjacent pixel grayscale values ​​to obtain a grayscale frequency matrix. .

[0146] This matrix indicates that in an image, a pixel with a grayscale value of 0 and a pixel with a grayscale value of 1 appear adjacent to each other once, a pixel with a grayscale value of 1 and a pixel with a grayscale value of 2 appear adjacent to each other twice, and so on. This matrix is ​​a form of gray-level co-occurrence matrix, also called a gray-level co-occurrence matrix (GLCM).

[0147] Normalize the gray-level co-occurrence matrix so that each element represents the probability of the corresponding gray-level value combination, resulting in a probability matrix of gray-level values, for example: .

[0148] This matrix indicates that the probability of a pixel with a grayscale value of 0 and a pixel with a grayscale value of 1 appearing next to each other in an image is 0.0625, the probability of a pixel with a grayscale value of 1 and a pixel with a grayscale value of 2 appearing next to each other is 0.125, and so on. This matrix is ​​another form of the gray-level co-occurrence matrix, also called the gray-level probability co-occurrence matrix (GPCM).

[0149] Based on the gray-level co-occurrence matrix or gray-level probability co-occurrence matrix, different texture features, such as contrast, homogeneity, energy, and entropy, are calculated. These features reflect the coarseness, uniformity, and complexity of the image's texture, thereby describing morphological parameters such as surface roughness, surface tension, and surface temperature distribution of the molten pool. The formulas for calculating these features are as follows:

[0150]

[0151] Where N is the number of gray values, and p(i,j) is the value of the (i,j)th element of the gray-level co-occurrence matrix or gray-level probability co-occurrence matrix. The values ​​of these features range from [0,1]. When the texture of the image is coarser, more uneven, and more complex, the contrast and entropy are greater, while the homogeneity and energy are smaller; and vice versa.

[0152] According to one aspect of this application, it also includes a temperature field construction module, which calculates the temperature distribution of the actual temperature of the molten pool based on a fitting method, obtains the temperature distribution map and temperature gradient of the molten pool, and analyzes the spatial distribution and variation law of the temperature field of the molten pool.

[0153] First, the molten pool needs to be divided into several grids. Then, at the center of each grid, a temperature data point is interpolated and used as the temperature value for that grid. Next, based on all the temperature data points, a temperature function is obtained using polynomial fitting, spline fitting, or other fitting methods to represent the relationship between temperature and spatial coordinates. In this way, a continuous temperature distribution function can be obtained, representing how the temperature of the molten pool changes with spatial location.

[0154] First, a temperature distribution map of the molten pool needs to be plotted based on the temperature distribution function, representing how the temperature of the molten pool changes with spatial location. Different colors or shades of gray can be used to represent different temperature ranges; for example, red represents a high-temperature region, and blue represents a low-temperature region. Then, the temperature gradient of the molten pool is calculated based on the derivative of the temperature distribution function, representing the rate of temperature change with spatial location. The temperature gradient can be represented by the direction and length of an arrow; for example, the direction of the arrow indicates the direction of temperature change, and the length of the arrow indicates the magnitude of the temperature change.

[0155] First, observe the temperature distribution map and temperature gradient to identify the high-temperature and low-temperature zones of the molten pool, as well as the trends and patterns of temperature changes. Then, based on the principles of thermodynamics and fluid mechanics, analyze the characteristics of heat flow and heat transfer in the molten pool, such as heat convection, heat diffusion, and heat radiation. Finally, based on the analysis results, evaluate the impact of the temperature field of the molten pool on its morphology, stability, and quality, and propose optimization suggestions or control strategies.

[0156] According to one aspect of this application, the vibration parameter extraction unit further comprises:

[0157] The final displacement field is smoothed, interpolated, and fitted, and then Fourier transform and wavelet transform are used to extract the vibration characteristics of the molten pool.

[0158] Perform a Fourier transform on the displacement function to obtain the spectrum of the molten pool vibration, i.e., the main frequency components and amplitude of the molten pool vibration; perform a wavelet transform on the displacement function to obtain the time-frequency diagram of the molten pool vibration, i.e., how the frequency of the molten pool vibration changes over time; based on the spectrum and the time-frequency diagram, extract the characteristic parameters of the molten pool vibration, such as vibration frequency, amplitude, phase, energy, entropy, etc.

[0159] First, the displacement function is sampled, meaning the function value is recorded at regular time intervals to obtain a discrete displacement sequence. The sampling time interval should satisfy the Nyquist sampling theorem, that is, the sampling frequency should be greater than or equal to twice the highest frequency of the displacement signal to avoid confusion or distortion. For example, if the highest frequency of the displacement signal is 100Hz, then the sampling frequency should be at least 200Hz, that is, the displacement value is recorded once every 0.005 seconds. Assuming the total sampling time is 1 second, then the number of sampling points is 200, resulting in a displacement sequence of length 200, denoted as x[n], where n=0,1,...,199.

[0160] Next, a Fourier transform is performed on the displacement sequence, i.e., the Discrete Fourier Transform (DFT) of the displacement sequence is calculated. The DFT can decompose the displacement sequence into a superposition of sine and cosine waves of different frequencies, resulting in a complex sequence denoted as X[k], where k = 0, 1, ..., 199. The magnitude of X[k] represents the amplitude of the k-th frequency component, and the argument of X[k] represents the phase of the k-th frequency component.

[0161] Because the DFT is computationally intensive, the Fast Fourier Transform (FFT) is typically used to accelerate the calculation. The FFT is an optimized algorithm that utilizes properties such as symmetry and periodicity to reduce the time complexity of the DFT from O(N^2) to O(NlogN). In MATLAB, the `fft` function can be used to compute the FFT.

[0162] Finally, the results of the Fourier transform are analyzed, i.e., the spectrum of the displacement signal is plotted to represent its frequency distribution and characteristics. Since X[k] is a complex sequence, it can be represented in different ways, such as by complex modulus, complex argument, real part, and imaginary part. Typically, the complex modulus is used to represent the amplitude spectrum of the displacement signal, and the complex argument is used to represent its phase spectrum. Because the displacement signal is a real signal, X[k] is a symmetrical complex sequence, i.e., the complex conjugate of X[k] = X[Nk], where Nk represents the negative remainder modulo N. Therefore, only the first half of X[k] needs to be plotted, i.e., k = 0, 1, ..., N / 2, to represent the complete spectral information of the displacement signal. Additionally, the frequency values ​​corresponding to X[k] need to be determined, i.e., the actual frequencies of the k frequency components. This depends on the sampling frequency and the number of sampling points. According to the definition of DFT, the frequency value corresponding to X[k] is: f k =kfracF s N;

[0163] Among them, F s Where f is the sampling frequency and N is the number of sampling points. Therefore, the amplitude spectrum of the displacement signal can be plotted based on the magnitudes of fk and X[k]. k Plot the phase spectrum of the displacement signal using the argument of X[k].

[0164] In a further embodiment, multi-channel data is input into a neural network module using an attention mechanism to construct a neural network model. A neural network module can be added to collect data from each channel, and then the data is processed using an attention mechanism. The specific data processing procedure is as follows: Vibration information, molten pool morphology information, and welding temperature information are used as inputs to the neural network module, and their respective feature vectors are extracted through different convolutional and pooling layers. The three feature vectors are concatenated into a long vector, which is then passed through a fully connected layer to obtain a comprehensive feature vector. Using the attention mechanism, the correlation between the comprehensive feature vector and the feature vectors of each channel is calculated to obtain an attention weight vector, representing the importance of each channel to welding quality. Based on the attention weight vector, the feature vectors of each channel are weighted and summed to obtain a final feature vector, which is used to predict welding quality evaluation indicators, such as weld width, depth, and shape. This effectively utilizes the information from each channel while considering the relationships between them, improving the accuracy of welding quality prediction. In some embodiments, CNN or GCN network modules can be used for implementation.

[0165] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details of the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the protection scope of the present invention.

Claims

1. A device for ultrasonically vibrating wire-assisted fused electrode arc additive manufacturing, characterized in that, include: The wire feeding mechanism is used to provide power to the welding wire and transport it towards the welding gun mechanism; The welding torch mechanism is used to apply current to the welding wire and move the welding wire to the working area; An ultrasonic vibration mechanism is set between the wire feeding mechanism and the welding torch mechanism. It applies ultrasonic vibration to the welding wire through a transducer and an amplitude transformer, so that during welding, the ultrasonic vibration is transmitted to the molten pool along with the welding wire and stirs the molten pool. The ultrasonic vibrating wire-assisted consumable electrode arc additive manufacturing device further includes a speckle structured light component that moves with the welding torch, comprising: A structural laser emitter is used to guide the laser to the welding work area and to emit the laser into the molten pool area; The structured laser receiver receives the laser reflected from the molten pool region and guides it to the laser imaging unit; The multi-channel laser processing module includes a vibration channel unit, a molten pool morphology channel unit, a welding temperature channel unit, and an attention module disposed between the channels; Data from the laser imaging unit is acquired and preprocessed, then sent to the vibration channel unit, molten pool morphology channel unit, and welding temperature channel unit for further processing; vibration information, molten pool morphology information, and welding temperature information are obtained. The vibration channel unit includes: The molten pool image extraction unit acquires data from the laser imaging unit, performs denoising and enhancement, acquires an image of the molten pool region, and divides it into several overlapping image sub-regions. Each image sub-region contains at least one calculation point and a predetermined number of pixels. The overall shape of each image sub-region is similar. The displacement field calculation unit compares the corresponding image sub-regions in the reference molten pool image and the target molten pool image, solves for the displacement vector of each calculation point, and thus obtains the displacement field of the molten pool region. The vibration parameter extraction unit extracts vibration parameters based on the variation characteristics of the displacement field. The displacement field calculation unit is further used for: Find the image sub-region in the target molten pool image that is most similar to the image sub-region in the reference molten pool image, and use the coordinate difference of its center point as the initial displacement value; Read and perform Taylor expansion on the relevant function or correlation coefficient, and solve the subpixel displacement using the iterative least squares method. During the iterative solution process, perform grayscale interpolation on the subpixel positions in the target image to obtain their grayscale values. The initial values ​​are diffused between neighboring calculation points using a reliability-guided method to calculate the total displacement and obtain the final displacement field. The final displacement field is smoothed, and different strains and deformations are calculated using different strain and deformation calculation formulas.

2. The apparatus for ultrasonic vibrating wire-assisted fused electrode arc additive manufacturing as described in claim 1, characterized in that, The ultrasonic vibration mechanism includes: The transducer is electrically connected to the ultrasonic power supply and converts electrical energy into vibration during operation. The amplitude transformer, connected to the transducer, is used to concentrate the energy of the transducer and apply it to the tool head; The tool head is fixed at one end to the amplitude transformer, and the other end abuts against the wire guide tube of the welding wire during operation, transmitting vibration energy to the welding wire.

3. The apparatus for ultrasonic vibrating wire-assisted fused electrode arc additive manufacturing as described in claim 1, characterized in that, The welding torch mechanism includes: The welding torch body has a wire passage hole through which the welding wire passes and reaches the welding work area; The wire feed roller, located at the wire inlet end of the welding torch body, is used to provide power to the welding wire; The gas cylinder is connected to the gas channel on the welding torch body and extends to the welding area. During operation, it delivers gas to the welding work area. Welding power source, connected to welding wire and substrate.

4. The apparatus for ultrasonic vibrating wire-assisted fused electrode arc additive manufacturing as described in claim 1, characterized in that, The welding temperature channel unit is further defined as follows: The temperature calculation unit reads the intensity and wavelength of the preprocessed laser reflection signal and calculates the temperature and absorptivity using a preset model. T = hc / (λ×k b ×ln(2hc 2 / λ 5 AND R +1))(α=(aI0-bI R ) / I0: h is Planck's constant, c is the speed of light, and k is the speed of light. b Where λ is the Boltzmann constant, I0 is the laser wavelength, and I is the incident laser intensity. R This represents the intensity of the reflected laser; a and b are correction factors. The temperature correction unit measures the actual temperature of the molten pool area using an infrared sensor, compares it with the estimated value of the laser reflection signal, and corrects for the error.

5. The apparatus for ultrasonic vibrating wire-assisted fused electrode arc additive manufacturing as described in claim 1, characterized in that, The molten pool morphology channel unit is further defined as follows: The edge detection unit acquires data from the laser imaging unit, generates an image sequence and converts it into a grayscale image, performs filtering and enhancement processing on the grayscale image, and uses the Sobel operator to extract the edge of the image, obtaining the region in the image where the grayscale value changes beyond the threshold, thus obtaining the edge image of the molten pool. The contour extraction unit extracts the contour of the edge image using the boundary tracking method to obtain the contour information of the molten pool, and represents the edge of the molten pool with a set of ordered points or line segments. The morphology generation unit constructs a gray-level co-occurrence matrix to extract features from the image or contour of the molten pool, obtaining molten pool parameters including surface roughness, surface tension, and surface temperature distribution, so as to convert the image information of the molten pool into numerical information for subsequent neural network processing.

6. The apparatus for ultrasonic vibrating wire-assisted fused electrode arc additive manufacturing as described in claim 4, characterized in that, It also includes a temperature field construction module, which calculates the temperature distribution of the molten pool based on the fitting method, obtains the temperature distribution map and temperature gradient of the molten pool, and analyzes the spatial distribution and variation law of the temperature field of the molten pool.

7. The apparatus for ultrasonic vibrating wire-assisted fused electrode arc additive manufacturing as described in claim 1, characterized in that, The vibration parameter extraction unit is further configured as follows: The final displacement field is smoothed, interpolated, and fitted, and then Fourier transform and wavelet transform are used to extract the vibration characteristics of the molten pool.

Images