A SLM printing method and device for real-time defect detection

Abstract

The present application relates to the technical field of metal additive manufacturing, and in particular to a SLM printing method and device for real-time defect detection, wherein the method comprises the following steps: S1, cladding metal powder according to the laser printing strategy of the nth layer; S2, performing defect detection on the nth layer of the part to be printed, and outputting the defect detection result; S3, determining the laser repair strategy of the nth layer according to the defect detection result. In step S2, the present application predicts that the voxel units whose defect detection results exceed a preset threshold value are marked as structural defects through a pre-trained mechanical detection model. If there are no defects in the nth layer, continue to print the n+1th layer. The present application performs real-time defect detection on SLM printing through a neural network model, and can quickly determine whether there are pores in each layer that affect the mechanical properties of the printed parts. Non-structural defects are not processed, which can speed up the SLM printing speed.

Description

Technical Field

[0001] The present application relates to the technical field of metal additive manufacturing, and in particular to a SLM printing method and device for real-time defect detection. Background Art

[0002] In metal 3D printing technology, selective laser melting (i.e., SLM) can produce metal parts with close to full density, whose mechanical properties such as tensile strength can even reach the level of forgings, and their micro-Vickers hardness can be higher than that of forgings.

[0003] The principle of SLM technology is: metal powder is laid layer by layer on the printing platform, and the laser beam melts the sliced ​​area of ​​the printed part model. When all the slices are melted layer by layer, the formed printed part can be obtained.

[0004] During SLM printing, the pores of each layer need to be detected and repaired after cladding. In current SLM technology, there is over-repair of non-structural defects. Some pore defects do not affect the mechanical properties of printed parts, but over-repair will increase printing time and lead to material waste.

[0005] In order to speed up SLM printing and ensure the success rate of printing, the present application provides an SLM printing method and device for real-time defect detection. Summary of the invention

[0006] In order to overcome the problems existing in the related art, the first aspect of the present application provides a SLM printing method for real-time defect detection, comprising:

[0007] S1, cladding metal powder according to the laser printing strategy of the nth layer;

[0008] S2, performing defect detection on the nth layer of the part to be printed, and outputting the defect detection result;

[0009] S3, determining a laser repair strategy for the nth layer according to the defect detection result; n is an integer greater than or equal to 1; wherein S2 specifically includes:

[0010] S21, obtaining the printed part image data of the nth layer;

[0011] S22, detecting pore defects on the printed part image data to obtain pore defect features;

[0012] S23, constructing printed part voxel data according to the three-dimensional part data and the pore defect characteristics;

[0013] S24, encoding the printed part voxel data and inputting it into a mechanical detection model, and outputting the defect detection result.

[0014] In one implementation, S21 specifically includes:

[0015] S211, obtaining a platform surface image of the nth layer;

[0016] S212, performing contour recognition and segmentation on the platform surface image to obtain first printed part image data; the first printed part image data is an image of the cladding area;

[0017] S213, performing CT scanning on the cladding area to obtain second printed part image data.

[0018] In one implementation, S22 specifically includes:

[0019] S221, training a pore defect detection model;

[0020] S222, preprocessing the first printing part image data and the second printing part image data;

[0021] S223, inputting the first printed part image data and the second printed part image data into the pore defect detection model to obtain pore defect features.

[0022] In one embodiment, S23 includes:

[0023] S231, acquiring the three-dimensional data of the part and the pore defect characteristics;

[0024] S232, constructing the printed part voxel data.

[0025] In one embodiment, S24 includes:

[0026] S241, encoding each voxel unit in the printed part voxel data;

[0027] S242, construct and train a mechanical detection model;

[0028] S243, inputting the encoded voxel unit into the mechanical detection model, and outputting the defect detection result;

[0029] In one embodiment, the loss function of the mechanical detection model is:

[0030]

[0031] in, For the The stress prediction value of each voxel unit, For the The true value of stress per voxel.

[0032] In one embodiment, The matrix of stress prediction values ​​of the voxel units is expressed as:

[0033]

[0034] in, Indicates The stress prediction value of each voxel unit, represents the normal stress along the x-axis, represents the normal stress along the y-axis, represents the normal stress along the z-axis, and represents the shear stress on the xoy plane, and represents the shear stress in the xoz plane, and represents the shear stress in the zoy plane.

[0035] In one implementation, S3 specifically includes:

[0036] The defect type of the nth layer is determined according to the defect detection result of the nth layer. If the defect type is a structural defect, the nth layer is clad and reprinted; if the defect type is a non-structural defect, the laser printing strategy of the n+1th layer is obtained.

[0037] The second aspect of the present application provides an SLM printing device for real-time defect detection, which is used to execute the SLM printing method described in the first aspect of the present application, comprising: a laser printing module, a defect detection module and a printing control module;

[0038] The laser printing module is used to clad metal powder layer by layer according to the laser printing strategy;

[0039] The defect detection module is used to perform defect detection on the nth layer of the part to be printed and output the defect detection result;

[0040] The printing control module is electrically connected to the laser printing module and the defect detection module, and is used to determine the laser printing strategy according to the defect detection result.

[0041] The technical solution provided by this application includes the following beneficial effects:

[0042] When performing SLM layer-by-layer printing, metal powder cladding is performed according to the laser printing strategy of the nth layer. After the metal powder solidifies, the pore defect detection is performed through the printed part image data of this layer and encoded as pore defect features. The printed part voxel data is constructed based on the part 3D data and pore defect features. After the printed part voxel data is encoded, it is input into the pre-trained mechanical detection model. The defect detection result of each voxel unit is obtained through the mechanical detection model. If the pore defect in one of the voxel units is a structural defect, the pore defect in the nth layer needs to be repaired; if the pore defect of the current voxel unit does not affect the structural strength, the n+1 layer continues to be printed.

[0043] This application uses a neural network model to perform real-time defect detection on SLM printing, which can quickly determine whether there are pores in each layer that affect the mechanical properties of the printed parts. Non-structural defects are not processed, which can speed up the SLM printing speed.

[0044] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The above and other objects, features and advantages of the present application will become more apparent through a more detailed description of exemplary embodiments of the present application in conjunction with the accompanying drawings, wherein the same reference numerals generally represent the same components in the exemplary embodiments of the present application.

[0046] Figure 1 A schematic diagram of the process of the SLM printing method shown in the embodiment of the present application;

[0047] Figure 2 Schematic diagram of the process of step S2 in the SLM printing method shown in the embodiment of the present application;

[0048] Figure 3 A schematic diagram of the network structure of the mechanical detection model shown in the embodiment of the present application;

[0049] Figure 4 Schematic diagram of the logical structure of the SLM printing device shown in the embodiment of the present application. DETAILED DESCRIPTION

[0050] The preferred embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the preferred embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art.

[0051] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0052] It should be understood that although the terms "first", "second", "third", etc. may be used in this application to describe various information, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of this application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0053] Embodiment 1

[0054] In metal 3D printing technology, selective laser melting (i.e., SLM) can produce metal parts with close to full density, whose mechanical properties such as tensile strength can even reach the level of forgings, and their micro-Vickers hardness can be higher than that of forgings.

[0055] The principle of SLM technology is: metal powder is laid layer by layer on the printing platform, and the laser beam melts the sliced ​​area of ​​the printed part model. When all the slices are melted layer by layer, the formed printed part can be obtained.

[0056] Therefore, in order to reduce metallurgical holes and keyholes during printing, the existing SLM technology will detect the pores slice by slice. When there are too many pores, the pores need to be filled and the scanning speed is reduced. Although this method can reduce the number of pores in the printed parts and improve the printing success rate, it further reduces the printing speed. The current SLM technology printing process usually takes more than 20 hours.

[0057] In the current SLM technology, there is excessive repair of non-structural defects. Some pore defects do not affect the mechanical properties of printed parts, but excessive repair will increase printing time and cause material waste.

[0058] In order to speed up SLM printing and ensure the success rate of printing, the present application embodiment provides a SLM printing method for real-time defect detection, such as Figure 1 As shown, the following steps are included:

[0059] S1, cladding metal powder layer by layer according to the laser printing strategy;

[0060] S2, performing defect detection on the nth layer of the part to be printed, and outputting the defect detection result;

[0061] S3. Determine the laser repair strategy according to the defect detection result; n is an integer greater than or equal to 1.

[0062] It can be understood that the SLM printing method of the embodiment of the present application is executed by an SLM printing device, and the SLM printing device includes a laser printing module, a defect detection module, and a printing control module.

[0063] Among them, the laser printing module is used to melt the metal powder layer by layer according to the laser printing strategy; the defect detection module is used to perform defect detection on the nth layer of the part to be printed and output the defect detection result; the printing control module is used to determine the laser printing strategy according to the data to be printed and the defect detection result.

[0064] First, the laser printing module is provided with a printing platform, a powder spreading mechanism and a laser cladding mechanism. The laser printing module is communicatively connected to the printing control module. After the laser printing module obtains the laser printing strategy of the nth layer, the powder spreading mechanism spreads the nth layer of powder on the printing platform. Finally, the laser cladding mechanism clads the nth layer contour of the part to be printed by laser.

[0065] In the layer-by-layer printing process, Figure 2 As shown, the defect detection module performs defect detection on the nth layer based on step S2.

[0066] Specifically, the defect detection module is provided with a camera module and a CT scanning module, which are used to obtain the surface image and internal structure of the part to be printed.

[0067] Specifically, step S2 specifically includes the following steps:

[0068] S21, obtaining the printed part image data of the nth layer;

[0069] S22, detecting pore defects on the printed part image data to obtain pore defect features;

[0070] S23, constructing printed part voxel data according to the three-dimensional part data and the pore defect characteristics;

[0071] S24, encoding the printed part voxel data and inputting it into a mechanical detection model, and outputting the defect detection result.

[0072] Wherein, step S21 includes:

[0073] S211, obtaining a platform surface image of the nth layer;

[0074] S212, performing contour recognition and segmentation on the platform surface image to obtain first printed part image data; the first printed part image data is an image of the cladding area;

[0075] S213, performing CT scanning on the cladding area to obtain second printed part image data.

[0076] In an embodiment of the present application, the camera module captures images of the printed part molding area and the metal powder area on the printing platform, that is, the platform surface image.

[0077] Therefore, it is necessary to perform contour recognition on the printed part molding area and segment it to obtain the first printed part image data.

[0078] In an embodiment of the present application, the SLM-printed part includes surface pores and internal pores. The embodiment of the present application detects the surface pores through the first printed part image data and detects the internal pores through the second printed part image data.

[0079] Specifically, S22 includes:

[0080] S221, training a pore defect detection model;

[0081] S222, preprocessing the first printing part image data and the second printing part image data;

[0082] Specifically, the pore defect detection model is constructed based on the yolo-v8n model, including a backbone structure, a neck structure and a detection head. The backbone network is used to extract features of the printed part structure, the neck structure is used to connect the backbone structure and the detection head; the detection head is used to output the pore defect coordinates.

[0083] S223, inputting the first printed part image data and the second printed part image data into the pore defect detection model to obtain pore defect features.

[0084] In the embodiment of the present application, the pore defect detection model is constructed based on the original yolo-v8n, and the color images and CT images of SLM printed parts with surface pores and internal pores are used as training sets and test sets to train the pore defect detection model. The pore defect detection model marks the internal or surface pores and outputs the coordinates, contour information and confidence of the pores.

[0085] Different from traditional convolutional neural networks, the pore defect detection model built in this application based on yolo-v8n can achieve a detection speed of 30 frames per second.

[0086] Further, S23, constructing printed part voxel data according to the part three-dimensional data and the pore defect characteristics, specifically including:

[0087] S231, acquiring the three-dimensional data of the part and the pore defect characteristics;

[0088] S232, constructing voxel data of printed parts.

[0089] In an embodiment of the present application, the pore defect features detected are added to the printed part voxel data generated from the three-dimensional data of the part.

[0090] It can be understood that the three-dimensional part data is data used to store and represent a three-dimensional model.

[0091] Specifically, the voxel data of the printed part is pixels in a three-dimensional space. That is, the three-dimensional space is divided into grid units of fixed volume. Each grid unit in the three-dimensional space is then filled according to the structure of the printed part. The voxel units are stacked to form the structure of the part to be printed as described in the embodiment of the present application.

[0092] In order to facilitate learning the structural features of a three-dimensional entity through a neural network model and complete the prediction of mechanical properties, in an embodiment of the present application, the currently printed n-layer printed part generates printed part voxel data, and the printed part voxel data includes pore defects.

[0093] In step S24, S24 specifically includes:

[0094] S241, encoding each voxel unit in the printed part voxel data;

[0095] Specifically, The encoding of each voxel unit is .in, , , Respectively represent three-dimensional coordinates; Indicates the state of an entity, either occupied or unoccupied.

[0096] In the embodiment of the present application, the three-dimensional coordinates of the voxel unit are the center coordinates.

[0097] S242, construct and train a mechanical detection model;

[0098] Specifically, Figure 3 As shown, the mechanical detection model is provided with 3 layers of fully connected layers, and Softmax activation functions are provided between the fully connected layers.

[0099] In the embodiment of the present application, the physical loss function of the mechanical detection model is:

[0100]

[0101] in, For the The stress prediction value of each voxel unit, For the The true value of stress per voxel.

[0102] In the embodiment of the present application, the mechanical detection model uses the physical loss function as a constraint condition of the neural network so that the output of the neural network can conform to the three-dimensional stress relationship. The three-dimensional stress relationship refers to the stress state of any voxel unit inside the printed part in three-dimensional space.

[0103] In the embodiment of the present application, each voxel unit involves calculating the stress tensor of nine components, namely three normal stress components and six shear stress components on the surface. i The force on each pixel unit is expressed as Matrix representation:

[0104]

[0105] in, Indicates The stress prediction value of each voxel unit, represents the normal stress along the x-axis, represents the normal stress along the y-axis, represents the normal stress along the z-axis, and represents the shear stress on the xoy plane, and represents the shear stress in the xoz plane, and represents the shear stress in the zoy plane.

[0106] It is understandable that in physics and engineering, stress is divided into nine components, including three normal stresses and six shear stresses. The voxel unit is a cube, and the voxel unit has six faces: xy, yx, xz, yz, zx, and zy. On the voxel unit, the forces applied perpendicular to the xyz axis are three normal stresses, and the stresses parallel to the above six faces are the shear stresses of the voxel unit.

[0107] In an embodiment of the present application, the mechanical detection model predicts 9 stress components of each voxel unit.

[0108] That is, , , ) represents the principal stress component on the xyz axis. , , , , , ) represents the shear stress components of the six faces of the voxel unit. This is used to detect and output the voxel unit code that exceeds the preset yield strength.

[0109] During the training phase, the true value of normal stress and shear stress of each voxel in the training set is calculated using finite element analysis software.

[0110] Taking metal rods as an example, the calculation formula for the true value of normal stress is:

[0111]

[0112] in, is the force in the x direction, represents the bending moment on the x-plane, express To The distance between individual voxel units, Indicates The projected area of ​​the voxel unit on the x-plane.

[0113] Furthermore, and represents the adjustment parameter, in the case of loads such as only tension or compression, is 0.

[0114] Similarly, taking a metal rod as an example, when the metal rod is subjected to shear stress caused by bending moment or torque, and The weight used as the adjustment parameter for adjustment calculation.

[0115] Specifically, the calculation formula for the true value of shear stress is:

[0116]

[0117] in, For the Shear force of a voxel unit in the xoy plane, Indicates The static moment of the individual pixel unit about the central axis; represents the moment of inertia of the xoy surface.

[0118] in, is the torque, is the distance from the voxel unit to the neutral axis, is the polar moment of inertia of the printed part.

[0119] In the embodiments of the present application, Indicates the shear force on the printed part when it is bent. Represents the shear force caused by torque.

[0120] S243, inputting the encoded voxel unit into the mechanical detection model, and outputting the defect detection result;

[0121] In the embodiments of the present application, Figure 3 As shown, Vector representation of voxel units , Pre-facilitation , preset torque And preset torque Input to the fully connected layer, and the stress prediction value of each voxel unit is predicted through the activation function .

[0122] During training, The gradient is optimized through the loss function until the output is a stress prediction value that meets the three-dimensional stress balance. The individual pixel units If it is greater than or equal to the preset threshold, the output defect detection result y is a structural defect; if it is less than the preset threshold, the output defect detection result y is a non-structural defect.

[0123] In the embodiment of the present application, step S3 specifically includes:

[0124] The defect type of the nth layer is determined according to the defect detection result of the nth layer. If the defect type is a structural defect, a laser repair strategy is executed to perform cladding and reprinting on the nth layer; if the defect type is a non-structural defect, the laser printing strategy of the n+1th layer is obtained.

[0125] In the embodiment of the present application, the defect detection result is predicted by a pre-trained mechanical detection model. The voxel unit output by the mechanical detection model that exceeds a preset threshold is marked as a structural defect. At this time, the pore defect in the nth layer needs to be repaired. If there is no structural defect in the nth layer, continue to print the n+1th layer.

[0126] The embodiment of the present application performs real-time defect detection on SLM printing through a neural network model, and can quickly determine whether there are pores in each layer that affect the mechanical properties of the printed parts. Non-structural defects are not processed, which can speed up the SLM printing speed.

[0127] Embodiment 2

[0128] A SLM printing device for real-time defect detection is used to execute the SLM printing method described in the first embodiment, and its structure is as follows: Figure 4 As shown, it includes: a laser printing module, a defect detection module and a printing control module;

[0129] The laser printing module is used to clad metal powder layer by layer according to the laser printing strategy;

[0130] The defect detection module is used to perform defect detection on the nth layer of the part to be printed and output the defect detection result;

[0131] The printing control module is electrically connected to the laser printing module and the defect detection module, and is used to determine the laser printing strategy according to the defect detection result.

[0132] Regarding the device in the above embodiment, the specific way in which each module performs the operation has been described in detail in the embodiment of the method, and will not be elaborated here.

[0133] The scheme of the present application has been described in detail above with reference to the accompanying drawings. In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments. Those skilled in the art should also be aware that the actions and modules involved in the description are not necessarily required for the present application.

[0134] In addition, it can be understood that the steps in the method of the embodiment of the present application can be adjusted in order, combined and deleted according to actual needs, and the modules in the device of the embodiment of the present application can be combined, divided and deleted according to actual needs.

[0135] In addition, the method according to the present application may also be implemented as a computer program or a computer program product, which includes computer program code instructions for executing some or all of the steps in the above method of the present application.

[0136] Alternatively, the present application can also be implemented as a non-temporary machine-readable storage medium (or computer-readable storage medium, or machine-readable storage medium) on which executable code (or computer program, or computer instruction code) is stored. When the executable code (or computer program, or computer instruction code) is executed by a processor of an electronic device (or electronic device, server, etc.), the processor executes part or all of the steps of the above-mentioned method according to the present application.

[0137] Those skilled in the art will further appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the application herein may be implemented as electronic hardware, computer software, or combinations of both.

[0138] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system and method according to multiple embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a part of a module, a program segment or a code, and the part of the module, the program segment or the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or the flow chart, and the combination of the square boxes in the block diagram and / or the flow chart can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0139] The embodiments of the present application have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A SLM printing method for real-time defect detection, characterized in that: include: S1, cladding metal powder according to the laser printing strategy of the nth layer; S2, performing defect detection on the nth layer of the part to be printed, and outputting the defect detection result; S3, determining a laser repair strategy for the nth layer according to the defect detection result; n is an integer greater than or equal to 1; wherein S2 specifically includes: S21, obtaining the printed part image data of the nth layer; S22, performing pore defect detection on the printed part image data to obtain pore defect features; S23, constructing printed part voxel data according to the three-dimensional part data and the pore defect characteristics; S24, encoding the printed part voxel data and inputting it into a mechanical detection model, and outputting the defect detection result; S24 includes: S241, encoding each voxel unit in the printed part voxel data; S242, construct and train a mechanical detection model; S243, inputting the encoded voxel unit into the mechanical detection model, and outputting the defect detection result; The loss function of the mechanical detection model is: in, For the The stress prediction value of each voxel unit, For the The true value of stress of each voxel unit; No. The matrix of stress prediction values ​​of the voxel units is expressed as: in, Indicates The stress prediction value of each voxel unit, represents the normal stress along the x-axis, represents the normal stress along the y-axis, represents the normal stress along the z-axis, and represents the shear stress on the xoy plane, and represents the shear stress in the xoz plane, and represents the shear stress in the zoy plane; S3 specifically includes: The defect type of the nth layer is determined according to the defect detection result of the nth layer. If the defect type is a structural defect, the nth layer is clad and reprinted; if the defect type is a non-structural defect, the laser printing strategy of the n+1th layer is obtained.

2. The SLM printing method for real-time defect detection according to claim 1, characterized in that: S21 specifically includes: S211, obtaining a platform surface image of the nth layer; S212, performing contour recognition and segmentation on the platform surface image to obtain first printed part image data; the first printed part image data is an image of the cladding area; S213, performing CT scanning on the cladding area to obtain second printed part image data.

3. The SLM printing method for real-time defect detection according to claim 2, characterized in that: S22 specifically includes: S221, training a pore defect detection model; S222, preprocessing the first printing part image data and the second printing part image data; S223, inputting the first printed part image data and the second printed part image data into the pore defect detection model to obtain pore defect features.

4. The SLM printing method for real-time defect detection according to claim 3, characterized in that: S23 includes: S231, acquiring the three-dimensional data of the part and the pore defect characteristics; S232, constructing the printed part voxel data.

5. A SLM printing device for real-time defect detection, used to execute the SLM printing method according to any one of claims 1 to 4, characterized in that: include: Laser printing module, defect detection module and printing control module; The laser printing module is used to clad metal powder layer by layer according to the laser printing strategy; The defect detection module is used to perform defect detection on the nth layer of the part to be printed and output the defect detection result; The printing control module is electrically connected to the laser printing module and the defect detection module, and is used to determine the laser printing strategy according to the defect detection result.

Images