An on-line detection method and device for internal three-dimensional defects during additive manufacturing of components

Abstract

The present invention relates to the technical field of component additive manufacturing, and discloses a method for online detection of internal three-dimensional defects during component additive manufacturing, comprising the following steps: printing a component, wherein a detection device follows the printing movement during component printing; scanning the print layers printed by a print head to obtain point cloud data of each print layer; removing invalid points in the obtained layer point cloud data and converting the data into a PCD point cloud format, sampling and performing point cloud data preprocessing to obtain single-layer point cloud data; detecting and quantitatively calculating defective areas of the single-layer point cloud data, and constructing a parameter matrix based on defect characteristics; and calculating the average height of the point cloud layer of each single layer of the component. The component additive manufacturing process of the present invention is a layer-by-layer manufacturing process, and the detection of internal defects can just utilize the basic principle process of the additive manufacturing process, that is, the manufacturing and detection have good process consistency, and the online detection performed during the manufacturing process achieves good process compatibility between the manufacturing and detection.

Description

Technical Field

[0001] The present invention belongs to the technical field of additive manufacturing of components, and more specifically, relates to a method and device for on-line detection of internal three-dimensional defects during additive manufacturing of components. Background Art

[0002] The following three technologies cannot meet the requirements for on-line detection of three-dimensional defects in the process of additive manufacturing of components.

[0003] (1) Surface detection devices such as microscopes and scanning electron microscopes mainly complete surface detection and it is difficult to detect internal defects of components.

[0004] (2) Traditionally, some industrial products often use methods such as cutting for destructive measurement, which destroys the original morphology of the components and has very low efficiency.

[0005] (3) Industrial CT scanning and ultrasonic detection can obtain internal information of workpieces for defect detection, but can only be carried out after the parts are manufactured, which belongs to off-line detection and cannot achieve on-line detection. It is difficult to integrate manufacturing and detection, affecting production efficiency.

[0006] Currently, traditional destructive measurement methods not only damage the workpiece structure, but also easily cause the workpiece to deform due to destructive testing, affecting the final measurement result; and during 3D printing, due to the special nature of workpiece forming, existing methods need to wait until the workpiece is printed before detection, consuming a large amount of time and causing waste of printing materials. Summary of the Invention

[0007] The purpose of the present invention is to provide a method and device for on-line detection of internal three-dimensional defects during additive manufacturing of components, (describing the beneficial effects and the advantages compared with the current technology).

[0008] The technical solution adopted by the present invention is specifically as follows: A method for on-line detection of internal three-dimensional defects during additive manufacturing of components, including the following steps:

[0009] S1. Perform component printing;

[0010] S2. Scan the printed layer printed by the print head to obtain the point cloud data of each printed layer;

[0011] S3. Remove invalid points from the obtained layer point cloud data, perform PCD point cloud format conversion, and sample for point cloud data preprocessing to obtain single-layer point cloud data;

[0012] S4. Detect and perform quantitative calculation on the defect area of the single-layer point cloud data, and construct a parameter matrix with defect features;

[0013] S5, calculating the average height of the point cloud layer of each single layer of the component, and performing transition processing of the point cloud data between layers, thereby constructing a three-dimensional body reconstruction;

[0014] S6. After calculating the center of gravity offset of each cube, the defect fusion is performed according to the parameter matrix to obtain the specific position and size of the defect inside the three-dimensional component.

[0015] Optionally, the component printing adopts 3D printing additive equipment to achieve corresponding printing, and the component content defects during the component 3D printing process are detected through online real-time three-dimensional detection. If defects are found, the 3D printing process parameters are corrected online to suppress the formation of internal defects in the component 3D printing process.

[0016] Optionally, the scanning print head is realized by a robotic arm carrying a line laser scanner that follows the moving print head. The robotic arm carries the print head and the laser scanner and moves simultaneously. The print head and the laser scanner adopt a parallel linkage layout, and the print head and the laser scanner are controlled by a host computer and software to achieve the integrated process characteristics of printing, scanning, measuring and repairing.

[0017] Optionally, according to the structural characteristics of the component and the printing process parameters, through filtering, median filtering or uniform filtering is selected within the layer to perform integrated processing of the sampled point cloud data, and the minimum vector spacing of the point cloud data is used between layers to make an interpolation transition to achieve a smooth transition of the reconstruction area between component layers.

[0018] Optionally, the division of the defect area can further shorten the point cloud processing range and clarify the basic characteristics of the range, shape and nature of single-layer defects.

[0019] Optionally, the component realizes high superposition of each layer of point cloud to form a three-dimensional reconstruction through self-stacking of point cloud.

[0020] Optionally, the calculation of the center of gravity offset of each cube is achieved by introducing the minimum enclosing cube of each layer defect.

[0021] Optionally, basic information about the internal defects of the component is fed back to the additive layer slicing software to correct process parameters and perform single-layer or multi-layer printing again.

[0022] A device for online detection of internal three-dimensional defects during additive manufacturing of a component, wherein the device for online detection of internal three-dimensional defects during additive manufacturing of a component executes any one of the aforementioned methods for online detection of internal three-dimensional defects during additive manufacturing of a component.

[0023] The technical effects achieved by the present invention are:

[0024] (1) This solution makes full use of the basic principle of the additive manufacturing process - the layer stacking principle, that is, the additive manufacturing process of components is a layer-by-layer manufacturing process, and the detection of internal defects can exactly utilize this basic principle process of the additive manufacturing process. Printing and detection are integrated together, and the detection is fed back to the printing process defects in real time, and the printing process is required to be corrected in real time and then detected again, achieving the process of printing - scanning - detecting - repairing, that is, manufacturing and detection have good process consistency, and online detection during the manufacturing process achieves good process compatibility between manufacturing and detection.

[0025] (2) This solution is for the detection of part defects during the manufacturing process, including internal and external defects, without damaging the parts, and can also reflect the synchronization of manufacturing and detection, realizing online manufacturing and detection in the true sense, and constructing an integrated platform for manufacturing and detection.

[0026] (3) This solution can realize the detection of errors caused by the manufacturing process, feed back and correct them in real time, and achieve the synchronization of manufacturing and detection.

[0027] (4) This solution is based on each printing layer and does not require destructive testing of workpieces like traditional destructive testing.

[0028] (5) Ultrasonic testing technology can only detect metal materials, requires a certain surface finish of the detection surface, and needs a coupling agent to ensure sufficient acoustic coupling, while this technology has lower requirements for the surface flatness of the printing layer and the selection of materials.

[0029] (6) Industrial CT scanning is mostly off-line and off-site detection. While obtaining high precision, it also requires extremely high detection costs and maintenance costs. This technology is for online detection, without the need for off-line and off-site detection, not only has relatively low costs, but also is simple and convenient to maintain, and the accuracy can meet the requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0031] Figure 1 It is a system structure block diagram of an online detection method for internal three-dimensional defects during additive manufacturing of components according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0032] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0033] Please refer to Figure 1 , some embodiments of the present invention provide an on-line detection method for internal three-dimensional defects during component additive manufacturing, including the following steps:

[0034] S1. Use a 3D printing additive device to print components.

[0035] The 3D printing additive device uses a 3D printer.

[0036] S2. Use a robotic arm to carry a line laser scanner to scan the printing layer following the moving print head to obtain the point cloud data of each printing layer;

[0037] In this embodiment, the line laser scanner follows the printing movement during printing. Further, during printing, the robotic arm moves the scanner to the best scanning position above the platform and makes the laser emission lens parallel to the printing plane. The format of the scanned file is CSV. The rows and columns of the file respectively correspond to the x-axis and y-axis during scanning, and the data corresponds to the height information of the point cloud. After importing the CSV file into the programmed Python program and setting the width and length point spacings (such as 0.005 mm and 0.01 mm), it is carried out in the order of rows first and columns second, and finally the conversion of the PCD file format can be completed. Here, PCD, the full name is PointCloud Data, is a file format for storing point cloud data.

[0038] S3. Use a programmed algorithm to remove invalid points from the obtained layer point cloud data and perform PCD point cloud format conversion to obtain a regular point cloud, and perform point cloud data preprocessing using straight-through filtering, median filtering, and uniform downsampling to obtain single-layer point cloud data;

[0039] Among them, the thickness of each layer of 3D printing is the same, and the platform height automatically decreases by one layer thickness every time a layer is printed. Therefore, set the straight-through filtering threshold T = d0 - w0 * 0.7, where w0 is the printing layer thickness and d0 is the height difference between the scanner and the printing platform. According to the difference in the height characteristic information of each point, remove the invalid point cloud (printing platform) and extract the object point cloud.

[0040] S4. Use the programming algorithm to detect the defective areas and perform quantitative calculations on the single-layer point cloud data, and construct a parameter matrix with defect features. Dividing the defective areas can further shorten the scope of point cloud processing and clarify the basic features such as the scope, shape, and nature of the single-layer defects.

[0041] S5. Calculate the average height of each single layer of the point cloud of the component, and perform the transition processing of the inter-layer point cloud data, and then construct the three-dimensional reconstruction;

[0042] Furthermore, due to the imbalance of height information, the range of the point cloud height information may be quite large, which affects the subsequent detection and visual effects. Therefore, define a scanning kernel, the kernel size depends on the point cloud spacing (e.g., length * width = 0.2mm * 0.1mm), scan from bottom to top and from left to right using the kernel, arrange the height information in the kernel in descending order, and change all the height values in the kernel to the median value to achieve the purpose of overall smoothing.

[0043] Among them, the point cloud simplification problem can be described as: given a point cloud model P = {p i} ∈ R 3 , the data point p i = (x i , y i , z i ), simplify it to a point cloud model Q = {q i} ∈ R 3 according to the actual situation, |Q| < |P|, and the two should be as close as possible to maintain the model features. Since the research object is a regular point cloud, which has a certain topological structure, on the premise of reducing the processing time and retaining the original morphology as much as possible, the uniform downsampling method of sampling every two points along the width direction can be used for point cloud simplification;

[0044] Compared with the original point cloud, after the preprocessing of extraction, smoothing, and simplification, irrelevant information is removed, the processing time is reduced, and the detectability is improved.

[0045] S6. Use the programming algorithm, by introducing the minimum bounding cube of each layer of defects, calculate the centroid offset of each cube, and perform defect fusion based on the parameter matrix to finally obtain the specific position and size of the defects inside the three-dimensional component.

[0046] S7. Feed back the basic information of the obtained defects to the layer slicing software for additive manufacturing, correct the process parameters, and perform single-layer or multi-layer printing again.

[0047] In some embodiments, an on-line detection device for internal three-dimensional defects during additive manufacturing of a component, the on-line detection device for internal three-dimensional defects during additive manufacturing of the component includes an on-line detection method for internal three-dimensional defects during additive manufacturing of a component according to any one of claims 1-8.

[0048] Defect area detection and quantization calculation of single-layer point cloud data, construction of a parameter matrix with defect features, and basic features of single-layer defects are as follows:

[0049] Single-layer defect recognition and quantization calculation are carried out using single-core scanning, region segmentation, and feature extraction. First, a rectangular scanning kernel is introduced, and attributes such as length, width, height, and scanning step size are set. The overall point cloud is scanned from bottom to top and from left to right, and the number of points in the kernel is calculated and recorded. Then, threshold conditions for the adaptive scanning kernel are set, and the kernel threshold is calculated according to the filling amount of the printing material and the density of the point cloud. The selection formula for the kernel threshold can be:

[0050]

[0051] where S is the filling amount and ζ is the adjustment coefficient. Then, the line where the defect is located is colored; through region segmentation, the line where the defect is located can be extracted separately, reducing the search range and further reducing the processing time. According to the characteristics that there are no points at the defect of the three-dimensional point cloud and the original serial number at the defect is removed and inherited by the next non-defect point, the contour at the defect is extracted. For the defect area, the distance between adjacent points is calculated in turn from left to right and from bottom to top. Set the distance threshold:

[0052]

[0053] Adjacent two points greater than the threshold d are marked as defect boundary points and colored black; when performing region segmentation, the normal region and multiple defect regions need to be segmented. The multiple defect regions are not connected to each other, and when performing the nth segmentation, the number of missing points in the first n - 1 defect regions needs to be calculated. The missing point calculation formula is:

[0054] Q i =(row*col)-Num#(3)

[0055] Since the distance between points is set relatively small, the defect boundary points can be approximately connected by straight line segments to replace the contour; the defect boundary length L is obtained by calculating the distance between adjacent defect points.

[0056]

[0057] Finally, the area of the defect region is calculated. The specific steps are as follows:

[0058] First, perform a projection transformation on the point cloud, and set different parameters of the projection matrix according to needs.

[0059]

[0060] where: N is the distance of the near plane, and F, B, L, and R are the up, down, left, and right vectors respectively.

[0061] The defect boundary can be approximately replaced by a polygon formed by connecting defect boundary points in sequence. The defect area formed between adjacent point cloud rows can be approximated as the area of a trapezoid. Based on the idea of differential superposition, the area of the polygon defect can be transformed into the sum of the areas of multiple trapezoids. Therefore, for the quantitative calculation of the area of each hole defect, it is transformed into solving the sum of the areas of multiple trapezoids:

[0062]

[0063] Position and superimpose multi-layer point clouds through initializing the coordinate origin and translation transformation. The translation matrix can be represented by the distances T x , T y , T z by which the coordinate origin moves around the x, y, and z axes respectively:

[0064]

[0065] According to the features obtained from the defect detection and quantitative calculation of single-layer point cloud data, align the lengths and widths of each layer of point clouds by initializing the coordinate origin. Use the correlation relationships between the feature parameters of each layer (such as: the difference in the central coordinate X between adjacent layers for the same defect is less than or equal to 0.01 mm, and the percentage difference in area does not exceed 10%) to perform translation transformation and achieve the height superposition of each layer of point clouds.

[0066] Finally, introduce the minimum bounding cube V i of the in-layer defect. After calculating the contour length, area, and depth of the single-layer defect, construct a defect feature parameter matrix with the central coordinate, defect area, and centroid of the minimum bounding cube.

[0067] The adopted programming algorithm is as follows:

[0068] From bottom to top, calculate the central coordinates (X i , Y i ) and the centroid (x i , y i , z i ) of each layer. Taking the first layer as the reference, calculate the difference in central coordinates:

[0069] (X′ (i+1)i = X i+1 - X i , Y′ (i+1)i = Y i+1 - Y i )#(6)

[0070] (X″ i = X i - X1, Y″ i = Y i - Y1)#(7)

[0071] Relative offset of the center of gravity:

[0072] (x′ (i+1)i = x i+1 - x i , y′ (i+1)i = y i+1 - yi ， z′ (i+1) = z i+1 - z i )#(8)

[0073] (x″ i = x i - x1, y″ i = y i - y1, z″ i = z i - z1)#(9)

[0074] Add offset threshold conditions, for example:

[0075]

[0076] For adjacent two layers that satisfy equation (10), defect fusion can be performed. It can be considered that the defects in the two adjacent layers are the distribution of the same defect at different heights, that is, the defects included in V i+1 and V i are the same defect.

[0077]

[0078] If equation (11) is also satisfied, then the defect can be considered as "vertical", otherwise it is determined as "tilted". For each layer of data read in, the above threshold conditions are judged, and the layer numbers of the fused point cloud are counted. The layers with the statistical layer number greater than or equal to 2 are regarded as defects. According to the numbers, the outlines of each printing layer are connected in sequence pairwise to obtain the complete outer contour of the defect, and the specific position, size, volume and surface area of the complete defect are obtained.

[0079] For example: A certain defect includes a total of 10 layers from i to i + 9.

[0080] Among them, the defect position is: (X i , Y i , Z i ) ∈ R 3 , and the defect volume is:

[0081] Feed back the defect information, and let the printing personnel judge the authenticity of the defect in the space and determine the harmfulness of the defect. If abnormal defects are found, the process can be corrected in time to improve the manufacturing accuracy of the parts and save time and costs.

[0082] Some other embodiments of the present invention provide an on-line detection device for internal three-dimensional defects during additive manufacturing of components. The on-line detection device includes a detection device and a processor. The detection device is connected to the processor. The detection device is, for example, a line laser scanner. The detection device is configured to scan the printed layer to obtain the point cloud data of each printed layer. The processor can be implemented by a central processing unit, a server, a terminal device or any other possible processing device. In some embodiments, the above-mentioned central processing unit, server, terminal device or other processing device can be implemented on a cloud platform. In some embodiments, the above-mentioned central processing unit, server or other processing device can be interconnected with various terminal devices, and the terminal device can complete information processing work or part of the information processing work. The processor is configured to execute the on-line detection method for internal three-dimensional defects during additive manufacturing of the above-mentioned components. In some embodiments, the on-line detection device further includes a printing device, for example, a 3D printer. The printing device is used for printing components.

[0083] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific embodiments described. Obviously, many modifications and variations can be made according to the content of this specification. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and utilize the present invention well. The present invention is only limited by the claims and their full scope and equivalents.

Claims

1. An on-line detection method for internal three-dimensional defects during additive manufacturing of components, characterized in that The following steps are involved: S1. Use 3D printing additive equipment to print components; S2, using a robotic arm equipped with a line laser scanner to follow the moving print head to scan the printed layer, and obtain point cloud data of each printed layer; When printing, the mobile robot arm places the scanner at the best scanning position above the platform, and makes the laser emission lens parallel to the printing plane; the file format obtained by scanning is CSV, the rows and columns of the file correspond to the x-axis and y-axis during scanning, and the data corresponds to the height information of the point cloud. The CSV file is imported into the compiled Python program, and after setting the width and length of the point spacing, it is carried out in the order of rows first and then columns, and finally the conversion of the PCD file format can be completed; S3. Use the compilation algorithm to remove invalid points in the acquired layer point cloud data and convert it into PCD point cloud format to obtain a regular point cloud. Use straight-through filtering, median filtering and uniform downsampling to preprocess the point cloud data to obtain a single layer point cloud data. The thickness of each layer of 3D printing is consistent, and the platform height is automatically reduced by one layer thickness for each layer printed; the straight-through filter threshold T=d0-w0*0.7 is set, where w0 is the printing layer thickness, and d0 is the height difference between the scanner and the printing platform. According to the difference in the height feature information of each point, the invalid point cloud is removed and the object point cloud is extracted; S4, detecting and quantifying the defective area of the single-layer point cloud data, and constructing a parameter matrix based on the defect characteristics; Among them, the division of defect areas can further shorten the point cloud processing range and clarify the scope, shape and nature of single-layer defects; S5, calculating the average height of the point cloud layer of each single layer of the component, and performing transition processing of the point cloud data between layers, thereby constructing a three-dimensional body reconstruction; S6. Using the compilation algorithm, by introducing the minimum bounding cube of each layer defect, calculating the center of gravity offset of each cube, and performing defect fusion according to the parameter matrix to obtain the specific position and size of the defect inside the three-dimensional component; S7, feeding back the basic information of the defects to the additive layer slicing software, correcting the process parameters, and performing single-layer or multi-layer printing again; In the step S4, the method of defect area detection and quantitative calculation of single-layer point cloud data and construction of parameter matrix based on defect features is as follows: single-layer defect identification and quantitative calculation are performed using single-core scanning, region segmentation, and feature extraction; First, a rectangular scanning kernel is introduced, and the length, width, height and scanning step size are set. The entire point cloud is scanned from bottom to top and from left to right, and the number of points in the kernel is calculated and recorded. Then, the threshold condition of the adaptive scanning kernel is set, and the kernel threshold is calculated according to the filling amount of the printing material and the density of the point cloud; the formula for selecting the kernel threshold is: Among them, S1 is the filling amount, ζ is the adjustment coefficient; Then, color the line where the defect is located; through region segmentation, the line where the defect is located can be extracted separately, narrowing the search range and reducing the processing time once again; according to the characteristics that there are no points at the defect of the three-dimensional point cloud and the original serial number at the defect is removed and then inherited by the next non-defect point, extract the contour at the defect; for the defect area, calculate the distance between adjacent points in sequence from left to right and from bottom to top; set a distance threshold d; where the formula for the distance threshold d is: Adjacent points greater than the threshold d are marked as defect boundary points and assigned black; when performing region segmentation, it is necessary to segment the normal region and multiple defect regions. The multiple defect regions are not connected to each other, and when performing the nth segmentation, it is necessary to calculate the missing point numbers of the previous n - 1 defect regions. The missing point number Q i The calculation formula is as follows: Q i = (row * col) - Num Since the distance between points is set to be small, a straight line segment is used to connect the defect boundary points to approximately replace the contour; the defect boundary length L1 is obtained by calculating the distance between adjacent defect points. Finally, calculate the defect area, and the specific steps are as follows: First, perform a projection transformation on the point cloud and set different parameters of the projection matrix according to needs. Where: N is the distance of the near plane, and F, B, L, and R are the up, down, left, and right vectors respectively. The defect boundary is approximately replaced by a polygon formed by connecting the defect boundary points in sequence. The defect area formed between adjacent point cloud lines is approximately the area of a trapezoid. Then, the area of the polygon defect is transformed into the sum of the areas of multiple trapezoids based on the idea of differential superposition; therefore, for the quantitative calculation of the area of each hole defect, it is transformed into solving the sum of the areas of multiple trapezoids, and the calculation formula for the sum of the areas of multiple trapezoids is: Position and superimpose multi-layer point clouds through initializing the coordinate origin and translation transformation; the translation matrix can be represented by the distances Tx, Ty, and Tz that the coordinate origin moves around the x, y, and z axes respectively. x , Ty y , Tz z to represent; According to the characteristics obtained from the defect detection and quantitative calculation of the single-layer point cloud data, align the lengths and widths of each layer of point cloud by initializing the coordinate origin; use the correlation relationship between the characteristic parameters of each layer to perform a translation transformation to achieve the height superposition of each layer of point cloud. Finally, introduce the minimum bounding cube within the layer. After calculating the contour length, area, and depth of the single-layer defect, construct a defect characteristic parameter matrix with the center coordinates of the minimum bounding cube, the defect area, and the centroid.

2. The on-line detection method for internal three-dimensional defects during additive manufacturing of components according to claim 1, wherein: The component printing is realized by a 3D printing additive manufacturing device, and during the 3D printing process of the component, the internal defects of the component content are detected online in real time. If a defect is found, the 3D printing process parameters are corrected online to inhibit the formation of internal defects during the 3D printing process of the component.

3. An on-line detection method for internal three-dimensional defects during additive manufacturing of components according to claim 1, characterized in that: The print head is realized by a print head carried by a robotic arm following a line laser scanner. The robotic arm carries the print head and the laser scanner and moves simultaneously. The print head and the laser scan adopt a parallel linkage layout, and the print head and the laser scan are linked and controlled by a host computer and software to achieve the integrated process characteristics of printing - scanning - measuring - repairing.

4. The on-line detection method for internal three-dimensional defects during additive manufacturing of components according to claim 1, characterized in that: According to the structural characteristics of the component and the printing process parameters, choose direct filtering or median filtering or uniform method for sampling point cloud data integration processing within the layer, and use the minimum vector distance of the point cloud data to judge and perform interpolation transition between layers to achieve a smooth transition in the reconstruction area between layers of the component.

5. An on-line detection device for internal three-dimensional defects during additive manufacturing of components, characterized in that: The on - line three - dimensional defect detection device during the additive manufacturing of the component executes the method for on - line three - dimensional defect detection during the additive manufacturing of the component as described in any one of claims 1 - 4.

Images