A method for aligning special-shaped surface additive manufacturing products
Through the combination of hydraulic tooling and three-dimensional data acquisition technology, the precise positioning and correction of special-shaped surface additive manufacturing products is achieved, solving the problem of complex surface product positioning in the existing technology, and improving processing accuracy and efficiency.
Patent Information
- Application Number
- CN202411088698.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-09
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2044-08-09
AI Technical Summary
The prior art is difficult to achieve efficient data processing and precise positioning correction when processing additive manufacturing products with complex or special-shaped surfaces, and lacks automation and intelligence, resulting in inefficiency and error prone.
Using hydraulic tooling and three-dimensional data acquisition technology combined with the public point least squares method, internal signature description algorithm and direction histogram signature algorithm, the precise registration of the three-dimensional point cloud data model and theoretical model is realized by iterating the nearest point algorithm, and the position pose transformation is calculated Matrix to adjust product position and orientation.
It improves the processing accuracy and efficiency of additive manufacturing products, reduces manual positioning errors, reduces the risk of deviation in the production process, and significantly improves production efficiency and product quality.
Smart Images

Figure CN118578670B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of mechanical processing, and in particular to a method for aligning a product manufactured by additive manufacturing of a special-shaped surface. Background Art
[0002] In modern manufacturing, especially in the field of additive manufacturing (commonly known as 3D printing), it is critical to accurately position and correct products to their theoretical design models. This process usually involves the use of various techniques to ensure the geometric and dimensional consistency of the manufactured physical product with its computer-generated design model. Traditional methods commonly used techniques include the use of mechanical tooling for preliminary positioning of the product and the use of three-dimensional data acquisition technologies such as laser scanning to capture the actual shape of the product. This data is then processed by software tools to evaluate the difference between the actual state of the product and its theoretical model.
[0003] However, although commonly used alignment methods can achieve the positioning and correction of regular additive manufacturing products, these methods often encounter difficulties and are inefficient when processing products with complex or special-shaped surfaces. Existing technologies often cannot achieve efficient data processing and accurate alignment, especially when the product has complex geometric structures or subtle design details. In addition, traditional methods lack sufficient automation and intelligence, resulting in inefficient operations and prone to errors.
[0004] Therefore, it is very necessary to develop a new alignment method for additively manufactured products with special-shaped surfaces. Summary of the invention
[0005] The present application provides a method for aligning a product with an additively manufactured surface to improve the processing accuracy and efficiency of the additively manufactured product.
[0006] The present application provides a method for aligning a product manufactured using additively manufactured products having special-shaped surfaces, including:
[0007] Use hydraulic tooling, contour blocks and ball studs to perform preliminary positioning of additively manufactured products;
[0008] Using 3D data acquisition technology, obtain the 3D point cloud data model of the surface of the additive manufacturing product after preliminary positioning;
[0009] By using the common point least squares method coordinate transformation principle, the best fit between the coordinate system of the three-dimensional point cloud data model and the coordinate system of the theoretical model is achieved to obtain the fitted three-dimensional point cloud data model; wherein the theoretical model is constructed according to the design parameters of the additive manufacturing product by computer-aided design software;
[0010] The key points of the point cloud data in the three-dimensional point cloud data model after fitting are extracted by using an internal signature description algorithm; the key points are extracted by using a directional histogram signature algorithm to obtain key point features; based on the key point features, accurate registration of the point cloud data is achieved by using an iterative closest point algorithm to obtain a pose transformation matrix between the point cloud data model after accurate registration and the theoretical model;
[0011] The distance and angle that need to be adjusted for the additively manufactured product are calculated according to the posture transformation matrix, and the additively manufactured product is aligned according to the calculated distance and angle.
[0012] Furthermore, the method of achieving the best fit between the coordinate system of the three-dimensional point cloud data model and the coordinate system of the theoretical model by using the common point least squares method coordinate transformation principle to obtain the fitted three-dimensional point cloud data model includes:
[0013] For the data points in the coordinate system of the theoretical model And the data points in the coordinate system of the 3D point cloud data model , the weight is calculated according to the following formula 1 :
[0014]
[0015] in, Yes The measurement uncertainty of is a positive constant; the point and Point correspond to each other;
[0016] According to the following formula 2, the total deviation is defined :
[0017]
[0018] in, is the rotation matrix; is the translation vector; is the total number of data points in the coordinate system of the theoretical model;
[0019] The Lagrangian function is constructed according to the following formula 3 :
[0020]
[0021] in, is the symmetric Lagrange multiplier matrix; represents the trace of a matrix; is the rotation matrix The transpose of is the identity matrix;
[0022] Solving the Lagrangian function using the Newton–Raphson iteration method , find Minimum target rotation matrix and the target translation variable ;
[0023] According to the target rotation matrix and the target translation variable , obtain the fitted three-dimensional point cloud data model.
[0024] Furthermore, the extracting key points of the point cloud data in the fitted three-dimensional point cloud data model by using the internal signature description algorithm includes:
[0025] Selecting a target point in the point cloud data in the fitted three-dimensional point cloud data model as an analysis center point;
[0026] A local neighborhood is defined, wherein a specified number of neighboring points are included around the analysis center point in the local neighborhood;
[0027] For each point in the analysis center point and the local neighborhood, the geometric properties between it and the neighboring points are calculated; wherein the geometric properties include relative position and normal direction difference;
[0028] Based on the geometric attributes, a signature matrix describing the local geometric shape of the analysis center point is generated, wherein the signature matrix includes the distances and angles between points and their statistical distribution characteristics;
[0029] The obtained signature matrix is compared with the preset threshold to determine whether the center point is a key point.
[0030] Furthermore, the use of a directional histogram signature algorithm to extract features of key points in the fitted three-dimensional point cloud data model to obtain key point features includes:
[0031] Step a: select a key point as the center point, determine a predefined neighborhood radius, and select a specified number of neighboring points around the center point with the neighborhood radius;
[0032] Step b: for each neighboring point, calculate the vector between it and the center point, and standardize the direction of the vector relative to the surface normal of the center point to eliminate the deviation caused by different acquisition angles;
[0033] Step c: quantizing the direction of the normalized vector into a predetermined number of histogram boxes, and calculating the number of vectors falling in each histogram box to form a direction histogram;
[0034] Repeat steps ac to generate a direction histogram for each key point, and use the direction histogram as the key point feature.
[0035] Furthermore, the accurate registration of the point cloud data is achieved by iterative closest point algorithm according to the key point features, and the pose transformation matrix between the point cloud data model and the theoretical model after accurate registration is obtained, including:
[0036] Using the extracted key point features as a source point set, and performing feature matching between the source point set and the corresponding target point set in the theoretical model to form an initial point pair relationship;
[0037] The following iterative registration steps are performed:
[0038] Step a1: Based on the key point features, for each point in the source point set, find the nearest point with the most similar features in the target point set, and redetermine the point pair relationship;
[0039] Step a2: For each set of redetermined point pairs, calculate a transformation including rotation and translation so that after applying this transformation, the source point set is more closely aligned with the target point set;
[0040] Step a3: Apply the calculated transformation to the source point set so that the source point set is better aligned with the target point set;
[0041] Check whether the change between consecutive iterations is less than a preset threshold. If so, it is considered that the registration process has converged and the iteration is stopped; if not, continue the iteration steps of steps a1-a3;
[0042] After the iteration is completed, the rotation matrix and translation vector that achieve accurate registration are recorded, and the pose transformation matrix from the source point set to the target point set is constructed based on the accurate registration rotation matrix and translation vector.
[0043] Furthermore, the calculating the distance and angle that the additive manufacturing product needs to be adjusted according to the posture transformation matrix, and aligning the additive manufacturing product according to the calculated distance and angle, includes:
[0044] Parsing the pose transformation matrix obtained from the iterative closest point algorithm ,in is a The rotation matrix of is a three-dimensional translation vector;
[0045] Calculate the distance to be adjusted according to the following formula 4 :
[0046]
[0047] in is the translation vector The amount of
[0048] Calculate the Euler angle according to the following formula 5-7 , the Euler angle Indicates the angles that the additive manufacturing product needs to rotate around the X-axis, Y-axis, and Z-axis:
[0049]
[0050]
[0051]
[0052] in, Represents the rotation matrix Middle Line Elements of a column;
[0053] According to the calculated distance and Euler angles , adjust the position and orientation of the additively manufactured product.
[0054] Furthermore, the preliminary positioning of the additively manufactured product using the hydraulic tooling profiling block and the ball stud includes:
[0055] In the design and manufacturing process of the additive manufacturing product, a coarse reference block is formed or integrated at a key part of the additive manufacturing product, and the coarse reference block clearly appears on the surface of the additive manufacturing product as a physical feature to support the subsequent clamping and positioning process;
[0056] Placing a hydraulic tooling profiling block to fit the coarse reference block, and ensuring accurate docking between the hydraulic tooling profiling block and the coarse reference block through mechanical adjustment, wherein the shape and size of the hydraulic tooling profiling block are designed to fully match the coarse reference block;
[0057] The hydraulic tooling copy block is mechanically fixed to the coarse reference block using a ball stud, ensuring that the additively manufactured part remains stable and in the correct position throughout the machining or measurement process.
[0058] Furthermore, the three-dimensional data acquisition technology is used to obtain a three-dimensional point cloud data model of the surface of the additively manufactured product after preliminary positioning, including:
[0059] Use a high-precision laser scanner to scan additively manufactured products from multiple angles to obtain complete surface point cloud data;
[0060] Preprocessing the acquired point cloud data, wherein the preprocessing includes denoising, filtering and data compression;
[0061] According to the preprocessed point cloud data, a three-dimensional point cloud data model of the surface of the additive manufacturing product after preliminary positioning is obtained.
[0062] Furthermore, the method of using the three-dimensional data acquisition technology to obtain the three-dimensional point cloud data model of the surface of the additively manufactured product after preliminary positioning also includes:
[0063] When using high-precision laser scanners to scan additively manufactured products from multiple angles, the scanning speed and resolution of the 3D scanning equipment are optimized according to the material characteristics and geometric complexity of the additively manufactured products.
[0064] Furthermore, the alignment method for additively manufactured products with special-shaped surfaces further includes:
[0065] The hydraulic pressure holding control mechanism is used to ensure that the additive manufacturing product is in a consistent state during the processing.
[0066] The beneficial effects of the technical solution provided by this application include:
[0067] (1) The use of hydraulic tooling profiling blocks and ball studs to perform preliminary positioning of additively manufactured products can ensure the stability and correct position of the products before processing, which is particularly important for complex or special-shaped products. This method reduces the error of manual positioning and improves the accuracy and repeatability of the processing process. (2) By utilizing advanced 3D data acquisition technologies such as laser scanning, the surface data of additively manufactured products can be quickly and accurately captured. This makes it possible to obtain a high-resolution 3D point cloud data model, providing detailed basic data for subsequent data processing and analysis. (3) The principle of common point least squares coordinate transformation is adopted to achieve the best fit between the 3D point cloud data model and the theoretical model of computer-aided design. This method can effectively reduce the difference between the actual product and the design model, thereby optimizing the entire manufacturing process. (4) The key points in the point cloud data are efficiently extracted using the internal signature description algorithm and the directional histogram signature algorithm, and the registration is accurately performed through the iterative closest point algorithm, further improving the accuracy and efficiency of the registration. The application of these algorithms makes the posture adjustment more accurate and reduces the risk of deviation in the production process. (5) By accurately calculating the required adjustment distance and angle and automatically aligning the additive manufacturing product according to these parameters, production efficiency and product quality are significantly improved. The automated adjustment process reduces manual intervention and reduces production costs and time. BRIEF DESCRIPTION OF THE DRAWINGS
[0068] Figure 1 This is a flow chart of a method for aligning a product made of additively manufactured products with special-shaped surfaces provided in the first embodiment of the present application. DETAILED DESCRIPTION
[0069] Many specific details are described in the following description to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of the present application, so the present application is not limited by the specific implementation disclosed below.
[0070] The first embodiment of the present application provides a method for aligning a product manufactured by additive manufacturing of a special-shaped surface. Figure 1 , which is a schematic diagram of the first embodiment of the present application. Figure 1 A first embodiment of the present application provides a method for aligning a product made of additively manufactured products with special-shaped surfaces, which is described in detail.
[0071] Step S101: Preliminary positioning of the additively manufactured product is performed using a hydraulic tooling profiling block and a ball stud.
[0072] In step S101, the detailed operation process involves the initial positioning of the special-shaped surface additive manufacturing product using the hydraulic tooling profiling block and the ball stud. The purpose of this step is to provide a stable and accurate initial position reference for subsequent precision processing and testing. In order to achieve this goal, the following detailed operation process needs to be followed:
[0073] First, choose the right hydraulic tooling blocks. These blocks must be designed to match the shape and size of the additively manufactured product to ensure that they can provide adequate support and stability. Tooling blocks are usually designed with adjustable contact surfaces to adapt to the specific shape of the product, thereby reducing pressure and potential damage to the product surface.
[0074] Secondly, the hydraulic tooling blocks are placed in predetermined positions, which are usually determined during the product design phase, and joints or grooves are reserved on the surface of the product. The design of the joints or grooves should ensure that the tooling blocks can be accurately positioned and firmly fixed on the product.
[0075] Next, the hydraulic tooling copy block is fixed to the AM product using a ball stud. The ball stud design allows for fine adjustment in multiple directions to accommodate minor unevenness or angular deviations in the product surface. Before fixing the ball stud, make sure the contact surface between the tooling block and the product is flat and free of any foreign matter or impurities that would prevent a tight fit.
[0076] After securing the tooling blocks and ball studs, make any necessary adjustments to ensure the correct position of the tooling. This may include adjusting the pressure in the hydraulic system to ensure that the tooling blocks provide firm support for the product without placing excessive pressure on the product, which could cause deformation or damage.
[0077] After completing the initial positioning, check all connection points to ensure that each component has been correctly installed and fixed to ensure the position stability and positioning accuracy of the product in subsequent steps.
[0078] The above process ensures that the additive manufacturing product can be firmly and accurately positioned before entering precision processing or testing, thus laying the foundation for obtaining high-quality final products. The successful execution of this step is crucial to the efficiency of the entire production process and product quality.
[0079] Furthermore, the preliminary positioning of the additively manufactured product using the hydraulic tooling profiling block and the ball stud includes:
[0080] In the design and manufacturing process of the additive manufacturing product, a coarse reference block is formed or integrated at a key part of the additive manufacturing product, and the coarse reference block clearly appears on the surface of the additive manufacturing product as a physical feature to support the subsequent clamping and positioning process;
[0081] Placing a hydraulic tooling profiling block to fit the coarse reference block, and ensuring accurate docking between the hydraulic tooling profiling block and the coarse reference block through mechanical adjustment, wherein the shape and size of the hydraulic tooling profiling block are designed to fully match the coarse reference block;
[0082] The hydraulic tooling copy block is mechanically fixed to the coarse reference block using a ball stud, ensuring that the additively manufactured part remains stable and in the correct position throughout the machining or measurement process.
[0083] In the alignment method for additively manufactured products with special-shaped surfaces, the use of hydraulic tooling profiling blocks and ball studs for initial positioning is a key step. This step ensures that the additively manufactured product can maintain the correct posture and position during subsequent processing or measurement, thereby improving processing accuracy and efficiency. The following is a detailed description of how to achieve this process.
[0084] First, considering the specific needs of additive manufacturing products, the design process requires the formation or integration of so-called coarse datum blocks at key locations of the product. Coarse datum blocks are part of the product design and usually appear explicitly on the surface of the product as physical features. The purpose of these datum blocks is to provide support for subsequent clamping and positioning processes, ensuring that the product is accurately positioned throughout the manufacturing and inspection process.
[0085] Next, hydraulic tooling profiling blocks are designed and placed according to the shape and size of the rough reference block. These tooling blocks are designed to fully match the rough reference block to ensure that they can be accurately adapted and docked to the rough reference block in actual operation. The hydraulic tooling profiling blocks are precisely mechanically adjusted to ensure that the contact with the rough reference block is accurate and stable.
[0086] In addition, the hydraulic tooling copy block is mechanically fixed to the coarse reference block using a ball stud. The design of the ball stud allows fine adjustment in multiple directions, which is critical to ensure the precise docking between the tooling block and the coarse reference block. When fixing, the bolt needs to be carefully adjusted to ensure that the tooling block does not move during the processing or measurement process, thereby maintaining the stability and correct posture of the additively manufactured product.
[0087] Through the above steps, it is possible to effectively ensure that the additive manufacturing products have a high degree of positional accuracy and stability in the preparation stage before processing. This preliminary positioning method not only simplifies subsequent processing and measurement operations, but also significantly improves processing efficiency and product quality. The successful implementation of this process is crucial to achieving high-quality additive manufacturing, especially when it comes to the manufacturing of products with complex geometries and high precision requirements.
[0088] Step S102: using three-dimensional data acquisition technology to obtain a three-dimensional point cloud data model of the surface of the additively manufactured product after preliminary positioning.
[0089] In step S102, a three-dimensional data acquisition technology is used to obtain a three-dimensional point cloud data model of the surface of the additively manufactured product after preliminary positioning.
[0090] First, choose a suitable 3D scanning device that can efficiently capture the details of the surface of the additive manufacturing product. When choosing a device, consider the scanning accuracy, scanning speed, and compatibility with the additive manufacturing product materials. Laser scanners or structured light scanners are generally recommended because they can provide high resolution and accurate measurement results.
[0091] Before you start scanning, you need to calibrate the 3D scanning device to ensure the accuracy of the data. The calibration process usually includes checking the focus of the device, the scanning range setting, and the sensor sensitivity. In addition, in order to obtain a more complete data model, it may be necessary to scan the product from multiple angles. Therefore, set the scanning path and scanning angle to cover all surfaces of the product, especially those areas that are difficult to observe.
[0092] When performing 3D scanning, ensure that the scanning environment is stable and avoid any vibration or sudden movement that may cause distortion of the scanned data. The additively manufactured product should be fixed on a vibration-free platform and the surrounding light conditions should be consistent to avoid data errors caused by light changes.
[0093] Once the scan is complete, the next step is to use specialized software to process the raw point cloud data obtained from the scan. Data processing includes noise removal, data point filtering, and necessary point cloud density adjustment. Through these processing steps, the data model can be optimized to make it more suitable for subsequent alignment and analysis work.
[0094] Finally, the processed point cloud data is saved in a format suitable for subsequent processing, such as STL or PLY files. Ensuring the integrity and accuracy of the data is the basis for effective analysis and alignment operations.
[0095] Through the above steps, a detailed three-dimensional point cloud data model of the surface of the additive manufacturing product can be effectively obtained, providing an important data basis for accurate product alignment and quality control.
[0096] Furthermore, the three-dimensional data acquisition technology is used to obtain a three-dimensional point cloud data model of the surface of the additively manufactured product after preliminary positioning, including:
[0097] Use a high-precision laser scanner to scan additively manufactured products from multiple angles to obtain complete surface point cloud data;
[0098] Preprocessing the acquired point cloud data, wherein the preprocessing includes denoising, filtering and data compression;
[0099] According to the preprocessed point cloud data, a three-dimensional point cloud data model of the surface of the additive manufacturing product after preliminary positioning is obtained.
[0100] In the alignment method for additively manufactured products with special-shaped surfaces, using 3D data acquisition technology to obtain a 3D point cloud data model of the surface of the additively manufactured product after preliminary positioning is a crucial step. This process ensures that the resulting data model can accurately reflect the physical form of the product, thereby providing accurate input for subsequent processing steps. The following is a detailed description of how to implement this process.
[0101] First, scanning the additively manufactured product with a high-precision laser scanner is a preliminary step in obtaining 3D point cloud data. The laser scanner determines the position of each point on the surface of an object by emitting a laser beam and measuring the time or phase change between the laser beam and the object's surface. In order to ensure that comprehensive 3D data of the product is obtained, it is necessary to scan from multiple different angles so that all external surfaces of the product can be captured, especially those parts that may not be scanned at a single angle due to their position.
[0102] After scanning, the next step is to preprocess the resulting point cloud data. The purpose of this stage is to improve data quality and ensure the accuracy of subsequent processing. The preprocessing process usually includes several key operations:
[0103] 1. Denoising: Point cloud data often contains noise, which may come from the scanning device itself, environmental factors or the characteristics of the reflective surface. The denoising process involves identifying and removing these unrealistic data points to ensure that the point cloud data is cleaner and truly reflects the physical world.
[0104] 2. Filtering: This step further cleans the point cloud data and smoothes out the errors in the data. Filtering can use various algorithms, such as Gaussian filtering, median filtering, etc. These methods can reduce the random variation between data points and improve the consistency of the data.
[0105] 3. Data Compression: Point cloud data is usually very large, and directly processing this data can be very time-consuming. Data compression aims to reduce the size of the data set without significant loss of information, which can increase processing speed and simplify data storage and transmission.
[0106] After these preprocessing steps, the final result is a clean, high-quality 3D point cloud data model, which will accurately reflect the geometric shape and features of the surface of the additively manufactured product, and provide the necessary foundation for the next steps in the alignment method, such as point cloud registration and feature extraction.
[0107] Through the above detailed steps, high-precision 3D point cloud data can be effectively extracted from the actual additive manufacturing product. This data is the basis for accurate alignment, which is essential for high-quality product manufacturing and ensures the accuracy and consistency of the final product.
[0108] Furthermore, the method of using the three-dimensional data acquisition technology to obtain the three-dimensional point cloud data model of the surface of the additively manufactured product after preliminary positioning also includes:
[0109] When using high-precision laser scanners to scan additively manufactured products from multiple angles, the scanning speed and resolution of the 3D scanning equipment are optimized according to the material characteristics and geometric complexity of the additively manufactured products.
[0110] In the alignment method for additive manufacturing products with special-shaped surfaces, it is crucial to obtain a 3D point cloud data model of the surface of the additive manufacturing product by using 3D data acquisition technology. To further improve the efficiency and accuracy of this process, it also involves optimizing the configuration of the scanning equipment according to the specific properties of the additive manufacturing product, which includes adjusting the scanning speed and resolution to adapt to the material characteristics and geometric complexity of the product. The following is a detailed description of how to achieve this process.
[0111] First, before starting the scan, technicians need to evaluate the specific characteristics of the additively manufactured product. This includes the complexity of the material and geometry. For example, metal materials with high reflectivity and products with complex concave and convex surfaces will make different requirements for the configuration of the scanning equipment. Highly reflective materials may require reducing the laser power or adjusting the sensitivity of the receiving sensor to avoid data loss or misreading due to excessive reflection. At the same time, for products with complex geometries, increasing the resolution of the scan will help capture details more accurately, but this may reduce the speed of the scan.
[0112] Next, based on these assessment results, adjust the settings of the 3D scanning equipment. The optimization configuration of scanning speed and resolution aims to find the best balance between scanning efficiency and data quality. For example, for small and complex components with rich details, it may be necessary to use a slow scan with a higher resolution to ensure that all details are accurately recorded. On the contrary, for large but relatively simple objects, a faster scanning speed and lower resolution can be used to improve efficiency.
[0113] During the scanning process, it is also important to continuously monitor the data quality and the performance of the scanning equipment. If the preliminary data shows problems, such as missing data or large errors, the equipment parameters should be adjusted in time, such as changing the scanning angle, adjusting the focal length, or fine-tuning the scanning speed and resolution again. This flexible adjustment strategy can ensure that the final point cloud data model not only reflects the true form of the product, but also meets the accuracy requirements of subsequent processing.
[0114] Through the steps detailed above, the 3D scanning process can be effectively adjusted and optimized according to the specific characteristics of the additive manufacturing product. This method improves the accuracy of data acquisition, optimizes processing time, and ensures the efficiency and adaptability of the data acquisition process. Ultimately, this provides high-quality data support for the alignment process, which is a key link in achieving precise additive manufacturing.
[0115] Step S103: By using the common point least squares coordinate transformation principle, the best fit between the coordinate system of the three-dimensional point cloud data model and the coordinate system of the theoretical model is achieved to obtain the fitted three-dimensional point cloud data model; wherein the theoretical model is constructed by computer-aided design software according to the design parameters of the additive manufacturing product.
[0116] Step S103 is the core step in the entire alignment method, which involves using the common point least squares coordinate transformation principle to achieve accurate fitting between the 3D point cloud data model and the theoretical model of computer-aided design. The purpose of this step is to ensure that the actual geometry of the additively manufactured product is perfectly aligned with the design expectations, thereby improving the quality and accuracy of the final product.
[0117] First, select appropriate common points from the 3D point cloud data model obtained in the previous step. These common points are the bridge between the actual scan data and the theoretical model generated by the computer. When selecting these points, usually those with obvious geometric features that are easy to identify in both models are selected, such as corners, edges or other significant geometric markers.
[0118] Next, these common points are paired with corresponding points in the theoretical model. This process requires precise identification and matching of identical geometric feature points in both models, which often involves complex data processing and analysis techniques to ensure that each pair of points is matched as accurately as possible.
[0119] Once the points are matched, the next step is to apply the least squares principle to calculate the best coordinate transformation parameters. This includes rotation (represented by a rotation matrix) and translation (represented by a vector), which will be used to transform (or fit) the actual point cloud data to the theoretical model. Mathematically, this involves constructing and solving an optimization problem whose goal is to minimize the total distance between each point in the actual point cloud and its paired theoretical point.
[0120] The actual calculation of the least squares method is usually performed through an iterative method, where each iteration adjusts the rotation and translation parameters to gradually reduce the total deviation. After each iteration, the current fit quality is evaluated to determine whether sufficient accuracy has been achieved or whether further iterations are needed.
[0121] After the fitting process is completed, the final 3D point cloud data model will be highly consistent with the theoretical model in space, that is, the best fit is achieved. This step ensures that the geometric shape of the additive manufacturing product accurately matches the design model, providing a solid foundation for subsequent processing and assembly work.
[0122] By implementing this step, the least squares coordinate transformation principle can be effectively applied to ensure the accurate fit between the 3D point cloud data model and the theoretical model. This process not only improves the processing accuracy of the product, but also significantly optimizes the quality control process of the entire additive manufacturing.
[0123] Furthermore, the method of achieving the best fit between the coordinate system of the three-dimensional point cloud data model and the coordinate system of the theoretical model by using the common point least squares method coordinate transformation principle to obtain the fitted three-dimensional point cloud data model includes:
[0124] For the data points in the coordinate system of the theoretical model And the data points in the coordinate system of the 3D point cloud data model , the weight is calculated according to the following formula 1 :
[0125]
[0126] in, Yes The measurement uncertainty of is a positive constant; the point and Point correspond to each other;
[0127] According to the following formula 2, the total deviation is defined :
[0128]
[0129] in, is the rotation matrix; is the translation vector; is the total number of data points in the coordinate system of the theoretical model;
[0130] The Lagrangian function is constructed according to the following formula 3 :
[0131]
[0132] in, is the symmetric Lagrange multiplier matrix; represents the trace of a matrix; is the rotation matrix The transpose of is the identity matrix;
[0133] Solving the Lagrangian function using the Newton–Raphson iteration method , find Minimum target rotation matrix and the target translation variable ;
[0134] According to the target rotation matrix and the target translation variable , obtain the fitted three-dimensional point cloud data model.
[0135] First, the method defines a way to establish corresponding point relationships between the theoretical model and the 3D point cloud data model. For each data point in the coordinate system of the theoretical model , find the corresponding point in the coordinate system of the 3D point cloud data model The weight between each pair of points Calculated by the following formula:
[0136]
[0137] in, Indicate point The measurement uncertainty is It is a normal number used to ensure that the denominator is not zero and avoid calculation anomalies.
[0138] The measurement uncertainty of the point is , which usually depends on the measurement equipment used, environmental conditions, and the location of specific points on the surface of the object. In 3D scanning or other similar data acquisition processes, methods for determining the measurement uncertainty of each point can include the following:
[0139] 1. Equipment specifications
[0140] Accuracy and resolution: Manufacturers of measurement equipment often provide specifications for the accuracy and resolution of their equipment, which directly affect the uncertainty of each measured point. For example, if a laser scanner has an accuracy of ±0.1 mm, then under ideal conditions, the positional uncertainty of the measured point will be within this range.
[0141] 2. Environmental factors
[0142] Temperature, humidity and light: These environmental factors may affect the performance of the instrument and the stability of the measured data. For example, extreme temperatures may cause the electronic components of the instrument to work unstably, thereby increasing the uncertainty of the data.
[0143] 3. Material and surface properties of the object
[0144] Reflectivity and color: The material (e.g. metal, plastic, fabric, etc.) and color of an object’s surface can affect the reflective properties of optical measurement devices (e.g. laser scanners), which in turn affects measurement accuracy. Dark or highly reflective surfaces can result in higher measurement errors.
[0145] 4. Statistical Analysis
[0146] Repeated measurements and statistical models: By taking multiple independent measurements of the same feature and applying statistical analysis techniques (such as standard deviation calculation), a more reliable estimate of measurement uncertainty can be obtained. This approach is suitable for error analysis in high-precision engineering applications.
[0147] In practice, determining the measurement uncertainty at each point usually requires considering one or more of the above factors.
[0148] Next, the total deviation is defined by the formula , which is a weighted sum of the differences between the point cloud data and the theoretical model:
[0149]
[0150] Here, is the rotation matrix, is the translation vector. These two parameters describe how to adjust the theoretical model to coincide with the point cloud data model through spatial transformation. is the total number of data points in the theoretical model.
[0151] To find the minimum total deviation The rotation matrix and translation vector of , which incorporates the orthogonality constraint of the rotation matrix into the optimization problem:
[0152]
[0153] here, is a symmetric Lagrange multiplier matrix, represents the trace of the matrix. This function takes into account the orthogonality that the rotation matrix should maintain (i.e. = I, where I is the identity matrix).
[0154] Use the Newton-Raphson iteration method to solve this Lagrangian function and find the minimum The rotation matrix and translation vectors This algorithm optimizes and , these two parameters are updated in each iteration to reduce the total deviation and maintain orthogonality.
[0155] Finally, use the obtained target rotation matrix and the target translation variable The 3D point cloud data model is updated to obtain a fitted model. In this way, the 3D point cloud data model is accurately adjusted to a position and orientation consistent with the theoretical model.
[0156] Step S104: using an internal signature description algorithm to extract key points of the point cloud data in the fitted three-dimensional point cloud data model; using a directional histogram signature algorithm to extract features of the key points to obtain key point features; based on the key point features, using an iterative closest point algorithm to achieve precise alignment of the point cloud data, and obtaining a pose transformation matrix between the point cloud data model and the theoretical model after precise alignment.
[0157] Step S104 is a key step in the additive manufacturing product alignment method, which involves using advanced point cloud data processing technology to extract key features and achieve accurate alignment through these features. This step first involves using the internal signature description algorithm to extract key points in the point cloud data, then further analyzing the features of these points through the directional histogram signature algorithm, and finally using the iterative closest point algorithm for accurate alignment.
[0158] Before starting the process, the fitted 3D point cloud data model needs to be processed to determine which points can be used as key points. An internal signature description algorithm is used to automatically identify these points, usually based on the local geometric characteristics of the points to determine whether they are key points. These characteristics may include the curvature of the point, the direction of the surface normal, or the distribution of neighboring points. In this way, the most critical points for the characterization and subsequent registration of the model can be effectively screened out from a large amount of point cloud data.
[0159] Once the key points are identified, the next step is to apply the directional histogram signature algorithm to each key point. This algorithm converts the point cloud data in the local neighborhood of each point into a set of directional histograms that describe the directional distribution of the surrounding points relative to the central point. This form of description not only captures the spatial characteristics of the point, but also has a certain invariance to rotation and scale changes, thereby improving the robustness of the feature.
[0160] After obtaining the feature description of each key point, the next task is to achieve accurate registration of the point cloud data through the iterative closest point algorithm. This algorithm aligns the actual point cloud data model with the theoretical model by gradually optimizing the calculation. Specifically, in each iteration, the algorithm will find the point in the theoretical model that is closest to each key point in the actual model based on the existing pose estimate, and then calculate an optimal transformation (including rotation and translation) to make the point in the actual model as close as possible to the corresponding point in the theoretical model. This process is repeated until the preset accuracy standard is reached or the improvement is lower than a certain threshold, at which time the registration is considered to be completed.
[0161] After completing this series of operations, a precisely aligned point cloud data model and its pose transformation matrix with the theoretical model are finally obtained. This transformation matrix is the basis for the subsequent alignment step, which provides the precise adjustment instructions required for the actual additive manufacturing product relative to its design model.
[0162] By implementing the above steps, feature extraction, feature analysis and data registration of point cloud data can be accurately performed. This not only enhances the reliability and effect of the alignment method, but also optimizes the production quality of the entire additive manufacturing product.
[0163] Furthermore, the extracting key points of the point cloud data in the fitted three-dimensional point cloud data model by using the internal signature description algorithm includes:
[0164] Selecting a target point in the point cloud data in the fitted three-dimensional point cloud data model as an analysis center point;
[0165] A local neighborhood is defined, wherein a specified number of neighboring points are included around the analysis center point in the local neighborhood;
[0166] For each point in the analysis center point and the local neighborhood, the geometric properties between it and the neighboring points are calculated; wherein the geometric properties include relative position and normal direction difference;
[0167] Based on the geometric attributes, a signature matrix describing the local geometric shape of the analysis center point is generated, wherein the signature matrix includes the distances and angles between points and their statistical distribution characteristics;
[0168] The obtained signature matrix is compared with the preset threshold to determine whether the center point is a key point.
[0169] First, select the target points from the fitted 3D point cloud data model, which are used as the center points of the analysis. The process of selecting these center points should be based on certain criteria, such as the distribution density of the points, the feature prominence, or the importance of the position in the model. The selected center points will be used as the benchmark for local geometric analysis.
[0170] Next, a local neighborhood is defined for each selected center point. This local neighborhood contains a certain number of neighboring points around the center point. These points are usually selected based on the spatial distance from the center point. A distance threshold can be set to determine which points should be included. The size and shape of the local neighborhood can be adjusted according to specific analysis needs to ensure that there is enough data to support local analysis without being disturbed by surrounding noise.
[0171] Subsequently, geometric properties are calculated for each point in the central point and its local neighborhood. These geometric properties include the relative position and normal direction difference between the point and its neighboring points. Normals are usually estimated by surface fitting in the point cloud data, while relative positions are directly calculated by the coordinate difference between the points. These geometric properties reflect the detailed characteristics of the local shape in the point cloud and are the basis for subsequent analysis.
[0172] Based on the calculated geometric properties, a signature matrix describing the local geometry of each center point is generated. The signature matrix is a mathematical expression that comprehensively reflects the distances, angles and statistical distribution characteristics between points. By quantifying local spatial relationships into numerical form, the signature matrix provides a powerful tool for further pattern recognition and data analysis.
[0173] Finally, the signature matrix of each center point is compared with a preset threshold to determine whether the point meets the criteria for being a key point. The selection criteria for key points can be based on specific features in the signature matrix, such as the saliency, complexity, or distinguishability of a specific shape from other points. This step ensures that only those points that are most informative for the model description are selected as key points and then used in subsequent alignment or other analysis processes.
[0174] The following is a sample code for extracting key points using the internal signature description algorithm. The code uses Python and uses common point cloud processing libraries such as `open3d` to process data.
[0175] import numpy as np
[0176] import open3d as o3d
[0177] def calculate_normals(point_cloud):
[0178] """
[0179] Compute the normal for each point in the point cloud.
[0180] """
[0181] radius = 0.1 # Defines the neighborhood radius considered when calculating the normal
[0182] point_cloud.estimate_normals(search_param=o3d.geometry.KDTreeSearchParamHybrid(radius=radius, max_nn=30))
[0183] return point_cloud
[0184] def find_key_points(point_cloud, radius=0.2, min_neighbors=20,threshold=0.5):
[0185] """
[0186] Identify key points based on an internal signature description algorithm.
[0187] """
[0188] key_points = []
[0189] kd_tree = o3d.geometry.KDTreeFlann(point_cloud)
[0190] # Traverse each point in the point cloud
[0191] for i in range(len(point_cloud.points)):
[0192] center_point = np.array(point_cloud.points[i])
[0193] center_normal = np.array(point_cloud.normals[i])
[0194] # Get points in the local neighborhood
[0195] [k, idx, _] = kd_tree.search_radius_vector_3d(center_point,radius)
[0196] if k < min_neighbors:
[0197] continue # If the number of neighboring points is too small, skip
[0198] signature_matrix = []
[0199] # Calculate the geometric properties of each point in the local neighborhood
[0200] for j in idx[1:]: # skip itself
[0201] neighbor_point = np.array(point_cloud.points[j])
[0202] neighbor_normal = np.array(point_cloud.normals[j])
[0203] # Geometric properties: distance between points
[0204] distance = np.linalg.norm(center_point - neighbor_point)
[0205] # Geometry properties: Normal direction difference
[0206] normal_diff = np.dot(center_normal, neighbor_normal)
[0207] # Store each element of the signature matrix
[0208] signature_matrix.append([distance, normal_diff])
[0209] # Calculate statistical features based on the signature matrix
[0210] signature_matrix = np.array(signature_matrix)
[0211] if signature_matrix.size == 0:
[0212] continue
[0213] # Signature features, such as average distance and average normal direction difference
[0214] avg_distance = np.mean(signature_matrix[:, 0])
[0215] avg_normal_diff = np.mean(signature_matrix[:, 1])
[0216] # Determine whether the center point is a key point
[0217] if avg_distance > threshold and avg_normal_diff > threshold:
[0218] key_points.append(i) # If the difference between the average distance and the average normal direction exceeds the threshold, it is considered a key point
[0219] return key_points
[0220] # Load point cloud data
[0221] pcd = o3d.io.read_point_cloud("path_to_your_point_cloud.ply")
[0222] # Calculate the normal of the point cloud
[0223] pcd = calculate_normals(pcd)
[0224] # Find key points
[0225] key_point_indices = find_key_points(pcd, radius=0.2, min_neighbors=20, threshold=0.5)
[0226] # Visualize key points
[0227] key_point_cloud = pcd.select_by_index(key_point_indices)
[0228] key_point_cloud.paint_uniform_color([1, 0, 0]) # Mark the key points in red
[0229] o3d.visualization.draw_geometries([pcd, key_point_cloud])
[0230] This code first calculates the normal of the point cloud, then iterates over each point in the point cloud and calculates the distance and normal direction difference of the points in its local neighborhood. Then, the statistical characteristics of these geometric properties are used to determine which points are key points. Finally, these key points are visualized for easy observation and further analysis. In this way, useful features can be effectively extracted from complex 3D data to support further data processing and analysis tasks.
[0231] Furthermore, the use of a directional histogram signature algorithm to extract features of key points in the fitted three-dimensional point cloud data model to obtain key point features includes:
[0232] Step a: select a key point as the center point, determine a predefined neighborhood radius, and select a specified number of neighboring points around the center point with the neighborhood radius;
[0233] Step b: for each neighboring point, calculate the vector between it and the center point, and standardize the direction of the vector relative to the surface normal of the center point to eliminate the deviation caused by different acquisition angles;
[0234] Step c: quantizing the direction of the normalized vector into a predetermined number of histogram boxes, and calculating the number of vectors falling in each histogram box to form a direction histogram;
[0235] Repeat steps ac to generate a direction histogram for each key point, and use the direction histogram as the key point feature.
[0236] First, a key point is selected as the center point. In the point cloud data model, each identified key point will be used as the focus of analysis in turn. After the key point is determined, a predefined neighborhood radius is set, which is used to define a circular (or spherical, depending on the dimension of the data) neighborhood around the center point and includes all points that fall within this range. These included points are called neighboring points, and they will be used to calculate the spatial characteristics of the center point.
[0237] Next, for each neighboring point, a vector pointing from the center point to the neighboring point is calculated. This vector reflects the relative position relationship between the two points. In order to ensure the consistency of the features and eliminate the deviation caused by the difference in observation angles, the direction of each vector needs to be normalized relative to the surface normal of the center point. This normalization process involves adjusting the direction of the vector so that it is comparable under different observation conditions.
[0238] Then, the direction of each normalized vector is quantized into a preset histogram. This means that the direction distribution of the vector is divided into a discrete direction interval, each interval represents a histogram. The number of vectors contained in each histogram is calculated to form a direction histogram. This histogram effectively describes the distribution characteristics of the vector direction in the space around the center point, thereby capturing the characteristics of the local geometric morphology of the center point.
[0239] This process is repeated for each key point, generating a direction histogram for each point that describes its local spatial structure. Each direction histogram is used as an important feature of the corresponding key point, which will be used for further data processing, such as point cloud registration, pattern recognition, or machine learning analysis.
[0240] The following is a reference implementation code of the directional histogram signature algorithm. This code example is written in Python.
[0241] import numpy as np
[0242] import open3d as o3d
[0243] def compute_vector(p1, p2):
[0244] """Calculate the vector between two points"""
[0245] return np.array(p2) - np.array(p1)
[0246] def normalize_vector(vector, normal):
[0247] """Normalize the vector direction according to the normal of the center point"""
[0248] # Normalize the vector to ensure it is in the same direction as the normal
[0249] norm_vector = vector - np.dot(vector, normal) * normal
[0250] norm_vector / = np.linalg.norm(norm_vector)
[0251] return norm_vector
[0252] def vector_to_histogram(vector, bins):
[0253] """Quantize the direction of a vector into histogram bins"""
[0254] # For simplicity, the vector angles are used here to assign them to histogram bins. In practice, a more complex quantization algorithm may be required
[0255] angle = np.arctan2(vector[1], vector[0])
[0256] bin_index = int((angle + np.pi) / (2 * np.pi) * bins)
[0257] return bin_index
[0258] def create_direction_histogram(key_point, point_cloud, radius, bins=36):
[0259] """Create a direction histogram for a given keypoint"""
[0260] histogram = np.zeros(bins)
[0261] kd_tree = o3d.geometry.KDTreeFlann(point_cloud)
[0262] [k, idx, _] = kd_tree.search_radius_vector_3d(key_point, radius)
[0263] if k > 1: # Make sure there are neighboring points
[0264] for i in idx[1:]: # exclude itself
[0265] vector = compute_vector(key_point, point_cloud.points[i])
[0266] normalized_vector = normalize_vector(vector, point_cloud.normals[idx[0]])
[0267] bin_index = vector_to_histogram(normalized_vector, bins)
[0268] histogram[bin_index] += 1
[0269] return histogram
[0270] # Load point cloud data
[0271] pcd = o3d.io.read_point_cloud("path_to_your_point_cloud.ply")
[0272] # Assume there is already a list of key points
[0273] key_points_indices = [10, 50, 100] # Example key point index
[0274] radius = 0.1 # Neighborhood radius
[0275] # Calculate the direction histogram for each key point
[0276] for index in key_points_indices:
[0277] key_point = pcd.points[index]
[0278] histogram = create_direction_histogram(key_point, pcd, radius)
[0279] print(f"Histogram for key point {index}: {histogram}")
[0280] Furthermore, the accurate registration of the point cloud data is achieved by iterative closest point algorithm according to the key point features, and the pose transformation matrix between the point cloud data model and the theoretical model after accurate registration is obtained, including:
[0281] Using the extracted key point features as a source point set, and performing feature matching between the source point set and the corresponding target point set in the theoretical model to form an initial point pair relationship;
[0282] The following iterative registration steps are performed:
[0283] Step a1: Based on the key point features, for each point in the source point set, find the nearest point with the most similar features in the target point set, and redetermine the point pair relationship;
[0284] Step a2: For each set of redetermined point pairs, calculate a transformation including rotation and translation so that after applying this transformation, the source point set is more closely aligned with the target point set;
[0285] Step a3: Apply the calculated transformation to the source point set so that the source point set is better aligned with the target point set;
[0286] Check whether the change between consecutive iterations is less than a preset threshold. If so, it is considered that the registration process has converged and the iteration is stopped; if not, continue the iteration steps of steps a1-a3;
[0287] After the iteration is completed, the rotation matrix and translation vector that achieve accurate registration are recorded, and the pose transformation matrix from the source point set to the target point set is constructed based on the accurate registration rotation matrix and translation vector.
[0288] In the alignment method for additive manufacturing products with special-shaped surfaces, the use of the Iterative Closest Point (ICP) algorithm for accurate registration of point cloud data is a key process. This process not only involves matching and transformation calculations between point sets, but also includes an iterative optimization process to ensure the best spatial alignment between the final point cloud data model and the theoretical model. The following is a detailed description of the steps.
[0289] First, the source point set is constructed using key point features extracted from the point cloud data. These features include the spatial position of the key points and other possible geometric or surface features, such as normal direction, curvature, etc. At the same time, these features are matched with the predefined features of the corresponding points in the theoretical model to form an initial point pair relationship. This preliminary matching provides a starting point for the iterative process.
[0290] The process of iterative registration involves the following detailed steps:
[0291] 1. Feature matching: At the beginning of each iteration, for each point in the source point set, the nearest point with the most similar features is found in the target point set based on its features. This process depends on the method of defining feature similarity, and the common ones are Euclidean distance, cosine similarity, etc. The re-determined point pair relationship will be used for the next transformation calculation.
[0292] 2. Transformation calculation: For each newly determined pair of points, calculate an optimal transformation, including rotation and translation. The purpose of this transformation is to minimize the distance from each point in the source point set to its matching target point. Usually, this involves solving a least squares problem to find the rotation matrix and translation vector that minimize the sum of the total distance squares.
[0293] 3. Transformation application: Apply the transformation calculated in the previous step to the source point set, update the position of the source point set to make it closer to the target point set. In this way, the source point set gradually moves closer to the position and direction of the target point set.
[0294] 4. Convergence check: After each iteration, check whether the total distance difference between the source point set and the target point set is less than the preset threshold. If so, the registration process is considered to have converged and the iteration is stopped. If not, continue to perform the above steps.
[0295] After the iteration is completed, the rotation matrix and translation vector when the accurate registration is achieved are recorded. Using these parameters, the pose transformation matrix from the source point set to the target point set is constructed, which describes how to transform the source point set from its current state to a state that is completely aligned with the target point set.
[0296] This process ensures precise geometric consistency between the theoretical model and the actual scanned additive manufacturing product, and is a key step in achieving high-quality additive manufacturing.
[0297] The following is a reference implementation code of the Iterative Closest Point (ICP) algorithm. The code example is written in Python.
[0298] import numpy as np
[0299] from scipy.spatial import cKDTree
[0300] from scipy.spatial.transform import Rotation as R
[0301] def find_nearest_neighbors(source_points, target_points):
[0302] """For each point in the source point set, find the closest point in the target point set"""
[0303] tree = cKDTree(target_points)
[0304] distances, indices = tree.query(source_points, k=1)
[0305] return indices, distances
[0306] def compute_transformation(source_points, target_points):
[0307] """Calculate the optimal transformation (rotation and translation) to align the source point set to the target point set"""
[0308] # Calculate the centroid
[0309] centroid_source = np.mean(source_points, axis=0)
[0310] centroid_target = np.mean(target_points, axis=0)
[0311] # Centered point set
[0312] source_centered = source_points - centroid_source
[0313] target_centered = target_points - centroid_target
[0314] # Calculate the covariance matrix
[0315] H = np.dot(source_centered.T, target_centered)
[0316] # SVD decomposition
[0317] U, _, Vt = np.linalg.svd(H)
[0318] # Calculate the rotation matrix
[0319] R = np.dot(Vt.T, UT)
[0320] # Ensure a correct rotation
[0321] if np.linalg.det(R) < 0:
[0322] Vt[-1, :] *= -1
[0323] R = np.dot(Vt.T, UT)
[0324] # Calculate the translation vector
[0325] t = centroid_target - np.dot(R, centroid_source)
[0326] return R, t
[0327] def iterative_closest_point(source, target, tolerance=1e-5, max_iterations=100):
[0328] """Execute the ICP algorithm and return the final transformation matrix and number of iterations"""
[0329] source_points = np.copy(source)
[0330] target_points = np.copy(target)
[0331] for i in range(max_iterations):
[0332] # Match point pairs
[0333] indices, distances = find_nearest_neighbors(source_points,target)
[0334] matched_target = target_points[indices]
[0335] # Calculate transformation
[0336] rotation, translation = compute_transformation(source_points,matched_target)
[0337] # Update source point set
[0338] source_points = np.dot(rotation, source_points.T).T +translation
[0339] # Check convergence
[0340] mean_error = np.mean(distances)
[0341] if mean_error < tolerance:
[0342] break
[0343] return rotation, translation, i
[0344] # Example: Source point set and target point set
[0345] source_points = np.random.rand(100, 3) # source point set
[0346] target_points = np.random.rand(100, 3) # target point set
[0347] # Execute ICP
[0348] rotation_matrix, translation_vector, iterations = iterative_closest_point(source_points, target_points)
[0349] print("Rotation matrix:", rotation_matrix)
[0350] print("Translation vector:", translation_vector)
[0351] print("Iterations:", iterations)
[0352] This code may need to be adjusted in actual application, especially to adjust the point cloud processing method and data structure according to the point cloud processing library actually used.
[0353] Step S105: Calculate the distance and angle that the additive manufacturing product needs to be adjusted according to the posture transformation matrix, and align the additive manufacturing product according to the calculated distance and angle.
[0354] Step S105 involves using the pose transformation matrix obtained from the previous step to calculate the necessary adjustment parameters, i.e., the distance and angle that the additively manufactured product needs to be adjusted, and then implementing these adjustments to ensure the precise alignment of the product. This step is the implementation stage of the actual adjustment operation in the entire alignment method, and the key lies in accurately converting theoretical data into actual operation instructions.
[0355] First, the pose transformation matrix is provided by the iterative closest point algorithm in the previous step, which contains the rotation and translation information that must be applied to the additively manufactured product to match its theoretical model. The pose transformation matrix is usually represented as a composite matrix containing a rotation matrix and a translation vector. Through this matrix, the spatial transformation required for the product from the current actual state to the theoretical design state can be clearly defined.
[0356] Next, the translation vector and rotation matrix are extracted from the pose transformation matrix. The translation vector directly gives the specific distance that the product needs to move along the XYZ axis, which can be directly used to adjust the position of the product. From the rotation matrix, the specific angles of rotation around each axis (usually X, Y, and Z axes) are calculated. These angles guide how to adjust the direction of the product to meet the design specifications.
[0357] When making adjustments, it is necessary to operate the corresponding mechanical components on the additive manufacturing equipment, such as the mobile platform, rotating frame, etc., to accurately move and rotate according to the calculated distance and angle. This is usually performed by a programmed controlled robotic arm or other automated mechanical facilities to ensure the accuracy and repeatability of the adjustments.
[0358] In addition, it is very important to continuously monitor and record the actual effects of the adjustments during implementation to verify whether the adjustments have achieved the expected accuracy. This may involve repeatedly using 3D scanning technology to confirm whether the actual state of the product is completely aligned with the theoretical model.
[0359] Finally, after all adjustments are made, a final inspection is performed to ensure that every parameter precisely meets the design requirements. If deviations are found, fine-tuning may be required to ensure that every aspect of the product strictly meets the design specifications.
[0360] Through the above steps, it can be ensured that the final position and orientation of the additive manufacturing product accurately matches the design model, thereby ensuring the optimization of product quality and function. This process not only improves the overall quality of the product, but also optimizes the efficiency and reliability of the manufacturing process.
[0361] Furthermore, the calculating the distance and angle that the additive manufacturing product needs to be adjusted according to the posture transformation matrix, and aligning the additive manufacturing product according to the calculated distance and angle, includes:
[0362] Parsing the pose transformation matrix obtained from the iterative closest point algorithm ,in is a The rotation matrix of is a three-dimensional translation vector;
[0363] Calculate the distance to be adjusted according to the following formula 4 :
[0364]
[0365] in is the translation vector The amount of
[0366] Calculate the Euler angle according to the following formula 5-7 , the Euler angle Indicates the angles that the additive manufacturing product needs to rotate around the X-axis, Y-axis, and Z-axis:
[0367]
[0368]
[0369]
[0370] in, Represents the rotation matrix Middle Line Elements of a column;
[0371] According to the calculated distance and Euler angles , adjust the position and orientation of the additively manufactured product.
[0372] In the alignment method of additively manufactured products with special-shaped surfaces, the pose transformation matrix is used to accurately calculate the distance and angle that the product needs to be adjusted, and the position and direction of the product are adjusted accordingly. This process ensures that technicians can effectively implement the alignment operation based on the pose transformation matrix to ensure the best geometric consistency between the product and the design model. The specific steps and calculation methods of this process are described in detail below.
[0373] First, parse the pose transformation matrix obtained from the iterative closest point algorithm. This matrix is usually expressed as an enhanced matrix, which includes a 3×3 rotation matrix R and a three-dimensional translation vector t, in the following form:
[0374]
[0375] Here, R describes the rotation of the AM product relative to its theoretical model, while t describes the translation relative to the theoretical model.
[0376] Next, according to the components of the translation vector t on the X-axis, Y-axis, and Z-axis , calculate the distance d that needs to be adjusted. This distance is the straight-line distance required to move from the current position to the theoretical position, and the calculation method is as follows:
[0377]
[0378] This distance provides the absolute amount of movement of the product in space to ensure that it is accurately positioned as designed.
[0379] Then, the Euler angles are calculated using the rotation matrix R, which is the angles at which the product needs to rotate around the X, Y, and Z axes. The Euler angles are calculated using the following formula:
[0380]
[0381]
[0382]
[0383] here, Represents the element in the i-th row and j-th column of the rotation matrix R. These three angles represent the rotation around the X-axis, Y-axis, and Z-axis, respectively, so that the product can be transformed from the current posture to a posture that is completely aligned with the theoretical model.
[0384] Finally, according to the calculated distance d and Euler angle , making actual position and orientation adjustments. This typically involves physically moving and rotating the product on the AM equipment, or adjusting the position and orientation relative to the stage in software.
[0385] In this way, additive manufacturing products can be precisely adjusted according to theoretical designs, ensuring that the manufactured products have high precision and geometric characteristics that meet design specifications. This process not only improves product quality, but also optimizes production efficiency and material usage efficiency.
[0386] Furthermore, the alignment method for additively manufactured products with special-shaped surfaces further includes:
[0387] The hydraulic pressure holding control mechanism is used to ensure that the additive manufacturing product is in a consistent state during the processing.
[0388] A key additional step in the alignment method for additively manufactured products with special-shaped surfaces is the use of a hydraulic pressure holding control mechanism to ensure that the additively manufactured product remains consistent throughout the entire processing process. This step is critical because it ensures the accuracy of the processing and the high standards of product quality. The following describes this process in detail.
[0389] First, the hydraulic pressure holding control mechanism is introduced to solve the problem of product displacement or deformation that may occur during the additive manufacturing process. Additive manufacturing, especially manufacturing involving complex geometric structures and high precision requirements, often requires the application of continuous and uniform force during the processing to maintain the stability of the product. This is because during the manufacturing process, especially under high temperature or gravity, the product may be displaced or deformed due to insufficient support.
[0390] The hydraulic pressure control mechanism maintains continuous and uniform support for the product by automatically adjusting the hydraulic pressure during the manufacturing process. This mechanism usually includes a set of hydraulic pumps, valves and sensors, which work together to adjust the hydraulic pressure based on real-time pressure data. These sensors can detect changes in pressure at the interface between the product and the tooling, and feed the data back to the control system, which adjusts the output of the hydraulic pump to ensure that the pressure applied to the product is always maintained at the preset optimal level.
[0391] In specific implementation, it is necessary to first set the pressure parameters of the hydraulic control system, which are usually based on the properties of the product material, the specific technology of the additive manufacturing process (such as laser melting, electron beam melting, etc.), and the geometric characteristics of the product. Once set, the system will continuously monitor and adjust the pressure during the processing to respond to any pressure changes caused by the processing.
[0392] In addition, maintaining the consistency of product status not only helps to ensure product quality, but also helps to reduce material waste and improve processing efficiency. Inconsistency in the status during processing often leads to product defects, requiring rework or scrapping, which increases costs and time.
[0393] Through the above detailed steps, the hydraulic pressure holding control mechanism becomes a key technology to ensure product quality and processing accuracy in the additive manufacturing process. The correct implementation of this mechanism can significantly improve production reliability and product consistency, and is an important guarantee for achieving high-quality additive manufacturing. The successful implementation of this process is especially important for additive manufacturing projects that require extremely high precision and structural integrity.
[0394] A second embodiment of the present application provides an electronic device, the electronic device comprising:
[0395] processor;
[0396] The memory is used to store a program, and when the program is read and executed by the processor, the program executes the alignment method for additively manufactured products with special-shaped surfaces provided in the first embodiment of the present application.
[0397] The third embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the method for aligning a product made of additively manufactured products with special-shaped surfaces provided in the first embodiment of the present application is executed.
[0398] Although the present application is disclosed as above in the form of a preferred embodiment, it is not intended to limit the present application. Any technical personnel in this field may make possible changes and modifications without departing from the spirit and scope of the present application. Therefore, the scope of protection of the present application shall be based on the scope defined by the claims of the present application.
Claims
1. A method for aligning a product made of additively manufactured products with special-shaped surfaces, characterized in that: include: Use hydraulic tooling profiling blocks and ball studs to perform preliminary positioning of additively manufactured products; Using 3D data acquisition technology, obtain the 3D point cloud data model of the surface of the additive manufacturing product after preliminary positioning; By using the common point least squares method coordinate transformation principle, the best fit between the coordinate system of the three-dimensional point cloud data model and the coordinate system of the theoretical model is achieved to obtain the fitted three-dimensional point cloud data model; wherein the theoretical model is constructed according to the design parameters of the additive manufacturing product by computer-aided design software; The key points of the point cloud data in the three-dimensional point cloud data model after fitting are extracted by using an internal signature description algorithm; the key points are extracted by using a directional histogram signature algorithm to obtain key point features; based on the key point features, accurate registration of the point cloud data is achieved by using an iterative closest point algorithm to obtain a pose transformation matrix between the point cloud data model after accurate registration and the theoretical model; Calculating the distance and angle that the additive manufacturing product needs to be adjusted according to the posture transformation matrix, and aligning the additive manufacturing product according to the calculated distance and angle; The method of achieving the best fit between the coordinate system of the three-dimensional point cloud data model and the coordinate system of the theoretical model by using the common point least squares method coordinate transformation principle to obtain the fitted three-dimensional point cloud data model includes: For the data points in the coordinate system of the theoretical model And the data points in the coordinate system of the 3D point cloud data model , the weight is calculated according to the following formula (1) : in, Yes The measurement uncertainty of is a positive constant; the point and Point correspond to each other; According to the following formula (2), the total deviation is defined as : in, is the rotation matrix; is the translation vector; is the total number of data points in the coordinate system of the theoretical model; The Lagrangian function is constructed according to the following formula (3): : in, is the symmetric Lagrange multiplier matrix; represents the trace of a matrix; is the rotation matrix The transpose of is the identity matrix; Solving the Lagrangian function using the Newton–Raphson iteration method , find Minimum target rotation matrix and the target translation variable ; According to the target rotation matrix and the target translation variable , obtain the fitted three-dimensional point cloud data model; The step of extracting key points of the point cloud data in the fitted three-dimensional point cloud data model by using an internal signature description algorithm includes: Selecting a target point in the point cloud data in the fitted three-dimensional point cloud data model as an analysis center point; A local neighborhood is defined, wherein a specified number of neighboring points are included around the analysis center point in the local neighborhood; For each point in the analysis center point and the local neighborhood, the geometric properties between it and the neighboring points are calculated; wherein the geometric properties include relative position and normal direction difference; Based on the geometric attributes, a signature matrix describing the local geometric shape of the analysis center point is generated, wherein the signature matrix includes the distances and angles between points and their statistical distribution characteristics; Compare the obtained signature matrix with a preset threshold to determine whether the center point is a key point; The step of using a directional histogram signature algorithm to extract features of key points in the fitted three-dimensional point cloud data model to obtain key point features includes: Step a: select a key point as the center point, determine a predefined neighborhood radius, and select a specified number of neighboring points around the center point with the neighborhood radius; Step b: for each neighboring point, calculate the vector between it and the center point, and standardize the direction of the vector relative to the surface normal of the center point to eliminate the deviation caused by different acquisition angles; Step c: quantizing the direction of the normalized vector into a predetermined number of histogram boxes, and calculating the number of vectors falling in each histogram box to form a direction histogram; Repeat steps ac to generate a direction histogram for each key point, and use the direction histogram as a key point feature; The step of calculating the distance and angle that the additive manufacturing product needs to be adjusted according to the posture transformation matrix, and aligning the additive manufacturing product according to the calculated distance and angle, includes: Parsing the pose transformation matrix obtained from the iterative closest point algorithm ,in is a 3×3 rotation matrix, is a three-dimensional translation vector; Calculate the distance to be adjusted according to the following formula (4): : in is the translation vector The amount of The Euler angles are calculated using the following formulas (5)-(7): , the Euler angle Indicates the angles that the additive manufacturing product needs to rotate around the X-axis, Y-axis, and Z-axis: in, Represents the rotation matrix Middle Line Elements of a column; According to the calculated distance and Euler angles , adjust the position and orientation of the additively manufactured product.
2. The alignment method for additively manufactured products with special-shaped surfaces according to claim 1, characterized in that: According to the key point features, the accurate registration of the point cloud data is achieved by iterative closest point algorithm, and the pose transformation matrix between the point cloud data model and the theoretical model after accurate registration is obtained, including: Using the extracted key point features as a source point set, and performing feature matching between the source point set and the corresponding target point set in the theoretical model to form an initial point pair relationship; The following iterative registration steps are performed: Step a1: Based on the key point features, for each point in the source point set, find the nearest point with the most similar features in the target point set, and redetermine the point pair relationship; Step a2: For each set of redetermined point pairs, calculate a transformation including rotation and translation so that after applying this transformation, the source point set is more closely aligned with the target point set; Step a3: Apply the calculated transformation to the source point set so that the source point set is better aligned with the target point set; Check whether the change between consecutive iterations is less than a preset threshold. If so, it is considered that the registration process has converged and the iteration is stopped; if not, continue the iteration steps of steps a1-a3; After the iteration is completed, the rotation matrix and translation vector that achieve accurate registration are recorded, and the pose transformation matrix from the source point set to the target point set is constructed based on the accurate registration rotation matrix and translation vector.
3. The alignment method for additively manufactured products with special-shaped surfaces according to claim 1, characterized in that: The method of using the hydraulic tooling profiling block and the ball stud to perform preliminary positioning of the additive manufacturing product includes: In the design and manufacturing process of the additive manufacturing product, a coarse reference block is formed or integrated at a key part of the additive manufacturing product, and the coarse reference block clearly appears on the surface of the additive manufacturing product as a physical feature to support the subsequent clamping and positioning process; Placing a hydraulic tooling profiling block to fit the coarse reference block, and ensuring accurate docking between the hydraulic tooling profiling block and the coarse reference block through mechanical adjustment, wherein the shape and size of the hydraulic tooling profiling block are designed to fully match the coarse reference block; The hydraulic tooling copy block is mechanically fixed to the coarse reference block using a ball stud, ensuring that the additively manufactured part remains stable and in the correct position throughout the machining or measurement process.
4. The alignment method for additively manufactured products with special-shaped surfaces according to claim 1, characterized in that: The method of using the three-dimensional data acquisition technology to obtain the three-dimensional point cloud data model of the surface of the additively manufactured product after preliminary positioning includes: Use a high-precision laser scanner to scan the additively manufactured product from multiple angles to obtain complete surface point cloud data; Preprocessing the acquired point cloud data, wherein the preprocessing includes denoising, filtering and data compression; According to the preprocessed point cloud data, a three-dimensional point cloud data model of the surface of the additive manufacturing product after preliminary positioning is obtained.
5. The alignment method for additively manufactured products with special-shaped surfaces according to claim 4, characterized in that: The method of using the three-dimensional data acquisition technology to obtain the three-dimensional point cloud data model of the surface of the additively manufactured product after preliminary positioning also includes: When using high-precision laser scanners to scan additively manufactured products from multiple angles, the scanning speed and resolution of the 3D scanning equipment are optimized according to the material characteristics and geometric complexity of the additively manufactured products.
6. The alignment method for additively manufactured products with special-shaped surfaces according to claim 1, characterized in that: Also includes: The hydraulic pressure holding control mechanism is used to ensure that the additive manufacturing product is in a consistent state during the processing.
Citation Information
Patent Citations
MEMS triaxial accelerometer calibration method based on maximum likelihood estimation algorithm
CN110174122A
Workpiece three-dimensional point cloud registration method based on direction histogram signature features
CN111862177A
Workpiece position alignment method using point cloud registration technology and three-dimensional reconstruction technology
CN113421291A
Three-dimensional curved surface 3D printing path planning method and system
CN117261232A
3D measurement model and spatial calibration method based on 1D displacement sensor
US20210095959A1
