Three-dimensional printing method and printing system

Abstract

The invention relates to the field of three-dimensional printing, and discloses a three-dimensional printing method and a three-dimensional printing system, which realize instant discovery and prediction of printing defects by monitoring the printing process in real time and comparing the difference between a digital model and a printing entity. By means of the innovation, the precision and efficiency of three-dimensional printing are improved, and material waste and time cost caused by defects are greatly reduced. By continuously optimizing printing setting, geometric and topological defects in the printing process can be automatically repaired, it is ensured that a printed object meets the design requirement, and the reliability and practicability of the three-dimensional printing technology are improved.

Description

Technical Field

[0001] The present invention relates to the field of three-dimensional printing, and in particular to a three-dimensional printing method and a printing system. Background Art

[0002] As a rapid prototyping technology, 3D printing has received widespread attention and application in recent years. It processes digital models in layers and then builds objects by stacking materials layer by layer. It has high flexibility and personalized customization capabilities. However, in practical applications, 3D printing technology still faces many challenges.

[0003] On the one hand, since the 3D printing process involves complex material deposition and curing processes, the printing results are often affected by multiple factors, such as printing speed, temperature, material properties, etc. Changes in these factors may cause defects in the printed objects, such as rough surface, poor interlayer adhesion, dimensional deviation, material unevenness or deformation, etc. These defects not only affect the aesthetics of the object, but may also reduce its mechanical properties and functional characteristics.

[0004] On the other hand, traditional 3D printing methods usually lack real-time monitoring and feedback mechanisms, and are unable to promptly detect and correct deviations in the printing process. This means that once a problem occurs, manual inspection and repair are often required after printing is completed, which not only increases costs and time, but may also affect the efficiency and accuracy of the entire production process. Summary of the invention

[0005] The purpose of the present invention is to provide a three-dimensional printing method and printing system. The method captures information during the printing process in real time, generates a digital model corresponding to the actual printing result, performs comparison and defect prediction, and adjusts the printing settings accordingly to achieve improved and continuous optimization of printing quality.

[0006] In order to achieve the above technical objectives and the above technical effects, the present invention discloses a three-dimensional printing method, which at least includes the following steps:

[0007] S1, scanning and modeling: obtain the geometric information of the actual object through three-dimensional scanning, and use modeling software to generate the corresponding digital body 1;

[0008] S2, printing and monitoring: starting the three-dimensional printing process, and setting a scanning camera in the printing chamber to capture the information during the printing process in real time, and transmitting the captured data back to the system to generate a digital body 2 corresponding to the actual printing result;

[0009] S3, comparison and defect prediction: compare digital body 1 with digital body 2, analyze the difference between them, so as to identify and predict future printing defects and their possible locations;

[0010] S4, defect repair: according to the comparison result, the printing settings of the digital body 2 are adjusted to repair or optimize the defects in the printing process;

[0011] S5. Correction: In the subsequent printing process, the comparison between the digital body 1 and the digital body 2 is continuously performed, and the printing settings are corrected based on the comparison results to achieve closed-loop control and continuous optimization of printing quality.

[0012] Preferably, the S2 step at least includes:

[0013] S21, deploying a number of scanning cameras in the printing chamber;

[0014] S22, during the printing process of the printer, the scanning camera is automatically triggered to collect images and obtain the image data of the layer;

[0015] S23, the system acquires scanning camera image data to form layered model data.

[0016] Furthermore, the step S21 further includes a motion track, the scanning camera is mounted on the motion track, and the motion track is a plurality of vertical motion tracks or a plurality of spiral ascending tracks;

[0017] Wherein, the vertical motion track is arranged symmetrically with respect to the printing area axis of the printing chamber;

[0018] Wherein, the spiral ascending track is arranged with equal pitch with the center of the printing area of ​​the printing chamber as the axis.

[0019] Further, the step S23 at least includes:

[0020] S231, extracting feature points from image data of multiple perspectives, and generating point cloud data;

[0021] S232, accurately registering and fusing point cloud data acquired from different perspectives;

[0022] S233, converting the point cloud data into a mesh model, performing triangulation processing, and generating a continuous surface triangulated mesh model.

[0023] Preferably, the S3 step at least includes:

[0024] S31, model alignment: spatially align digital volume 1 and digital volume 2 using spatial transformation to ensure that the two models are in the same coordinate system before comparison, eliminating errors caused by position, rotation or scaling;

[0025] S32, defect classification: using a deep learning model to identify and classify defects on the adjusted digital body 2 to obtain printing defect information;

[0026] S33, defect prediction: import the printing defect information and printer data information in the digital body 2 into the time series model, and predict the location and type of possible defects in the future through continuous data analysis.

[0027] Wherein: the time series model adopts an autoregressive differential moving average model or a vector autoregressive model, and establishes a prediction relationship between time and defects by using the printing defect information in the digital body 2 obtained in continuous time to obtain predicted defect information.

[0028] Preferably, the following is further included between step S31 and step S32:

[0029] Difference calculation: Determine the degree of deviation by calculating the distance error between the corresponding feature points in digital body 1 and digital body 2;

[0030] Hotspot identification: Based on the calculation results, an error threshold is set and feature point differences exceeding this threshold are identified as hotspots.

[0031] Preferably, in step S4, "repairing or optimizing defects in the printing process" includes repairing printing defects and optimizing predicted defects.

[0032] The present invention also discloses a printing system, which is based on the above three-dimensional printing method and at least comprises:

[0033] A scanning module, which is used to obtain geometric information of an actual object by three-dimensional scanning, generate a corresponding digital model using modeling software, and transmit the obtained digital model information to a data processing module and / or a printing module;

[0034] Printing module: The printing module is connected to the data processing module and performs two-way data interaction;

[0035] Scanning camera module: The scanning camera module is arranged in the printing module, and is connected with the data processing module and / or the printing module, and performs two-way data interaction;

[0036] Data processing module: The data module is used to receive data from the scanning module, printing module, and scanning camera module for processing, and generate control instructions to be sent to the corresponding modules.

[0037] The present invention has the following beneficial effects:

[0038] 1. The present invention monitors the printing process in real time and compares the differences between the digital model and the printed entity, thereby realizing the instant discovery and prediction of printing defects. This innovation not only improves the accuracy and efficiency of 3D printing, but also greatly reduces the material waste and time cost caused by defects. By continuously optimizing the printing settings, the present invention can automatically repair geometric and topological defects in the printing process, ensure that the printed object meets the design requirements, and improve the reliability and practicality of 3D printing technology.

[0039] 2. Through reasonable design and layout, the present invention deploys multiple scanning cameras in the printing chamber, which can capture information in the printing process from all directions and angles. Combined with advanced image processing technology and deep learning models, the present invention can accurately identify and analyze the type and location of printing defects, providing strong data support for subsequent optimization and repair. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is an overall flow chart of Example 1 of the present invention. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments.

[0042] Example 1

[0043] like Figure 1 As shown, this embodiment discloses a three-dimensional printing method, which includes the following steps:

[0044] S1, scanning and modeling: obtain the geometric information of the actual object through three-dimensional scanning, and use modeling software to generate the corresponding digital body 1;

[0045] S2, printing and monitoring: starting the three-dimensional printing process, and setting a scanning camera in the printing chamber to capture the information during the printing process in real time, and transmitting the captured data back to the system to generate a digital body 2 corresponding to the actual printing result;

[0046] S3, comparison and defect prediction: compare digital body 1 with digital body 2, analyze the difference between them, so as to identify and predict future printing defects and their possible locations;

[0047] S4, defect repair: according to the comparison result, the printing settings of the digital body 2 are adjusted to repair or optimize the defects in the printing process;

[0048] S5. Correction: In the subsequent printing process, the comparison between the digital body 1 and the digital body 2 is continuously performed, and the printing settings are corrected based on the comparison results to achieve closed-loop control and continuous optimization of printing quality.

[0049] Wherein, the S2 step comprises:

[0050] S21, deploying a number of scanning cameras in the printing chamber;

[0051] S22, during the printing process of the printer, the scanning camera is automatically triggered to collect images and obtain the image data of the layer;

[0052] S23, the system acquires scanning camera image data to form layered model data.

[0053] Further, the step S23 includes:

[0054] S231, extracting feature points from image data of multiple perspectives, and generating point cloud data;

[0055] S232, accurately registering and fusing point cloud data acquired from different perspectives;

[0056] S233, converting the point cloud data into a mesh model, performing triangulation processing, and generating a continuous surface triangulated mesh model.

[0057] Specifically, when the print job needs to be based on a physical object, we use high-precision scanning technology to obtain detailed geometric data of the entity. In this way, we can directly generate a digital model from the scanned data, thus avoiding the measurement errors or inaccuracies in model reconstruction that may exist in traditional methods. If the physical object already exists in the form of a digital model, then this scanning process can be omitted and the existing digital model can be directly used for printing preparation.

[0058] A plurality of scanning cameras form a circular array around the printing area to ensure that the printing status of each layer can be captured from different angles. After each layer is printed, the scanning camera in the circular array is automatically triggered to take a photo of the current layer and obtain the image data of the layer.

[0059] By processing the acquired image data of each layer into point cloud, key feature points such as edges, corners, and unique texture areas are extracted from the images captured by each camera.

[0060] These feature points are converted into point cloud data in three-dimensional space, where each point contains its coordinates (x, y, z) in space. Point cloud data formula example: P = {(x1, y1, z1), (x2, y2, z2), …, (x n ,y n ,z n )}.

[0061] Using the iterative closest point ICP algorithm or other feature matching techniques, point cloud data from different perspectives are registered to align their positions in a unified coordinate system. Here, the ICP algorithm is used as an example to align point cloud data acquired from multiple perspectives into a common reference coordinate system:

[0062] S2321, find the closest point match:

[0063] For each point in the source point cloud, find the nearest point in the target point cloud. The formula is:

[0064] Match(P s ,P t )={(p s1 ,p t1 ),(p s2, p t2 ),…,(p sn ,p tn )} Among them, P s is the source point cloud, P t is the target point cloud, p si is a point in the source point cloud, p ti is the closest point in the target point cloud.

[0065] S2322, Rigid body transformation estimation:

[0066] Estimate the best rigid body transformation, i.e. translation and rotation, so that the distance between the matching point pairs is minimized, where:

[0067] The estimation of the rotation matrix R and the translation vector t can be achieved by minimizing the following cost function:

[0068] Among them, E(R,t) is the cost function, and |||| represents the Euclidean distance.

[0069] S2323, Transformation Application:

[0070] Apply the estimated transformation to the source point cloud, update the position of the source point cloud, and obtain the updated point cloud: P′ s = {Rp s1 +t,Rp s2 +t,...,Rp sn +t}, where P s ′ represents the source point cloud P after the transformation by the rotation matrix R and the translation vector t s The point in .

[0071] S2324, iterative optimization: repeat the steps of closest point matching, transformation estimation and transformation application until the change of transformation or error is lower than the preset threshold. After each iteration, compare the error change of the current iteration with that of the previous iteration. If the change is less than the threshold, or the change of the rotation matrix and translation vector is small, the algorithm is considered to have converged.

[0072] S2325. Output the final rotation matrix R and translation vector t. These parameters define the precise transformation required to transform the source point cloud to align with the target point cloud. Traverse all point clouds and repeat steps S2321-S2324.

[0073] Apply a triangulation algorithm, such as Delaunay triangulation or greedy triangulation, to convert the point cloud data into a continuous triangular mesh model. In this embodiment, Delaunay triangulation is a method for converting point cloud data into a triangular mesh model, which specifically includes:

[0074] S2331, Delaunay triangulation: For every three non-collinear points in the point cloud, construct a triangle.

[0075] The key to Delaunay triangulation is to ensure that the circumcircle of any triangle does not contain other points, which is guaranteed by the Delaunay condition: Delaunay(P) = {T1, T2, …, T m} Among them, T i It is the point p i1 ,p i2 ,p i3 The triangle formed satisfies the Delaunay condition.

[0076] S2332. Construct a triangular mesh: combine all triangles that meet the Delaunay condition to form a complete triangular mesh model. Triangular mesh model: M = {V, E, F}, where V is a vertex set, E is an edge set, and F is a face set, and each face is defined by three vertices.

[0077] Through the above steps, the point cloud data acquired from multiple perspectives can be accurately registered through the ICP algorithm, and then the registered point cloud data can be converted into a continuous triangular mesh model using Delaunay triangulation to provide an accurate three-dimensional model for subsequent defect detection and quality control.

[0078] Wherein, the S3 step includes:

[0079] S31, model alignment: spatially align digital volume 1 and digital volume 2 using spatial transformation to ensure that the two models are in the same coordinate system before comparison, eliminating errors caused by position, rotation or scaling;

[0080] S32, difference calculation: determining the degree of deviation by calculating the distance error between corresponding feature points in digital body 1 and digital body 2;

[0081] S34, hotspot identification: according to the calculation results, an error threshold is set, and feature point differences exceeding this threshold are identified as hotspots;

[0082] S34, defect classification: using a deep learning model to identify and classify defects on the adjusted digital body 2 to obtain printing defect information;

[0083] S35, defect prediction: import the printing defect information and printer data information in the digital body 2 into the time series model, and predict the location and type of possible defects in the future through continuous data analysis.

[0084] Wherein: the time series model adopts an autoregressive differential moving average model or a vector autoregressive model, and establishes a prediction relationship between time and defects by using the printing defect information in the digital body 2 obtained in continuous time to obtain predicted defect information.

[0085] Specifically, digital body 1 and digital body 2 are first spatially aligned through spatial changes, such as appropriate scaling, rotation, translation and other operations, and then the distance error between the two is calculated. By determining a suitable deviation value as an error threshold, such as 2% or 5%, the feature point differences that exceed this threshold are identified as hotspots and used as key focus areas in subsequent deep learning recognition.

[0086] By pre-screening out the point cloud areas with large differences, the amount of data that the deep learning model needs to process can be reduced. This means that the model only needs to focus on those areas where defects may exist, rather than the entire point cloud dataset. By focusing on potential defect areas, the model can learn the characteristics of these areas more effectively, which may improve the accuracy of defect classification.

[0087] The hot spot area or the whole point cloud data set is analyzed by a data learning model such as PointCNN model or PointNet, and the geometric properties such as curvature, area and topological properties such as connected components and holes in the point cloud are identified, and the related defects are further analyzed. This embodiment is described by taking the PointCNN model as an example:

[0088] The designed PointCNN model includes:

[0089] Input Layer:

[0090] Input data: Receive preprocessed point cloud data, where each point cloud contains N points, each point has three-dimensional coordinates (x, y, z). Coordinate dimensions such as color can be added under conditions that need to be judged.

[0091] Data preprocessing: Normalize the point cloud data to ensure that the coordinate values ​​of all points are within a reasonable range.

[0092] Feature extraction layer:

[0093] For each point, use the K nearest neighbor algorithm to find the K points around it.

[0094] Calculate geometric features of the local area, such as distance between points, normals, etc. (Although we only use coordinates in this example, these features may be useful in real applications).

[0095] Use a multi-layer perceptron (MLP) to extract features from points in the local area and generate a local feature vector, f i =MLP(x i ,N(x i ) where f i It is i The local eigenvector of the point, x i is the coordinate of the point, N(x i ) represents point x i 's neighborhood.

[0096] X-Transform:

[0097] Apply X-transform to the local feature vector to extract more discriminative features. X-transform is a learnable transformation matrix that is adaptively adjusted according to the point cloud distribution in the local area. i ′=Xf i Among them, X is the learnable transformation matrix, f i ′ is the transformed eigenvector.

[0098] Feature Propagation Layer:

[0099] The extracted local features are propagated back to each point in the original point cloud. This can be achieved by interpolation or other methods to ensure that each point has a feature vector associated with it. This embodiment uses spherical convolution: the features are propagated in the point cloud through spherical convolution operations to capture a wider range of local structural information.

[0100] Pooling layer:

[0101] Through farthest point sampling, M farthest points are selected from the point cloud as pooling center points; local area division is further adopted, and for each pooling center point, the local area around it is found using ball query or K nearest neighbor algorithm; the points in each local area are feature aggregated to generate the pooled feature vector. This can be achieved through operations such as summation, averaging, and maximum pooling.

[0102] Global feature extraction layer:

[0103] The pooled feature vector is input into a multi-layer perceptron (MLP) to extract global features. This MLP can contain multiple hidden layers, each of which is followed by a ReLU activation function.

[0104] Fully connected layer:

[0105] The global feature vector is input into a fully connected layer whose number of output neurons is equal to the number of defect categories. The output of the fully connected layer is normalized using the Softmax activation function to obtain the probability value of each category.

[0106] Output Layer

[0107] The output layer receives the output of the fully connected layer and gives the probability value of each category. The category with the maximum probability value is taken as the prediction result.

[0108] By training the PointCNN model, we can perform deep learning analysis on the point cloud data generated during the 3D printing process, effectively identify and classify various defects including surface roughness, poor interlayer adhesion, dimensional deviation, material unevenness and material deformation, and further classify the defects to obtain printing defect information.

[0109] In step S35, for defect prediction, the printing defect information and printer data information in the digital body 2 are imported into the time series model, and the location and type of possible defects in the future are predicted through continuous data analysis. The time series model can adopt an autoregressive difference moving average model or a vector autoregressive model. Specifically:

[0110] When the defects in the 3D printing process are relatively simple, or present defects in continuous areas, such as dimensional deviations and material unevenness caused by local material loss, the use of the autoregressive differential moving average (ARIMA) model for prediction may be a better choice. The ARIMA model can handle non-stationary time series, convert non-stationary series into stationary series through differential operations, and then apply autoregressive (AR) and moving average (MA) components for modeling and prediction. The advantage of this method is that it can capture short-term dependencies in time series and is sensitive to periodic changes.

[0111] In this case, the ARIMA model can effectively analyze the data of historical printing layers and identify the relationship between factors such as material deposition, temperature control and printing speed and defect formation. Through in-depth analysis of these factors, the model can predict similar defects that may occur in future printing layers. For example, if the model identifies that under specific printing conditions, there is a trend of increasing defects such as uneven materials, it can issue an early warning, allowing the operator to adjust the printing parameters to avoid or reduce the occurrence of such defects.

[0112] For defects with various forms or mutations, such as poor interlayer adhesion, the vector autoregression (VAR) model may be more appropriate. The VAR model is a multivariate model that allows for interdependence between multiple time series. This model is particularly suitable for analyzing and predicting defects that may be caused by multiple factors.

[0113] In the case of poor interlayer adhesion, factors that may be involved include print head movement speed, material extrusion volume, printing environment conditions, etc. The VAR model is able to consider these variables simultaneously and analyze their interactions and impact on the final print quality. By building a multivariate model that includes these factors, VAR can help identify which combination of factors may lead to poor interlayer adhesion problems. Once these factors are identified, the model can predict the risk of poor interlayer adhesion under specific printing conditions, allowing the printing operator to take preventive measures, such as adjusting the print path, changing the print speed, or optimizing the material extrusion strategy.

[0114] Through this multivariate approach, the VAR model provides a more comprehensive tool to analyze and predict complex defects that may be caused by multiple factors. The predictive power of this method is based on in-depth analysis of historical data, making the monitoring and quality control of the 3D printing process more accurate and effective.

[0115] In the 3D printing process, the S4 step focuses on using the data and analysis results collected in the previous steps to repair the discovered printing defects and optimize the potential problems foreseen by the predictive model. This step is a key link in printing quality control, which involves fine-tuning the printing parameters to ensure that the printed objects meet the design specifications and quality standards.

[0116] The core of the S4 step is to implement a feedback loop in which the repair and optimization of printing defects are carried out in parallel. First, for identified defects such as surface roughness, dimensional deviation or material unevenness, the system will automatically adjust the printing parameters, such as adjusting the printing speed, temperature, material extrusion rate, etc., to repair these defects. For example, if poor interlayer adhesion is detected, the print head may increase the dwell time on each layer or adjust the printing path to enhance the interlayer bonding.

[0117] At the same time, for those potential defects predicted by the time series model, the S4 step will adjust the printing settings in advance to avoid the occurrence of defects. This may include changing the amount of material extruded, adjusting the movement speed of the print head, or changing the printing direction. Through these preventive measures, the uncertainty and potential risk of failure in the printing process can be significantly reduced.

[0118] Step S5 is about continuous correction and optimization during the printing process. In the subsequent stages of printing, the system will continuously compare digital body 1 and digital body 2. This comparison allows the system to monitor the printing quality in real time and make necessary adjustments based on the comparison results. This closed-loop control ensures the stability and consistency of the printing process, while also allowing the entire printing design to be continuously optimized based on the actual printing situation.

[0119] Through this continuous monitoring and adjustment, the S5 step ensures dynamic optimization of the printing process, making the printed object as close to the original design as possible in terms of geometry and topology. This approach not only improves the success rate of a single printing task, but also accumulates valuable data and experience for future printing tasks, further improving the overall performance and reliability of the printing system.

[0120] Example 2

[0121] This embodiment is based on the embodiment 1. For the three-dimensional printing method proposed in the embodiment 1, a necessary printing system is established. The printing system includes:

[0122] A scanning module, which is used to obtain geometric information of an actual object by three-dimensional scanning, generate a corresponding digital model using modeling software, and transmit the obtained digital model information to a data processing module and / or a printing module;

[0123] Printing module: The printing module is connected to the data processing module and performs two-way data interaction;

[0124] Scanning camera module: The scanning camera module is arranged in the printing module, and is connected with the data processing module and / or the printing module, and performs two-way data interaction;

[0125] Data processing module: The data module is used to receive data from the scanning module, printing module, and scanning camera module for processing, and generate control instructions to the corresponding modules.

[0126] Specifically, the scanning module can be a scanner, a handheld scanning device, etc. The accuracy of the image captured by the scanning device directly affects the amount of data in subsequent calculations;

[0127] The printing module is a printer or a printer with integrated data processing and display functions, which can complete basic three-dimensional material printing or can realize data processing and command printing at the same time;

[0128] The scanning camera module can be an industrial camera, a camera module with depth of field photography, or even a camera module with infrared photography and X-ray photography. Similarly, the image accuracy of the scanning camera module directly determines the amount of data in subsequent calculations.

[0129] The data processing module can be a host computer, a micro data processing platform with high computing power, or a data center used in high-precision printing. The data processing module needs to be equipped with an operable computer interface, an executable data model, and a communication protocol that can obtain information interaction among multiple devices.

[0130] Furthermore, in the field of three-dimensional printing, there are two main ways to move the printing platform: one is to build the object layer by layer by sinking the printing platform, and the other is to print on a fixed platform by moving the print head. For those devices that cannot realize the sinking of the printing platform in design, the present invention proposes an innovative solution, that is, to set up a motion track system inside the printing chamber to adapt to different printing needs and improve the flexibility and efficiency of the printing process.

[0131] For devices that cannot sink the printing platform, this embodiment proposes to design motion tracks in the printing chamber to support the dynamic adjustment of the scanning camera. These tracks include several vertical motion tracks and spiral ascending tracks, which together form a flexible monitoring network. The vertical motion tracks are symmetrically arranged with the printing area axis of the printing chamber as the reference, ensuring the possibility of comprehensive monitoring of the printing process from different heights and angles. This design enables the scanning camera to capture images from multiple perspectives during the printing process, providing rich data for subsequent defect detection and analysis.

[0132] The spiral ascending track is set up with the center of the print area of ​​the print chamber as the axis and arranged according to equal pitch. The design of this track allows the scanning camera to move along the spiral path, so as to achieve full monitoring of the printed object without interfering with the work of the print head. This layout is particularly suitable for complex printing tasks that require precise monitoring from multiple angles. It can ensure that every detail of the object can be carefully observed and recorded during the printing process.

[0133] Through this innovative track design, this embodiment not only improves the monitoring quality during the 3D printing process, but also provides strong technical support for the real-time discovery and prediction of printing defects. This optimized monitoring system can significantly improve printing accuracy, reduce material waste, and ultimately improve the practicality and reliability of 3D printing technology.

[0134] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A three-dimensional printing method, characterized in that: The method comprises at least the following steps: S1, scanning and modeling: obtain the geometric information of the actual object through three-dimensional scanning, and use modeling software to generate the corresponding digital body 1; S2, printing and monitoring: starting the three-dimensional printing process, and setting a scanning camera in the printing chamber to capture the information during the printing process in real time, and transmitting the captured data back to the system to generate a digital body 2 corresponding to the actual printing result; S3, comparison and defect prediction: compare digital body 1 with digital body 2, analyze the difference between them, so as to identify and predict future printing defects and their possible locations; S4, defect repair: according to the comparison result, the printing settings of the digital body 2 are adjusted to repair or optimize the defects in the printing process; S5. Correction: In the subsequent printing process, the comparison between the digital body 1 and the digital body 2 is continuously performed, and the printing settings are corrected based on the comparison results to achieve closed-loop control and continuous optimization of printing quality.

2. A three-dimensional printing method according to claim 1, characterized in that: The step S2 at least includes: S21, deploying a number of scanning cameras in the printing chamber; S22, during the printing process of the printer, the scanning camera is automatically triggered to collect images and obtain the image data of the layer; S23, the system acquires scanning camera image data to form layered model data.

3. A three-dimensional printing method according to claim 2, characterized in that: The step S21 also includes a motion track, the scanning camera is mounted on the motion track, and the motion track is a plurality of vertical motion tracks or a plurality of spiral ascending tracks; Wherein, the vertical motion track is arranged symmetrically with respect to the printing area axis of the printing chamber; Wherein, the spiral ascending track is arranged with equal pitch with the center of the printing area of ​​the printing chamber as the axis.

4. A three-dimensional printing method according to claim 2, characterized in that: The step S23 at least includes: S231, extracting feature points from image data of multiple perspectives, and generating point cloud data; S232, accurately registering and fusing point cloud data acquired from different perspectives; S233, converting the point cloud data into a mesh model, performing triangulation processing, and generating a continuous surface triangulated mesh model.

5. A three-dimensional printing method according to claim 1, characterized in that: The S3 step at least includes: S31, model alignment: spatially align digital volume 1 and digital volume 2 using spatial transformation to ensure that the two models are in the same coordinate system before comparison, eliminating errors caused by position, rotation or scaling; S32, defect classification: using a deep learning model to identify and classify defects on the adjusted digital body 2 to obtain printing defect information; S33, defect prediction: import the printing defect information and printer data information in the digital body 2 into the time series model, and predict the location and type of possible defects in the future through continuous data analysis.

6. A three-dimensional printing method according to claim 5, characterized in that: The time series model adopts an autoregressive differential moving average model or a vector autoregressive model, and establishes a prediction relationship between time and defects by using the printing defect information in the digital body 2 obtained in continuous time to obtain predicted defect information.

7. A three-dimensional printing method as claimed in claim 6, characterized in that: The steps S31 and S32 also include: Difference calculation: Determine the degree of deviation by calculating the distance error between the corresponding feature points in digital body 1 and digital body 2; Hotspot identification: Based on the calculation results, an error threshold is set and feature point differences exceeding this threshold are identified as hotspots.

8. A three-dimensional printing method according to claim 6, characterized in that: In the step S4, “repairing or optimizing defects in the printing process” includes repairing printing defects and optimizing predicted defects.

9. A printing system, characterized in that: The printing system is based on the three-dimensional printing method according to any one of claims 1 to 8, and at least comprises: A scanning module, which is used to obtain geometric information of an actual object by three-dimensional scanning, generate a corresponding digital model using modeling software, and transmit the obtained digital model information to a data processing module and / or a printing module; Printing module: The printing module is connected to the data processing module and performs two-way data interaction; Scanning camera module: The scanning camera module is arranged in the printing module, and is connected with the data processing module and / or the printing module, and performs two-way data interaction; Data processing module: The data module is used to receive data from the scanning module, printing module, and scanning camera module for processing, and generate control instructions to be sent to the corresponding modules.

Images