CN106182765A - 3D printer model scale error Forecasting Methodology based on support vector machine - Google Patents

Abstract

The invention provides a kind of 3D printer model scale error Forecasting Methodology based on support vector machine, step includes: 1, threedimensional model design；2, set different print parameters and print；3, model key point, Most Vital Edge position determine；4, by obtaining scale error with master pattern Least squares matching；5, the model parameter obtained and corresponding print parameters are formed data base, data base is randomly divided into two classes: training group and prediction group；6, use training group that SVM model is trained, and use prediction group that the prediction accuracy of trained SVM model is verified, filter out satisfactory SVM model；7, using new print parameters as the scale error of input prediction model.The present invention can predict the scale error scope of this parameter series drag by print parameters, is conducive to improving printing effect, reduces unnecessary spillage of material.

Description

3D printing model size error prediction method based on support vector machine

technical field

The present invention relates to the technical field of auxiliary manufacturing, in particular to a 3D printing model size error prediction method based on a Support Vector Machine (Support Vector Machine, SVM).

Background technique

The purpose of 3D printing model size prediction is to divide the printing model into two categories that meet the size error requirements, and assist the printer to consider the printing results before printing and when setting printing parameters, improve printing quality, reduce waste of consumables and time loss. So far, there are endless 3D printer products on the market, and training materials are also available. However, the control of 3D printing printing results still requires the experience of printer operators. The use of machine learning prediction algorithms based on support vector machines effectively reduces the requirements for printer operators. , which has a good meaning for enhancing the economical practicability of the 3D printer itself.

SVM is a linear classifier based on statistical learning theory. Its algorithm is a convex optimization problem, and its local optimal solution is the global optimal solution. Its characteristic is to seek the best between the complexity of the model and the generalization learning ability based on the principle of structural risk minimization with limited sample information, which can effectively avoid the shortcomings of over-learning or falling into local optimum. Taking two-dimensional data as an example, two types of data points are distributed in a two-dimensional plane. The basic principle is to find the classification counties that can separate the two types of data points through training. Although there are many such classification lines, there is one and only one dividing line that satisfies the classification line with the shortest distance to the two types of data points. For multidimensional data, the data points are distributed in multidimensional space, and the SVM classifier obtains the optimal classification hyperplane.

After searching, the Chinese invention patent application with publication number CN105643944A and application number 201610200113.5 discloses a 3D printer stability control method and control system. The 3D printer stability control method and control system of the invention construct the optimal threshold error value Based on the model, the optimal threshold error value can be selected in real time by the attenuation integral balance point of the recommended control value, which can better solve the phenomenon that the recommended control value of the 3D printer molding process undergoes balanced and unbalanced attenuation with the change of the threshold value. Improved printer stability. However, this patent focuses on improving the printing quality by solving the stability problem of the hardware part of the printer, and cannot comprehensively consider the influence of various factors in the entire printing process.

Contents of the invention

In view of the defects in the prior art, the object of the present invention is to provide a 3D printing model size error prediction method based on a support vector machine, the method uses a support vector machine to model the relationship between the printing parameters and the model size error, and the same The size error of the model under different parameter settings is predicted, and the prediction result of the possibility of the model size error exceeding the threshold value in the parameter setting is given for the reference of the printer, so as to optimize the printing quality.

In order to achieve the above object, the present invention provides a 3D printing model size error prediction method based on a support vector machine, the method includes the following steps:

The first step, three-dimensional model design, described three-dimensional model is the three-dimensional model of finished product, is used for testing the printing level of 3D printer;

The second step is to import the 3D model obtained in the first step into the 3D printer, and set different printing parameters on the 3D printer to print;

The third step is to determine the position of the key points and key edges of the 3D model and measure the point cloud data;

Inductively determine the main geometric features of the 3D model obtained in the first step, including the center of the circular hole, the radian of the arc, the length of each side and its intersection point; the determined main geometric features are measured under the measuring instrument to obtain point cloud data;

The fourth step is to obtain the size error by least squares matching with the standard model;

Perform least squares matching on the point cloud data obtained in the third step and the 3D model obtained in the first step to obtain the size error of the finished product printed by the 3D printer;

The fifth step is to form a database with the dimensional error of the finished product obtained in the fourth step and the corresponding printing parameters obtained in the second step, and divide the database into two categories according to the random grouping of finished products: training group and prediction group;

The sixth step is to use the training group to train the SVM model, and use the prediction group to verify the prediction accuracy of the trained SVM model, and select the SVM model that meets the requirements, that is, the prediction model;

The seventh step is to use the new printing parameters as input to predict the size error of the model.

Preferably, in the first step, the geometric features of the three-dimensional model include points, lines, surfaces, and holes of commonly used models, and are convenient for measurement with an optical measuring instrument or a three-dimensional probe.

Preferably, in the fourth step, the dimensional error obtained by least squares matching with the standard model refers to: matching the point cloud data obtained in the third step with the three-dimensional model of the finished product designed in the first step, to get the least squares error value.

Preferably, in the fifth step, the random grouping is multi-time grouping, and the grouping depends on the total number of data groups.

Preferably, in the fifth step, the database is a multi-dimensional array, which consists of two parts: printing parameters and label values.

Preferably, the prediction model exists as a software system independent of the 3D printer, or is built into the 3D printing parameter setting system as an algorithm module, and serves as a printing effect feedback after parameter setting to assist and guide the printer to set the parameters.

Compared with the prior art, the present invention has the following beneficial effects:

The dimensional error prediction of the present invention is based on the entire printing process, which comprehensively considers various factors that may affect the dimensional error of the printed product in the printing process, establishes a support vector machine model on this basis, and based on historical printing data. Dimensional error prediction.

The method of the invention can predict the size error range of the finished product printed by the three-dimensional model under the parameter series through the printing parameters, which is beneficial to improve the printing efficiency and reduce unnecessary material loss.

Description of drawings

Other characteristics, objects and advantages of the present invention will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings:

Fig. 1 is a flowchart of an embodiment of the present invention;

Fig. 2 is a schematic diagram of a three-dimensional model structure of an embodiment of the present invention.

detailed description

The present invention will be described in detail below in conjunction with specific embodiments. The following examples will help those skilled in the art to further understand the present invention, but do not limit the present invention in any form. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

As shown in Figure 1, a kind of 3D printing model size error prediction method based on support vector machine, described method comprises the following steps:

Step 1. Design 3D model

Described three-dimensional model is the three-dimensional model of finished product, is used for testing the printing level of 3D printer;

The 3D model needs to have the geometric features of common workpieces, including but not limited to points: lines (straight lines, curves, arcs), surfaces (curved surfaces, planes), holes (through holes, stepped holes), etc., as shown in Figure 2.

Step 2, import the 3D model obtained in step 1 into a 3D printer, and set different printing parameters on the 3D printer for printing; the printing parameters are the printing parameters of the 3D model, including: the thickness of the printing layer, the model during the printing process The placement angle of the model, the parameters of the support during the model printing process such as support height, support inclination, contact area between the support structure and the model, the density of the support structure, environmental parameters such as temperature, humidity, and material property code.

Usually, taking an SLA printer as an example, there are more than ten printing parameters that need to be set. The printer can determine the key parameter settings through his own judgment, including but not limited to: printing layer thickness, model placement angle, support The size of the contact point of the structure, the angle of the structure, etc., the environmental conditions such as temperature, humidity, etc., and the printing materials such as model 1, model 2, etc.

Step 3. Determine the position of the key points and key edges of the 3D model and measure the point cloud data;

Inductively determine the main geometric features of the 3D model obtained in step 1, including the center of the circular hole, the radian of the arc, the length of each side and its intersection point; the determined main geometric features are measured under the measuring instrument to obtain point cloud data;

The three-dimensional key points include: the starting point and end point of the line, the center of the arc, the geometric features of the hole and the protruding column, and the contour information of the model. In the specific point taking, the image measurement obtains the point cloud data through image processing, and the contact method The measurement obtains point cloud data by setting the number of points per unit area cm ² or unit length cm. Generally, five points per unit area cm ² are set in a matrix or ten points are equidistant per unit length cm.

The key edges are the edges that play a decisive role in the shape and function of the three-dimensional model parts, including the edges that form the outline of the three-dimensional model, such as straight lines and arcs, and the outline circles formed on a certain projection plane that form the key holes and columns of the three-dimensional model. Arc or line, curved edge.

For key points and key edges, the point cloud data of key points and key edges can be obtained by using contact imager measurement, and the point cloud data of key points and key edges can be obtained by using contact measurement such as probe measurement; among them, contact measurement requires Determines the number of measuring points per unit area or unit length.

Step 4, obtain the size error by performing least squares matching with the standard model;

The point cloud data obtained in step 3 is matched with the actual points, lines, and surfaces of the 3D model in step 1 to obtain the dimensional error result of the finished product printed by the 3D printer.

Step 5, form a database with the size error of the finished product obtained in step 4 and the corresponding printing parameters obtained in step 2, and divide the database into two categories according to the random grouping of finished products: training group and prediction group;

The database is a multi-dimensional array, which consists of two parts: print parameters and label values. The label value is converted from the dimensional error obtained after the measurement and matching of each set of printing parameters corresponding to the finished part, a certain dimensional error threshold is set, and the dimensional error value greater than the threshold is set to -1, less than or equal to The size error value of the threshold is set to +1, assuming that there are n printing parameters and m models are printed, the data set is a matrix of m×(n+1).

Random grouping requires multiple random groupings, and the grouping depends on the total number of data groups. Generally, three groups, five groups, etc. are selected; and the error result in step 4 is set to +1 according to a certain threshold (the size error is less than the threshold) , -1 (the size error is greater than or equal to the threshold).

Step 6, using the training group to train the SVM model, and using the prediction group to verify the prediction accuracy of the trained SVM model, and selecting the SVM model that meets the requirements, that is, the prediction model;

Different random grouping results train different SVM models, use the corresponding test group to test the SVM model, and select the SVM model with high prediction accuracy. Generally, the accuracy rate is greater than or equal to 85%.

Step 7, using the new printing parameters as the size error of the input prediction model;

Determine the model with a high prediction accuracy rate in step 6 as the prediction model, and predict the size error of the model under the new printing parameters, and determine whether it exceeds a specific size error (the size error threshold set according to the actual product quality requirements) limit.

Said method of the present invention has analyzed the factor (comprising environmental parameter, model parameter and equipment parameter: setting parameter before printer printing) that may have effect on the quality of finished product in the printing process, has proposed to comprehensively consider the whole printing process and adopt The method of mathematical statistics establishes the prediction model of the influence factor and the size error of the finished product, which can be used as the built-in software of the equipment or as a peripheral program to predict the size error of the finished product under a certain printing parameter setting in advance, and the printing quality is based on the history. The prediction of experience improves the success rate of printing finished parts.

The invention can predict the size error range of the model under the parameter series through the printing parameters, which is beneficial to improving the printing efficiency and reducing unnecessary material loss.

Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art may make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention.

Claims (8)

1. a 3D printer model scale error Forecasting Methodology based on support vector machine, it is characterised in that described method includes Following steps:

The first step, threedimensional model design, and described threedimensional model is the threedimensional model of finished parts, is used for testing 3D printer stamping ink Flat；

Second step, the threedimensional model first step obtained import 3D printer, and set different printing ginsengs on 3D printer Number prints；

3rd step, threedimensional model key point, Most Vital Edge position determine and record cloud data；

The main geometric properties of the threedimensional model first step obtained carries out conclusion and determines, including the circular hole center of circle, circular arc radian, each The length of side and intersection point thereof；The main geometric properties determined measures acquisition cloud data under measuring instrument；

4th step, by with master pattern Least squares matching obtain scale error；

The threedimensional model that 3rd step gained cloud data and the first step obtain is carried out Least squares matching, it is thus achieved that 3D printer is beaten The scale error of print finished parts；

The corresponding print parameters that 5th step, the finished parts scale error the 4th step obtained and second step obtain forms number According to storehouse, and data base is divided into two classes by finished parts random packet: training group and prediction group；

SVM model is trained, and uses the prediction group prediction to trained SVM model by the 6th step, employing training group Accuracy is verified, filters out the satisfactory i.e. forecast model of SVM model；

7th step, using new print parameters as the scale error of input prediction model.

A kind of 3D printer model scale error Forecasting Methodology based on support vector machine the most according to claim 1, it is special Levying and be, in the first step, the geometric properties of described threedimensional model comprises the point, line, surface of common model, hole, and is easy to use Optical measuring instrument or three-dimensional probe measurement.

A kind of 3D printer model scale error Forecasting Methodology based on support vector machine the most according to claim 1, it is special Levying and be, in second step, described print parameters is the print parameters of threedimensional model, including: printing thickness, model is printed Angles in journey, the parameter such as bearing height supported in model print procedure, support gradient, supporting construction and model Contact area, the density of supporting construction, ambient parameter such as temperature, humidity, material properties code name.

A kind of 3D printer model scale error Forecasting Methodology based on support vector machine the most according to claim 1, it is special Levying and be, in the 3rd step, described three-dimensional key point, including: the starting point of line and terminal, the center of circle of circular arc, hole and prominent post Geometric properties, and the profile information of model, in specifically taking a little, radiographic measurement obtains cloud data, contact by image procossing Measure by setting unit are cm²Or the quantity that takes in unit length cm obtains cloud data, typically set unit are cm²Take a little five or unit length cm by matrix-style and equidistantly take a little ten；

Described Most Vital Edge, is the limit playing a decisive role threedimensional model External Shape and function, including composition threedimensional model The limit of contours profiles, composition threedimensional model closes keyhole, the profile circular arc formed on a certain perspective plane of post or straight line, curved side.

A kind of 3D printer model scale error Forecasting Methodology based on support vector machine the most according to claim 1, it is special Levying and be, in the 5th step, described random packet is for be repeatedly grouped, and packet is according to depending on data set sum.

6. according to a kind of based on support vector machine the 3D printer model scale error prediction side described in any one of claim 1-5 Method, it is characterised in that in the 5th step, described data base is a Multidimensional numerical, and it is made up of two parts: print parameters and mark Label value.

A kind of 3D printer model scale error Forecasting Methodology based on support vector machine the most according to claim 6, it is special Levy and be, what described label value obtained after being measured coupling by each group of print parameters correspondence finished parts scale error conversion and Coming, set a scale error threshold value, the scale error value that will be greater than this threshold value is set as-1, misses less than or equal to the size of this threshold value Difference is set as+1, it is assumed that has n print parameters, and have printed m model, then this data set is the matrix of m × (n+1).

8. according to a kind of based on support vector machine the 3D printer model scale error prediction side described in any one of claim 1-5 Method, it is characterised in that described forecast model exists independent of 3D printer as software system, or as in algoritic module Put 3D print parameters and arrange in system, the printing effect feedback after arranging as parameter, auxiliary direction printing person's setup parameter.