CN112172128A - Method for rapidly optimizing polylactic acid fused deposition molding process based on dual-index orthogonal test combined with support vector machine - Google Patents

Abstract

The invention relates to a method for quickly optimizing polylactic acid fused deposition molding process based on a double-index orthogonal test and a support vector machine, which comprises the following steps: determining process parameters, constructing a standard spline model and adjusting the process parameters, printing test splines by using an FDM printer and carrying out mechanical property test on the test splines, combining an orthogonal test with a support vector machine, establishing a printing product process parameter-mechanical property model, generating a large number of virtual samples, carrying out high-throughput screening, and rapidly forecasting the performance of a printing product. The method simply and quickly optimizes the parameters of the polylactic acid fused deposition molding process by using a machine learning modeling method, and avoids the problem that the optimal molding process parameter combination is difficult to obtain only by blind trial and error based on experience; based on reliable experimental data and a modeling method, the established fused deposition modeling process parameter-mechanical property forecasting model has the advantages of simplicity, convenience, rapidness, low cost and no pollution.

Description

Method for rapidly optimizing polylactic acid fused deposition molding process based on dual-index orthogonal test combined with support vector machine

Technical Field

The invention relates to the field of process optimization of fused deposition molded products, in particular to a method for rapidly optimizing a polylactic acid fused deposition molding process based on a dual-index orthogonal test and a support vector machine.

Technical Field

Additive Manufacturing (AM) is an intelligent manufacturing technology that integrates a plurality of subjects such as electronics, software, machinery and materials. It has the advantages of rapid prototype design, high fidelity, reduced material consumption, etc. Fused Deposition Modeling (FDM) is one of the mainstream additive manufacturing technologies, and has been widely applied to the fields of biomedical science, aerospace, national defense and military, instrument manufacturing, and the like. The current factors restricting the development of fused deposition modeling technology are not only modeling equipment and modeling materials, but also the influence of fused deposition modeling process parameters on the quality of printed products is a significant factor. Research shows that even if the same forming consumable material is used by the same FDM printer and printing is carried out under different forming process conditions, the mechanical property and the surface quality of the obtained product have obvious difference.

In recent years, research on parameters of the FDM forming process mainly focuses on the influence of parameters such as the lamination thickness, the bottom plate temperature, the nozzle temperature, the deposition mode, the printing speed and the like on single mechanical strength, and when the parameter change is ignored, synchronous change of various mechanical properties is ignored, and research on comprehensive mechanical properties is less. The multi-index orthogonal test is an important analysis method for researching multi-factor influence tests, can efficiently and economically find the influence of each factor, and more comprehensively and preferably selects the optimal test combination of comprehensive performance.

Polylactic acid (PLA) is taken as a mainstream consumable of FDM forming technology, has the advantages of high strength, high elastic modulus, good biocompatibility and the like, and the quality and the performance of a printed product of the PLA are widely concerned.

A Support Vector Machine (SVM) is a new machine learning method established by mathematician vladimirn, vapnik and the like on the basis of Statistical Learning Theory (SLT), and includes a Support Vector Classification (SVC) algorithm and a Support Vector Regression (SVR) algorithm. The method has a solid theoretical foundation, better analyzes the problems of 'overfitting' and 'underfitting', and provides a corresponding solution. The method provides a rich kernel function method, is particularly suitable for data modeling under the condition of a small sample set, and can improve the reliability of a forecast result to the maximum extent.

Disclosure of Invention

The invention aims to overcome the blindness of a trial-and-error method of an experiment, and provides a method for quickly optimizing a polylactic acid fused deposition molding process based on a double-index orthogonal test combined with a support vector machine.

The purpose of the invention is realized by the following technical scheme:

a method for rapidly optimizing a polylactic acid fused deposition molding process based on a dual-index orthogonal test and a support vector machine comprises the following steps:

1) determining technological parameters:

determining technological parameters and horizontal numbers of fused deposition molding, and selecting a proper orthogonal test table;

2) constructing a standard spline model and adjusting process parameters:

designing a standard spline model by adopting a stretching and impacting spline standard and through CAD software; exporting the constructed mechanical property test spline model as a stl file, slicing by using Cura software, and adjusting process parameters according to an orthogonal test scheme;

3) and (3) printing a test sample strip by using an FDM printer and testing the mechanical property of the test sample strip:

after the parameters are adjusted, importing the set model data into an FDM printer for printing; after all test sample strips are printed, respectively testing the mechanical properties according to the standard, and recording the test results;

4) the orthogonal test is combined with a support vector machine to establish an FDM printing product process parameter-mechanical property model:

converting the process parameter data according to a formula, importing the converted process parameter data and the orthogonal test result into modeling software, and preparing for machine learning modeling; taking root mean square error and correlation coefficient obtained by cross validation of each model by one-out-of-one method as evaluation indexes, and establishing a polylactic acid fused deposition molding product process parameter-mechanical property model by using a support vector machine;

5) generating a large number of virtual samples, carrying out high-throughput screening, and rapidly forecasting the performance of a printed product:

a large number of virtual samples are automatically generated through programming, the virtual samples are put into a constructed relation model of process parameters and mechanical properties, high-throughput screening is carried out, and the mechanical properties of the FDM printed product under the process conditions are rapidly predicted.

In the step 2), the independent variables used for modeling are variables obtained by converting four process parameters through formula calculation, and the target variables are the tensile strength and the impact strength of the FDM printed product.

In the step 4), the polylactic acid fused deposition modeling product process parameters and mechanical property data for data conversion are obtained by designing a dual-index orthogonal test.

In the step 5), the high-throughput screening comprises the following steps:

5-1) the design conditions are as follows:

the range of the layering thickness is 0-0.4 mm, and the step length of the layering thickness is not more than 0.05 mm; the printing speed range is 30-90 mm/s, and the printing speed step length is not more than 5 mm/s; the printing temperature range is 190-220 ℃, and the printing temperature step length is not more than 5 ℃; the filling angle range is 0-45 degrees, the filling angle step is not more than 5 degrees, and a large number of virtual samples are generated by utilizing python programming;

5-2) putting a large number of generated virtual samples into the constructed prediction model, and outputting mechanical property prediction values under different process parameters;

5-3) selecting the process parameter combination with higher index value from the forecast result, and printing by utilizing the optimized process parameter combination as a high-throughput screening result.

Compared with the prior art, the invention has the following obvious outstanding characteristics and obvious advantages:

1. the method simply and quickly optimizes the parameters of the polylactic acid fused deposition molding process by using a machine learning modeling method, thereby avoiding the problem that the optimal molding process parameter combination is difficult to obtain only by blind trial and error based on experience;

2. the method of the invention utilizes the double-index orthogonal test and the support vector machine to quickly predict the mechanical properties of the polylactic acid fused deposition molding product under different process conditions, the required test times are less, and the cost is low;

3. according to the method, the mechanical property of the FDM printed product printed under different process parameters is predicted in advance through model prediction, samples meeting requirements are selected for experimental verification, the efficiency of experiment is improved, a guiding effect is achieved, and the blindness of the experiment of a trial-and-error method is avoided.

Description of the drawings:

fig. 1 is a block diagram of the FDM printed article forming process optimization routine of the present invention.

FIG. 2 is a leave-one-out cross-validation result of the regression model of the tensile strength support vector machine of the present invention.

FIG. 3 is a leave-one-out cross-validation result of the regression model of the impact strength support vector machine of the present invention.

FIG. 4 shows the results of high throughput screening of samples for tensile strength and impact strength under different process conditions.

The specific implementation mode is as follows:

the present invention will be described in detail below with reference to the accompanying drawings and examples.

The first embodiment is as follows:

referring to fig. 1, a method for rapidly optimizing a polylactic acid fused deposition modeling process based on a dual-index orthogonal test combined with a support vector machine comprises the following steps:

1) determining technological parameters:

determining technological parameters and horizontal numbers of fused deposition molding, and selecting a proper orthogonal test table;

2) constructing a standard spline model and adjusting process parameters:

designing a standard spline model by adopting a stretching and impacting spline standard and through CAD software; exporting the constructed mechanical property test spline model as a stl file, slicing by using Cura software, and adjusting process parameters according to an orthogonal test scheme;

3) and (3) printing a test sample strip by using an FDM printer and testing the mechanical property of the test sample strip:

after the parameters are adjusted, guiding the set spline model into an FDM printer for printing; after all test sample strips are printed, respectively testing the mechanical properties according to the standard, and recording the test results;

4) the orthogonal test is combined with a support vector machine to establish an FDM printing product process model:

converting the process parameter data according to a formula, importing the converted process parameter data and the orthogonal test result into modeling software, and preparing for machine learning modeling; taking root mean square error and correlation coefficient obtained by cross validation of each model by one-out-of-one method as evaluation indexes, and establishing a polylactic acid fused deposition molding product process parameter-mechanical property model by using a support vector machine;

5) generating a large number of virtual samples, carrying out high-throughput screening, and rapidly forecasting the performance of a printed product:

a large number of virtual samples are automatically generated through programming, the virtual samples are put into a constructed relation model of process parameters and mechanical properties, high-throughput screening is carried out, and the mechanical properties of the FDM printed product under the process conditions are rapidly predicted.

The method of the embodiment simply and quickly optimizes the parameters of the polylactic acid fused deposition molding process by using a machine learning modeling method, thereby avoiding the problem that the optimal molding process parameter combination is difficult to obtain only by blind trial and error based on experience.

The method of the embodiment utilizes the double-index orthogonal test and the support vector machine to quickly predict the mechanical properties of the polylactic acid fused deposition molding product under different process conditions, and has the advantages of less test times and low cost.

Example two:

this embodiment is substantially the same as the first embodiment, and is characterized in that:

referring to fig. 1, in said step 2), the independent variables used for modeling are variables after the four process parameters are formulaically converted, and the target variables are the tensile strength and impact strength of the FDM printed article.

In the step 4), the polylactic acid fused deposition modeling product process parameters and mechanical property data for data conversion are obtained by designing a dual-index orthogonal test.

In the step 5), the high-throughput screening comprises the following steps:

5-1) the design conditions are as follows:

the range of the layering thickness is 0-0.4 mm, and the step length of the layering thickness is not more than 0.05 mm; the printing speed range is 30-90 mm/s, and the printing speed step length is not more than 5 mm/s; the printing temperature range is 190-220 ℃, and the printing temperature step length is not more than 5 ℃; the filling angle range is 0-45 degrees, the filling angle step is not more than 5 degrees, and a large number of virtual samples are generated by utilizing python programming;

5-2) putting a large number of generated virtual samples into the constructed prediction model, and outputting mechanical property prediction values under different process parameters;

5-3) selecting the process parameter combination with higher index value from the forecast result, and printing by utilizing the optimized process parameter combination as a high-throughput screening result.

According to the embodiment, a large number of virtual samples are automatically generated through programming, the samples are put into a constructed relation model of process parameters and mechanical properties for high-throughput screening, and the mechanical properties of the printed product under the process conditions are rapidly predicted. The embodiment is based on reliable experimental data and a modeling method, and the established fused deposition modeling process parameter-mechanical property forecasting model has the advantages of simplicity, convenience, rapidness, low cost, no pollution and the like.

Example three:

this embodiment is substantially the same as the above embodiment, and is characterized in that:

in this embodiment, the method for rapidly optimizing the poly lactic acid fused deposition modeling process based on the orthogonal test and the support vector machine includes the following steps:

(1) in the optimization process of the fused deposition modeling process, a plurality of factors need to be examined, the included level of each factor is more than one, and if all the factor levels are matched with each other for testing, the number of required tests is extremely large, and a large amount of manpower and material resources are needed. Analytical statistical studies were performed on the results obtained. The orthogonal test can perform balanced sampling in a factor change interval, and a better test condition is found through a smaller number of test times, so that each test has certain representativeness and is more efficient and economical. Therefore, in the embodiment, the process parameters and the level number are determined according to the prior knowledge and the limitation of the existing equipment conditions, a proper orthogonal test table is selected, the factor level table is shown in table 1, and a four-factor four-level orthogonal test table is selected;

TABLE 1 factor horizon

Level of	Thickness per mm of the layer	Printing speed/(mm/s)	Nozzle temperature/. degree.C	Filling angle/°
					1	0.1	30	190	45
2	0.2	50	200	30
					3	0.3	70	210	15
4	0.4	90	220	0

(2) As the mechanical property is used as an evaluation index, the forming process parameters of FDM printing are optimized, so that the mechanical property test standard of molding and extrusion molding is referred. Designing a standard spline model by adopting GB/T1040.2-2006 tensile spline standard and GB/T9341-2000 impact spline standard through CAD software;

(3) exporting the constructed mechanical property test spline three-dimensional model as stl file, slicing by using Cura software and adjusting process parameters according to an orthogonal test scheme, wherein part of the test scheme is shown in Table 2;

TABLE 2 partial orthogonal test parameter setup protocol

Thickness per mm of the layer	Printing speed/(mm/s)	Nozzle temperature/. degree.C	Filling angle-o
				0.1	30	190	45
0.1	50	200	30
				0.1	70	210	15
0.1	90	220	0
				0.2	30	200	15
0.2	50	190	0
				0.2	70	220	45
0.2	90	210	30

(4) After the parameters are adjusted, guiding the set spline model into an FDM printer for printing;

(5) after all test sample strips are printed, respectively testing the tensile strength and the notch impact strength according to GB/T1040.2-2006 and GB/T9341-2000 standards, and recording test results, wherein partial orthogonal test results are shown in Table 3;

TABLE 3 results of partial orthogonality test

(6) Converting the process parameter data according to a formula and the mechanical property value, wherein the partial data after the process parameter conversion is shown in a table 4;

6-1) for tensile strength, independent variable conversion is performed according to the following formula:

6-1-1) X1 ═ 5.020[ layer thickness ] -0.01027[ print speed ] +0.02890[ nozzle temperature ] +0.04068[ fill angle ] -4.969

6-1-2) X2 ═ 0.3523[ layered thickness ] -0.004580[ printing speed ] +0.07710[ nozzle temperature ] -0.02447[ filling angle ] -15.068

6-1-3) X3 ═ 4.093[ layer thickness ] -0.02656[ print speed ] +0.01938[ nozzle temperature ] +0.03336[ fill angle ] -2.107

6-1-4) X4 ═ 12.383[ layer thickness ] -0.02422[ printing speed ] +0.06698[ nozzle temperature ] +0.09386[ filling angle ] -11.294

6-2) for impact strength, independent variable conversion is performed according to the following formula:

6-2-1) X1 ═ 7.413[ layer thickness ] +0.009036[ printing speed ] -0.002103[ nozzle temperature ] +0.02727[ filling angle ] -2.578

6-2-2) X2 ═ 1.980[ layered thickness ] -0.01492[ printing speed ] -0.01036[ nozzle temperature ] -0.01549[ filling angle ] +2.872

6-2-3) X3 ═ 7.517[ layered thickness ] -0.008698[ printing speed ] -0.01657[ nozzle temperature ] -0.02289[ filling angle ] +6.314

6-2-4) X4 ═ 32.573[ split thickness ] -0.03536[ print speed ] +0.01070[ nozzle temperature ] -0.1165[ fill angle ] +10.6926-1-2) X2 ═ 0.3523[ split thickness ] -0.004580[ print speed ] +0.07710[ nozzle temperature ] -0.02447[ fill angle ] -15.068.

TABLE 4 partial data after conversion of Process parameters

X1	X2	X3	X4
				1.542758	-1.622485	1.870475	3.69112
1.016285	-0.5760595	1.032795	2.468641
				0.4898113	0.4703658	0.1951152	1.246161
-0.03666221	1.516791	-0.6425649	0.02368099
				0.1094869	-0.08219375	0.6542746	0.306843

(7) Importing the converted process parameter data and the orthogonal test result into Expminer software developed in a laboratory, and preparing for machine learning modeling;

(8) respectively taking the tensile strength and the impact strength obtained by an orthogonal test as target variables, taking the variables obtained after the conversion of the four process parameters as independent variables, and establishing a relation model between the polylactic acid fused deposition molding process parameters and the mechanical properties by using a support vector machine;

(9) high-throughput screening: according to the two built prediction models of the mechanical properties and the generated process parameter virtual samples, rapidly optimizing the polylactic acid fused deposition modeling process parameters;

the high-throughput screening comprises the following specific steps:

9-1) the design conditions are as follows:

the range of the layering thickness is 0-0.4 mm, and the step length of the layering thickness is 0.05 mm; the printing speed range is 30-90 mm/s, and the printing speed step length is 5 mm/s; the printing temperature range is 190-220 ℃, and the printing temperature step length is 5 ℃; the filling angle range is 0-45 degrees, the filling angle step is 5 degrees, and a large number of virtual samples are generated by utilizing python programming;

9-2) putting a large number of generated virtual samples into the constructed prediction model, and outputting mechanical property prediction values under different process parameters;

9-3) selecting the process parameter combination with higher index value from the forecast result, and printing by utilizing the optimized process parameter combination as a high-throughput screening result.

Compared with the traditional trial-and-error experiment, the method has better directivity, and can greatly save the cost and time consumption when optimizing materials.

Example four:

this embodiment is substantially the same as the above embodiment, and is characterized in that:

in the present embodiment, the modeling result of the FDM printed product tensile strength quantitative prediction model established based on 16 fused deposition modeling orthogonal test data in combination with the support vector machine is shown in fig. 2.

In the embodiment, a support vector machine regression algorithm is adopted, variables obtained by converting four process parameters in a data set through a formula are used as independent variables, a linear kernel function is used as a modeling kernel function, and a relation model between the fused deposition modeling process parameters and the tensile strength is constructed. The model is obtained by the established leave-one-out modeling model: the predicted tensile strength correlated with the actual tensile strength found experimentally, R, to 0.77 with a root mean square error RMSE of 1.959.

Example five:

this embodiment is substantially the same as the above embodiment, and is characterized in that:

in the present embodiment, the modeling result of the FDM printed product impact strength quantitative prediction model established based on 16 fused deposition modeling orthogonal test data in combination with the support vector machine is shown in fig. 3.

In the embodiment, a support vector machine regression algorithm is adopted, variables obtained by converting four process parameters in a data set through a formula are used as independent variables, and a linear kernel function is used as a modeling kernel function, so that a relation model between fused deposition modeling process parameters and impact strength is constructed. The model is obtained by the established leave-one-out modeling model: the correlation coefficient R between the predicted impact strength and the experimentally measured impact strength reaches 0.90, and the root mean square error RMSE is 0.523.

Example six:

this embodiment is substantially the same as the above embodiment, and is characterized in that:

in this embodiment, based on two mechanical property models that have been constructed and a large number of virtual samples generated, high-throughput screening is performed, and the screening results of some samples are shown in fig. 4.

In this embodiment, 7280 generated virtual samples are put into the established model, the corresponding tensile strength and impact strength are predicted, and the process parameter combination with higher tensile strength and impact strength is selected to guide the experiment.

To sum up, the embodiment of the invention is based on a dual-index orthogonal test combined with a support vector machine to rapidly optimize the polylactic acid fused deposition molding process, and the method comprises the following steps: determining technological parameters: determining technological parameters and horizontal numbers of fused deposition molding, and selecting a proper orthogonal test table; constructing a standard spline model and adjusting process parameters: designing a standard spline model by adopting national stretching and impact spline standards and through CAD software; exporting the constructed mechanical property test spline model as a stl file, slicing by using Cura software and adjusting process parameters according to an orthogonal test scheme; and (3) printing a test sample strip by using an FDM printer and testing the mechanical property of the test sample strip: after the parameters are adjusted, guiding the set spline model into an FDM printer for printing; after all test sample strips are printed, respectively testing the mechanical properties according to the national standard, and recording the test results; and (3) establishing a product process model by combining an orthogonal test with a support vector machine: converting the process parameter data according to a formula, importing the converted process parameter data and the orthogonal test result into Expminer software developed in a laboratory, and preparing for machine learning modeling; taking root mean square error and correlation coefficient obtained by cross validation of each model by one-out-of-one method as evaluation indexes, and establishing a polylactic acid fused deposition molding product process parameter-mechanical property model by using a support vector machine; generating a large number of virtual samples, carrying out high-throughput screening, and rapidly forecasting the performance of a printed product: a large number of virtual samples are automatically generated through programming, the samples are put into a constructed relation model of process parameters and mechanical properties for high-throughput screening, and the mechanical properties of the printed product under the process conditions are quickly predicted. The embodiment is based on reliable experimental data and a modeling method, and the established fused deposition modeling process parameter-mechanical property forecasting model has the advantages of simplicity, convenience, rapidness, low cost and no pollution.

The embodiments of the present invention have been described with reference to the accompanying drawings, but the present invention is not limited to the embodiments, and various changes and modifications can be made according to the purpose of the invention, and any changes, modifications, substitutions, combinations or simplifications made according to the spirit and principle of the technical solution of the present invention shall be equivalent substitutions, as long as the purpose of the present invention is met, and the present invention shall fall within the protection scope of the present invention without departing from the technical principle and inventive concept of the present invention.

Claims (4)

1. A method for rapidly optimizing a polylactic acid fused deposition molding process based on a double-index orthogonal test and a support vector machine is characterized by comprising the following steps:

1) determining technological parameters:

determining technological parameters and horizontal numbers of fused deposition molding, and selecting a proper orthogonal test table;

2) constructing a standard spline model and adjusting process parameters:

designing a standard spline model by adopting a stretching and impacting spline standard and through CAD software; exporting the constructed mechanical property test spline model as a stl file, slicing by using Cura software, and adjusting process parameters according to an orthogonal test scheme;

3) and (3) printing a test sample strip by using an FDM printer and testing the mechanical property of the test sample strip:

after the parameters are adjusted, importing the set model data into an FDM printer for printing; after all test sample strips are printed, respectively testing the mechanical properties according to the standard, and recording the test results;

4) the orthogonal test is combined with a support vector machine to establish an FDM printing product process parameter-mechanical property model:

converting the process parameter data according to a formula, importing the converted process parameter data and the orthogonal test result into modeling software, and preparing for machine learning modeling; taking root mean square error and correlation coefficient obtained by cross validation of each model by one-out-of-one method as evaluation indexes, and establishing a polylactic acid fused deposition molding product process parameter-mechanical property model by using a support vector machine;

5) generating a large number of virtual samples, carrying out high-throughput screening, and rapidly forecasting the performance of a printed product:

a large number of virtual samples are automatically generated through programming, the virtual samples are put into a constructed relation model of process parameters and mechanical properties, high-throughput screening is carried out, and the mechanical properties of the FDM printed product under the process conditions are rapidly predicted.

2. The method for rapidly optimizing the poly (lactic acid) fused deposition modeling process based on the dual-index orthogonal test combined with the support vector machine as claimed in claim 1, wherein in the step 2), the independent variables used for modeling are variables obtained by converting four process parameters through formula calculation, and the target variables are the tensile strength and the impact strength of the FDM printed product.

3. The method for rapidly optimizing the poly (lactic acid) fused deposition modeling process based on the dual-index orthogonal test combined with the support vector machine as claimed in claim 1, wherein in the step 4), the process parameters and the mechanical property data of the poly (lactic acid) fused deposition modeling product for data transformation are obtained by designing the dual-index orthogonal test.

4. The method for rapidly optimizing the poly (lactic acid) fused deposition modeling process based on the dual-index orthogonal test combined with the support vector machine as claimed in claim 1, wherein in the step 5), the high-throughput screening comprises the following steps:

5-1) the design conditions are as follows:

the range of the layering thickness is 0-0.4 mm, and the step length of the layering thickness is not more than 0.05 mm; the printing speed range is 30-90 mm/s, and the printing speed step length is not more than 5 mm/s; the printing temperature range is 190-220 ℃, and the printing temperature step length is not more than 5 ℃; the filling angle range is 0-45 degrees, the filling angle step is not more than 5 degrees, and a large number of virtual samples are generated by utilizing python programming;

5-2) putting a large number of generated virtual samples into the constructed prediction model, and outputting mechanical property prediction values under different process parameters;

5-3) selecting the process parameter combination with higher index value from the forecast result, and printing by utilizing the optimized process parameter combination as a high-throughput screening result.

Images