Method for predicting mechanical properties of metal 3D printed components based on SVR and OpenCV

Abstract

The application discloses a metal 3D printing component mechanical property prediction method based on SVR and OpenCV, and comprises the following steps: S1, a multi-level multi-factor orthogonal test scheme is designed and a sample is prepared; S2, an electronic universal testing machine is used to test the mechanical properties of samples in each group in the orthogonal test; S3, an electron microscope is used to obtain a metallographic picture of the middle region of the sample; S4, a prediction model between the additive manufacturing process parameters and the mechanical properties is built; S5, two important parameters in the support vector machine algorithm are optimized by using a grey wolf algorithm (GWO); S6, an optimal process parameter combination of the additive manufacturing process is obtained; S7, a grain size range distribution of each metallographic picture is obtained; S8, a support vector machine prediction model of higher-dimensional input parameters is re-established; S9, the prediction value of the re-established GWO-SVR model is compared with the actual value, the prediction accuracy of the model is verified, the optimal process parameter combination obtained by the traversal algorithm is used for additive manufacturing process sampling, and the test verification is re-performed.

Description

Technical Field

[0001] This invention belongs to the field of metal 3D printing technology, specifically involving a method for predicting the mechanical properties of metal 3D printed components based on SVR and OpenCV. Background Technology

[0002] In recent years, with the continuous growth of the aerospace and defense industries, production demand will continue to increase. The rapid development and technological innovation of high-end manufacturing are also gradually increasing the requirements for the geometric complexity of structures. Traditional production methods are difficult to process complex structures. Therefore, additive manufacturing (AM) technology, which enables the integrated forming of complex parts, has gradually come to the forefront. It can manufacture high-precision, complex part shapes, and its applications are very wide-ranging, including aerospace, medical, automotive, and chemical industries.

[0003] Aerospace components are generally characterized by complex structures, customized products, and the inability to conduct destructive testing. For metal 3D printing materials, tensile strength refers to the maximum load-bearing capacity of the material under tensile load, and is a crucial indicator of its tensile properties. In industrial applications, high-strength metal parts are often required to meet extreme working conditions, such as in chemical, medical, and aerospace fields, where high strength is essential. Yield strength refers to the stress value at which a material begins to undergo plastic deformation under external force. Elongation refers to the increase in length of a material when it undergoes plastic deformation under external force. The yield strength and elongation of 3D printed materials are important indicators of their performance. For aerospace components, obtaining accurate mechanical properties through experimental and simulation analysis allows for a better understanding of the part's characteristics and application value, especially for irregularly shaped metal 3D printed aerospace components, where obtaining their mechanical properties is of great significance.

[0004] When manufacturing metal parts using additive manufacturing technology, their mechanical properties are influenced by various factors, such as material grain size and manufacturing process parameters. Therefore, in manufacturing 3D printed parts, it is necessary to control the parameters of the additive manufacturing process to obtain better mechanical properties. Furthermore, the Holpage relation characterizes the relationship between the lower yield point of a material and its grain size; that is, within the micrometer scale, the smaller the grain size, the higher the material strength. This invention can predict the mechanical properties of irregularly shaped 3D printed metal components. Specifically, under the same combination of process parameters, by acquiring metallographic images of the measured location, the mechanical property parameters can be quickly obtained through an algorithmic model.

[0005] The patent (application number 202210572726.7) establishes the relationship between the indentation profile snapshot in a spherical indentation and the material constitutive parameters based on an artificial neural network. It obtains the hyperparameters of the artificial neural network using a simulated annealing algorithm and utilizes single-objective programming to obtain the plastic tensile properties parameters of the tested metallic material. However, this method has a low prediction accuracy, with a tensile strength prediction accuracy of only 90.17%. Furthermore, the input parameters of the prediction model in this patent fail to fully characterize the main factors influencing mechanical properties. Due to the limitations of the prediction method, it is not suitable for parts with complex mechanical mechanisms.

[0006] The patent (application number: 202111261550.5) considers two strengthening mechanisms in metallic materials—dislocation strengthening and grain refinement strengthening—based on a crystal plastic constitutive framework. It establishes a crystal plastic constitutive model that considers both dislocation strengthening and grain refinement strengthening, thereby predicting the mechanical properties of metallic materials. The biggest problem with this method is its high experimental cost. It requires multiple observations and measurements using scanning electron microscopy, electron backscattering microscopy, and optical microscopy, incurring significant manpower and time costs. Furthermore, the patent does not provide specific examples illustrating the prediction accuracy, nor does it compare the method with other methods, failing to demonstrate or verify its reliability. Summary of the Invention

[0007] The purpose of this invention is to solve the above-mentioned problems and provide a method for predicting the mechanical properties of printed parts, including tensile strength, yield strength, and elongation, by using additive manufacturing process parameters and the grain size distribution of the printed part. The prediction model established by the process parameters and mechanical properties can determine the optimal combination of process parameters for the part through an ergonomic algorithm, guiding the selection of processing parameters for the part. This is a method for predicting the mechanical properties of metal 3D printed components based on SVR and OpenCV.

[0008] To solve the above-mentioned technical problems, the technical solution of the present invention is: a method for predicting the mechanical properties of metal 3D printed components based on SVR and OpenCV, comprising the following steps:

[0009] S1. Select key parameters of additive manufacturing process, design multi-level and multi-factor orthogonal experimental scheme and prepare samples;

[0010] S2. Use an electronic universal testing machine to test the mechanical properties of each group of specimens in the orthogonal test, including yield strength, tensile strength and elongation.

[0011] S3. Obtain metallographic images of the central region of the sample using an electron microscope;

[0012] S4. Based on the support vector machine regression algorithm (SVR), a predictive model between additive manufacturing process parameters and mechanical properties is built.

[0013] S5. Use the Grey Wolf Algorithm (GWO) to optimize two important parameters in the Support Vector Machine algorithm: the penalty function c and the radial basis function parameter g.

[0014] S6. Establish a GWO-SVR prediction model, optimize the parameters by taking the mean square error as the single parameter optimization objective, compare the predicted value of the prediction model with the actual value to verify the prediction accuracy of the model, and obtain the optimal combination of process parameters for additive manufacturing process through traversal algorithm.

[0015] S7. Based on OpenCV image processing technology, preprocess, segment and statistically analyze the metallographic images of each group of orthogonal test specimens to obtain the grain size range distribution of each metallographic image.

[0016] S8. Add the grain size range distribution as an input parameter to the input variables of the GWO-SVR prediction model, and re-establish a support vector machine prediction model with higher-dimensional input parameters.

[0017] S9. Compare the predicted values ​​of the re-established GWO-SVR model with the actual values ​​to verify the model's prediction accuracy. Use the optimal combination of process parameters obtained through the traversal algorithm to prepare additive manufacturing process samples and conduct experimental verification again.

[0018] Furthermore, the multi-level, multi-factor process in step S1 includes key parameters in the SLM process, laser power, scanning speed, and scanning spacing.

[0019] Furthermore, before using an electron microscope in step S3, the middle area of ​​the SLM sample is selected and inlaid with AB glue. The sample is then finely ground and polished using a polishing machine. After cleaning with an alcohol reagent, an etchant is placed on the sample surface for etching. The etchant is aqua regia, which is a mixture of concentrated nitric acid and concentrated hydrochloric acid in a 1:3 ratio.

[0020] Furthermore, the metallographic images in step S3 are obtained by taking pictures under an electron microscope at 500x magnification.

[0021] Furthermore, in step S5, the upper and lower bounds of the penalty function c and the radial basis function parameter g are 0.01 and 100, respectively.

[0022] Further, in step S6, the prediction model is rapidly calculated using an ergonomic algorithm to obtain the optimal combination of process parameters: laser power of 325W, scanning speed of 892mm / s, and scanning spacing of 0.16mm. This yields the optimal mechanical properties, predicting a yield strength of 570MPa, a tensile strength of 651MPa, and an elongation of 10.42%. The sample manufactured using the optimal process combination is subjected to a tensile test using an electronic universal testing machine, repeated three times. The final actual mechanical properties are a yield strength of 562MPa, a tensile strength of 643MPa, and an elongation of 9.88%.

[0023] Furthermore, in step S7, based on OpenCV image processing technology, the GWO-SVR model has a higher prediction accuracy for tensile strength, from the original 92.32% to the current 98.26%, achieving high-precision prediction of mechanical properties under SLM process.

[0024] Furthermore, the OpenCV image processing technology in step S7, through secondary development, enables the creation of a Python program capable of segmenting and statistically analyzing metallographic images, including the following sub-steps:

[0025] S71. Add the OpenCV code library to call different image processing functions, such as filtering, dilation, and erosion;

[0026] S72. Adjust each parameter in each function to its optimal value;

[0027] S73. First, import the cv2 image processing development library, use the cv2.blur function for filtering, the unsharp_mask function for sharpening, the cv2.erode function for erosion, the cv2.dilate function for dilation, and the color.label2rgb function for coloring to complete all image preprocessing and image segmentation operations.

[0028] S74. Then use the measure.regionprops function to calculate the range of different grain sizes.

[0029] The beneficial effects of this invention are:

[0030] 1. The mechanical property prediction method for metal 3D printed components based on SVR and OpenCV provided by this invention has the advantages of high prediction accuracy, ability to predict irregularly shaped components, and low test cost.

[0031] 2. This invention further optimizes key parameters of the support vector machine model using the Grey Wolf algorithm and compares them. OpenCV image processing technology is used for image segmentation and parameter statistics, replacing costly experiments such as EBSD, to obtain the required parameter set. Existing mathematical formulas or models cannot characterize the relationship between process parameters, grain size distribution, and mechanical properties. Therefore, this invention establishes a support vector machine prediction model optimized using the Grey Wolf algorithm, building a model of the relationship between 3D printing process parameters, grain size distribution, and mechanical properties, providing a scientific method for preparing 3D printed workpieces with high mechanical properties.

[0032] 3. This invention studies the relationship between process parameters, grain size and mechanical properties. Compared with the prior art, this invention increases the number of inputs to the prediction model. The input variables are characterized by corresponding parameters from both macroscopic and microscopic perspectives. The input parameters are not just a single tensile strength, but also include indicators that characterize the mechanical properties of the sample, namely yield strength, tensile strength and elongation, as output variables of the model.

[0033] 4. This paper uses OpenCV-based image processing technology, writes Python scripts and debugs the parameters of each image processing function, making it easy to encapsulate the code in the system or for secondary development.

[0034] 5. This invention explores the relationship between the microstructure and mechanical properties of metal parts, and builds a high-precision mechanical property prediction model with multiple inputs and multiple outputs by using the GWO-SVR algorithm and OpenCV image processing technology.

[0035] 6. This invention has high scalability and is not limited to a single metal material, making it applicable to various additive manufacturing processes. Experimental verification shows that the GWO-SVR model has high robustness and prediction accuracy, and can quickly calculate the optimal combination of process parameters. It provides process guidance for the fabrication of 3D printed metal parts with high mechanical properties. Attached Figure Description

[0036] Figure 1 This is the overall technical roadmap of the method for predicting the mechanical properties of metal 3D printed components based on SVR and OpenCV in this invention;

[0037] Figure 2 This is a schematic diagram of the mechanical tensile specimen dimensions of the present invention;

[0038] Figure 3 This is the technical roadmap for the GWO-SVR prediction model of this invention;

[0039] Figure 4 This is a comparison chart of the predicted and actual values ​​of the GWO-SVR algorithm model of this invention;

[0040] Figure 5 This is a demonstration diagram of the metallographic pretreatment process of the present invention;

[0041] Figure 6 This is a statistical chart of the metallographic size range corresponding to the nine sets of orthogonal experiments of this invention;

[0042] Figure 7 This is a comparison chart of the predicted and actual values ​​output by the GWO-SVR algorithm model rebuilt after adding grain size distribution as an input variable in this invention. Detailed Implementation

[0043] The present invention will be further described below with reference to the accompanying drawings and specific embodiments:

[0044] like Figures 1 to 7 As shown, the method for predicting the mechanical properties of metal 3D printed components based on SVR and OpenCV provided by the present invention is characterized by comprising the following steps:

[0045] S1. Select key parameters of additive manufacturing process, design multi-level and multi-factor orthogonal experimental scheme and prepare samples.

[0046] The multi-level, multi-factor process in step S1 includes key parameters of the SLM process, laser power, scanning speed, and scanning spacing. This embodiment uses 316L stainless steel and selective laser melting (SLM) as an example.

[0047] This invention uses 316L stainless steel powder with a particle size of 15–53 μm, and its powder composition is shown in Table 1. Key parameters in the SLM process, namely laser power, scanning speed, and scanning spacing, are selected as horizontal parameters to design an orthogonal experimental scheme. The dimensions of the tensile specimens are shown in Table 1. Figure 2 As shown.

[0048] Table 1. Composition of 316L stainless steel powder (mass fraction, %)

[0049] 0.043 0.73 1.45 0.02 0.004 16.8 12.6 2.6 margin

[0050] S2. Use an electronic universal testing machine to test the mechanical properties of each group of specimens in the orthogonal test, including yield strength, tensile strength and elongation.

[0051] Tensile tests were conducted using an electronic universal testing machine. Nine orthogonal experimental schemes were used to test their mechanical properties, including yield strength, tensile strength, and elongation. The orthogonal experimental results are shown in Table 2. As can be seen from the table, the mechanical properties of the ninth group were generally better.

[0052] Table 2 Orthogonal Experiment Table

[0053]

[0054] S3. Obtain metallographic images of the central region of the sample using an electron microscope.

[0055] Before using the electron microscope in step S3, select the middle area of ​​the SLM sample and mount it with AB glue. Then, use a polishing machine to finely grind and polish the sample. After cleaning with alcohol reagent, place the etchant on the sample surface for etching. The etchant used is aqua regia, which is a mixture of concentrated nitric acid and concentrated hydrochloric acid in a 1:3 ratio.

[0056] After cleaning, the samples were examined using an electron microscope to observe the metallographic images of the central region of nine sets of orthogonal test metal samples. Each image was taken under an electron microscope at 500x magnification.

[0057] S4. Based on the support vector machine regression algorithm (SVR), a predictive model is built between additive manufacturing process parameters and mechanical properties.

[0058] The technical approach of using MATLAB software to build a GWO-SVR prediction model and optimizing the support vector machine model using the Grey Wolf algorithm is as follows: Figure 3 As shown, the technical approach involves importing seven sets of orthogonal experimental data into the model for training.

[0059] S5. Use the Grey Wolf (GWO) algorithm to optimize two important parameters in the support vector machine algorithm: the penalty function c and the radial basis function parameter g.

[0060] In step S5, the upper and lower bounds of the penalty function c and the radial basis function parameter g are 0.01 and 100, respectively.

[0061] S6. Establish a GWO-SVR prediction model, optimize the parameters by taking the mean square error as the single parameter optimization objective, compare the predicted value of the prediction model with the actual value to verify the prediction accuracy of the model, and obtain the optimal combination of process parameters for additive manufacturing process through traversal algorithm.

[0062] The wolf pack size was set at 50, and the error value of tensile strength was used as the accuracy criterion for the prediction model. The comparison between the model's predicted values ​​and actual values ​​is as follows: Figure 4 As shown in the figure. After calculation, the tensile strength calculated by the GWO-SVR model achieved an accuracy rate of 92.93%. The statistical errors of the predictions for other mechanical property parameters are shown in Table 3. Comparing the calculated mechanical properties with the actual values ​​when the two combinations of process parameters were substituted into the model, the error rates were 92.33% and 93.26%, respectively, with an average accuracy rate of 92.80%. Figure 4 It can be seen that the model has good prediction accuracy, and the fluctuation trend of the predicted data is the same as that of the actual value.

[0063] In step S6, the prediction model is quickly calculated using an ergonomic algorithm to obtain the optimal combination of process parameters: laser power of 325W, scanning speed of 892mm / s, and scanning spacing of 0.16mm. This yields the optimal mechanical properties, predicting a yield strength of 570MPa, a tensile strength of 651MPa, and an elongation of 10.42%. The sample manufactured using the optimal process combination is subjected to a tensile test using an electronic universal testing machine, repeated three times. The final actual mechanical properties are a yield strength of 562MPa, a tensile strength of 643MPa, and an elongation of 9.88%.

[0064] Table 3 Analysis of GWO-SVR Model Prediction Results

[0065]

[0066] S7. Based on OpenCV image processing technology, preprocess, segment and statistically analyze the metallographic images of each group of orthogonal test specimens to obtain the grain size range distribution of each metallographic image.

[0067] In step S7, based on OpenCV image processing technology, the GWO-SVR model achieves a higher prediction accuracy for tensile strength, increasing from 92.32% to 98.26%, thus realizing high-precision prediction of mechanical properties under SLM technology. Figure 5 As shown, Figure 5 In the diagram, (a) represents grayscale, (b) represents filtering, (c) represents sharpening, (d) represents erosion, (e) represents dilation, and (f) represents shading.

[0068] Step S7 involves OpenCV image processing technology, followed by the development of a Python program capable of segmenting and statistically analyzing metallographic images. This includes the following sub-steps:

[0069] S71. Add the OpenCV code library to call different image processing functions, such as filtering, dilation, and erosion.

[0070] S72. Adjust each parameter in each function to its optimal value.

[0071] S73. First, import the cv2 image processing development library, use the cv2.blur function for filtering, the unsharp_mask function for sharpening, the cv2.erode function for erosion, the cv2.dilate function for dilation, and the color.label2rgb function for coloring to complete all image preprocessing and image segmentation operations.

[0072] S74. Then use the measure.regionprops function to calculate the range of different grain sizes.

[0073] In this embodiment, the statistical distribution of grain size in the nine orthogonal experimental groups is as follows: Figure 6 As shown.

[0074] S8. Add the grain size range distribution as an input parameter to the input variables of the GWO-SVR prediction model, and re-establish a support vector machine prediction model with higher-dimensional input parameters.

[0075] In this embodiment, the optimized model becomes a 6-input, 3-output prediction model. Experiments were then conducted again using the new combination of process parameters, and the results are as follows: Figure 7 As shown, based on OpenCV image processing technology, the GWO-SVR model has a higher prediction accuracy for tensile strength, from the original 92.32% to the current 98.26%, achieving high-precision prediction of mechanical properties under SLM process.

[0076] S9. Compare the predicted values ​​of the re-established GWO-SVR model with the actual values ​​to verify the model's prediction accuracy. Use the optimal combination of process parameters obtained through the traversal algorithm to prepare additive manufacturing process samples and conduct experimental verification again.

[0077] This invention comprehensively considers the coupling effect between different process parameters, and can reliably and accurately predict the mechanical properties of metal samples prepared by SLM, thereby further optimizing the process parameters of selective laser melting and effectively shortening the process development cycle of products.

[0078] The process of this invention is as follows: First, a multi-level, multi-factor orthogonal experiment is designed. Mechanical property tests are then conducted on the workpieces from each group of the orthogonal experiments to obtain the corresponding yield strength, tensile strength, and elongation. The orthogonal experiment data is used as the dataset for a support vector machine (SVM) prediction model. Metallographic experiments are then conducted on workpieces processed under the same combination of process parameters to obtain metallographic images of the central region of the workpiece. The metallographic images from each group of the orthogonal experiments are used as an image set, and image segmentation and statistical analysis are performed using the OpenCV library written in Python. The grain size distribution statistically derived from each metallographic image is added as an input parameter to the SVM prediction model, achieving high-precision prediction of the mechanical properties of 3D-printed metal workpieces.

[0079] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. A method for predicting the mechanical properties of metal 3D printed components based on SVR and OpenCV, characterized in that, Includes the following steps: S1. Select key parameters of additive manufacturing process, design multi-level and multi-factor orthogonal experimental scheme and prepare samples; The multi-level and multi-factor process in step S1 includes key parameters in the SLM process, including laser power, scanning speed and scanning spacing. S2. Use an electronic universal testing machine to test the mechanical properties of each group of specimens in the orthogonal test, including yield strength, tensile strength and elongation. S3. Obtain metallographic images of the central region of the sample using an electron microscope; S4. Based on the support vector machine regression algorithm (SVR), a predictive model is built between additive manufacturing process parameters and mechanical properties; S5. Use the Grey Wolf (GWO) algorithm to optimize two important parameters in the support vector machine algorithm: the penalty function c and the radial basis function parameter g. S6. Establish a GWO-SVR prediction model, optimize the parameters by taking the mean square error as the single parameter optimization objective, compare the predicted value of the prediction model with the actual value to verify the prediction accuracy of the model, and obtain the optimal combination of process parameters for additive manufacturing process through traversal algorithm. S7. Based on OpenCV image processing technology, preprocess, segment and statistically analyze the metallographic images of each group of orthogonal test specimens to obtain the grain size range distribution of each metallographic image. S8. Add the grain size range distribution as an input parameter to the input variables of the GWO-SVR prediction model, and re-establish a support vector machine prediction model with higher-dimensional input parameters. S9. Compare the predicted values ​​with the actual values ​​of the re-established GWO-SVR model to verify the prediction accuracy of the model. Use the optimal combination of process parameters obtained through the traversal algorithm to prepare additive manufacturing process samples and conduct experimental verification again.

2. The method for predicting the mechanical properties of metal 3D printed components based on SVR and OpenCV according to claim 1, characterized in that: Before using an electron microscope in step S3, the middle area of ​​the SLM sample is selected and inlaid with AB glue. The sample is then finely ground and polished using a polishing machine. After cleaning with alcohol reagent, an etchant is placed on the sample surface for etching. The etchant used is aqua regia, which is a mixture of concentrated nitric acid and concentrated hydrochloric acid in a 1:3 ratio.

3. The method for predicting the mechanical properties of metal 3D printed components based on SVR and OpenCV according to claim 1, characterized in that: The metallographic images in step S3 are obtained by taking pictures under a 500x electron microscope.

4. The method for predicting the mechanical properties of metal 3D printed components based on SVR and OpenCV according to claim 1, characterized in that: In step S5, the upper and lower bounds of the penalty function c and the radial basis function parameter g are 0.01 and 100, respectively.

5. The method for predicting the mechanical properties of metal 3D printed components based on SVR and OpenCV according to claim 1, characterized in that: In step S6, the prediction model is rapidly calculated using an traversal algorithm to obtain the optimal combination of process parameters: laser power of 325W, scanning speed of 892mm / s, and scanning spacing of 0.16mm. This yields the optimal mechanical properties, predicting a yield strength of 570MPa, a tensile strength of 651MPa, and an elongation of 10.42%. The sample manufactured using the optimal process combination is subjected to a tensile test using an electronic universal testing machine, repeated three times. The final actual mechanical properties are a yield strength of 562MPa, a tensile strength of 643MPa, and an elongation of 9.88%.

6. The method for predicting the mechanical properties of metal 3D printed components based on SVR and OpenCV according to claim 1, characterized in that: The OpenCV image processing technology in step S7 involves secondary development of a Python program capable of segmenting and statistically analyzing metallographic images, including the following sub-steps: S71. Add the OpenCV code library to call different image processing functions, including filtering, dilation, and erosion; S72. Adjust each parameter in each function to its optimal value; S73. First, import the cv2 image processing development library, use the cv2.blur function for filtering, the unsharp_mask function for sharpening, the cv2.erode function for erosion, the cv2.dilate function for dilation, and the color.label2rgb function for coloring to complete all image preprocessing and image segmentation operations. S74. Then use the measure.regionprops function to calculate the range of different grain sizes.

Images