Porosity classification prediction method for metal additive manufacturing samples based on ultrasonic testing

Abstract

A porosity classification prediction method for metal additive manufacturing samples based on ultrasonic testing, the steps are as follows: S1, preparing porosity calibration samples; S2, testing porosity calibration samples with different porosities to obtain actual porosity, formulating porosity classification standards and classifying each porosity calibration sample; S3, S4, using ultrasonic longitudinal waves and ultrasonic shear waves to perform ultrasonic testing on each porosity calibration sample, and obtaining transverse and longitudinal wave velocities and transverse and longitudinal wave attenuation coefficients of ultrasonic testing of each sample; S5, using a PSO-BP neural network to establish a corresponding mapping relationship model between transverse and longitudinal wave velocities and transverse and longitudinal wave attenuation coefficients and porosity classification; S6, using ultrasonic transverse and longitudinal waves to perform ultrasonic testing on metal additive manufacturing samples to be tested, respectively, to obtain transverse and longitudinal wave velocities and transverse and longitudinal wave attenuation coefficients; S7, inputting the ultrasonic testing data obtained in step S6 into the corresponding mapping relationship model established in step S5, and obtaining the porosity classification of the metal additive manufacturing sample to be tested.

Description

Technical Field

[0001] The invention relates to a porosity classification prediction method for metal additive manufacturing samples based on ultrasonic testing, and belongs to the technical field of ultrasonic nondestructive testing. Background Art

[0002] 3D printing (3D Printing), also known as additive manufacturing, is a new material forming technology compared to traditional subtractive manufacturing such as turning, milling, planing, and grinding. In the actual additive manufacturing process, pores are inevitable. Porosity plays a decisive role in the mechanical properties of the product, such as tensile and fatigue, and is the most important indicator for testing product quality. Therefore, accurate and convenient measurement of the porosity of additive manufacturing samples plays a vital role in their wide application in the industrial field. There are currently four main porosity detection methods: Archimedes method obtains the total porosity of the sample based on the relative density of liquid and air. This method is simple to operate, but has low accuracy and is difficult to detect for large products; the 2D imaging method based on cross-sectional imaging observes a specific cutting plane through an optical or scanning electron microscope (SEM) to obtain the distribution and morphology of pores in the sample, but the porosity generated based on the statistics of each cross-sectional sample has a large error and requires the destruction of the sample; the X-ray-based computed tomography (CT) method models the intensity change of the ray in three dimensions, which can accurately obtain the three-dimensional spatial distribution, morphology and volume size of the pores in the sample, but this method has a small detection area, shallow depth, and is expensive and complex to operate; the ultrasonic testing (UT) method based on the interaction between internal defects and ultrasonic waves can obtain time domain diagrams and 2D / 3D images related to porosity by extracting ultrasonic characteristic parameters in the echo signal. In recent years, ultrasonic testing has become the most powerful method for characterizing the internal structure of materials due to its excellent portability, high penetration depth, non-destructive testing, and low cost of use.

[0003] However, in the existing ultrasonic porosity detection methods, due to the limitation of the wavelength of ultrasound itself, the resolution of small pores is low, and the overall porosity quantification ability is insufficient. It is necessary to combine other tools and data post-processing to improve the accuracy of ultrasonic detection. Most of the existing methods to improve the accuracy of ultrasonic porosity detection use deep learning methods, use a single longitudinal wave for testing, and need to greatly increase the amount of test data or use multiple ultrasonic feature parameters to improve the accuracy of the model. Summary of the invention

[0004] The purpose of the present invention is to provide a porosity classification and prediction method for metal additive manufacturing samples based on ultrasonic testing. The method is based on ultrasonic transverse and longitudinal wave signals of a PSO-BP neural network and can classify and predict the porosity of metal additive manufacturing samples simply, quickly, at low cost and with high accuracy.

[0005] The technical solution adopted by the present invention to achieve its invention purpose is: a porosity classification prediction method for metal additive manufacturing samples based on ultrasonic testing, comprising the following steps:

[0006] S1. Preparing porosity calibration samples: using the same additive manufacturing method as the metal additive manufacturing sample to be tested, by adjusting the process parameters, preparing porosity calibration samples with different porosities and made of the same material as the metal additive manufacturing sample to be tested;

[0007] S2. Using micro-ultrasound to test the porosity calibration samples with different porosities, obtain the actual porosity of each porosity calibration sample, formulate a porosity classification standard and classify each porosity calibration sample according to the actual porosity;

[0008] S3, using ultrasonic longitudinal waves to perform ultrasonic testing on each porosity calibration sample, collecting ultrasonic signals and performing analysis and calculation to obtain the longitudinal wave velocity and longitudinal wave attenuation coefficient of the ultrasonic testing of each porosity calibration sample;

[0009] S4, using ultrasonic shear waves to perform ultrasonic testing on each porosity calibration sample, collecting ultrasonic signals and performing analysis and calculation to obtain the shear wave velocity and shear wave attenuation coefficient of the ultrasonic testing of each porosity calibration sample;

[0010] S5. According to the classification of each porosity calibration sample in step S2 and the longitudinal wave velocity, longitudinal wave attenuation coefficient, shear wave velocity and shear wave attenuation coefficient of each porosity calibration sample ultrasonically detected obtained in steps S3 and S4, a mapping relationship model corresponding to the longitudinal wave velocity, longitudinal wave attenuation coefficient, shear wave velocity and shear wave attenuation coefficient and the porosity classification is established using a PSO-BP neural network;

[0011] S6. Use ultrasonic shear waves and ultrasonic longitudinal waves to perform ultrasonic testing on the metal additive manufacturing sample to be tested, collect ultrasonic signals and perform analysis and calculation to obtain the shear wave velocity, shear wave attenuation coefficient, longitudinal wave velocity and longitudinal wave attenuation coefficient of the ultrasonic testing of the metal additive manufacturing sample to be tested;

[0012] S7, inputting the shear wave velocity, shear wave attenuation coefficient, longitudinal wave velocity and longitudinal wave attenuation coefficient of the ultrasonic detection of the metal additive manufacturing sample to be tested into the corresponding mapping relationship model established in step S5 to obtain the porosity classification of the metal additive manufacturing sample to be tested.

[0013] Furthermore, the metal additive manufacturing specimen to be tested in the present invention is an aluminum alloy component.

[0014] Furthermore, the metal additive manufacturing sample to be tested in the present invention is an AlSi10Mg component.

[0015] Furthermore, the metal additive manufacturing sample to be tested in the present invention is an aluminum alloy component prepared by selective laser melting additive manufacturing.

[0016] Furthermore, the porosity grading standards formulated in step S2 of the present invention are as follows: Grade 1: porosity 0-1%, Grade 2: porosity 1-2%, Grade 3: porosity 2-5%, Grade 4: porosity 5-10%, Grade 5: porosity>10%.

[0017] Furthermore, the specific operation of preparing a porosity calibration sample with different porosities made of the same material as the metal additive manufacturing sample to be tested by adjusting the process parameters using the same additive manufacturing method as the metal additive manufacturing sample to be tested is:

[0018] s11. Prepare additive manufacturing raw materials of the same material as the metal additive manufacturing sample to be tested, determine the optimal process parameters through preliminary experiments, and obtain the optimal laser volume energy density. The optimal process parameters include laser power, scanning speed, scanning spacing and layer thickness of selective laser melting additive manufacturing;

[0019] s12. By adjusting the laser power and scanning speed in the optimal process parameters upward or downward to change the laser volume energy density, N porosity calibration samples with different porosities are prepared, N ≥ 15.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] Existing porosity detection technologies often use a single longitudinal wave for characterization. The accuracy of using longitudinal waves or transverse waves alone in the same model is extremely low. A large amount of detection data is required to train the model to improve the prediction accuracy. In addition, ultrasonic waveforms are often used directly during training, which inevitably introduces noise and irrelevant parameters that affect the accuracy of the model. This technical solution combines the advantages of transverse waves and longitudinal waves in porosity detection, greatly reducing the amount of data required to establish the model; the sound velocity and attenuation coefficient are extracted through the cross-correlation algorithm and peak algorithm to avoid the introduction of ultrasonic waveform noise and irrelevant variables. The mapping relationship between each ultrasonic parameter and the porosity classification is established through the particle swarm optimization (PSO) BP neural network algorithm, which improves the accuracy of the porosity classification prediction of metal additive manufacturing samples. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a schematic diagram of ultrasonic testing of a porosity calibration sample according to an embodiment of the present invention. DETAILED DESCRIPTION

[0023] Example

[0024] A porosity classification prediction method for metal additive manufacturing samples based on ultrasonic testing comprises the following steps:

[0025] S1. Preparing porosity calibration samples: using the same additive manufacturing method as the metal additive manufacturing sample to be tested, by adjusting the process parameters, preparing porosity calibration samples with different porosities and made of the same material as the metal additive manufacturing sample to be tested;

[0026] S2. Using micro-ultrasound to test the porosity calibration samples with different porosities, obtain the actual porosity of each porosity calibration sample, formulate a porosity classification standard and classify each porosity calibration sample according to the actual porosity;

[0027] S3, using ultrasonic longitudinal waves to perform ultrasonic testing on each porosity calibration sample, collecting ultrasonic signals and performing analysis and calculation to obtain the longitudinal wave velocity and longitudinal wave attenuation coefficient of the ultrasonic testing of each porosity calibration sample;

[0028] S4, using ultrasonic shear waves to perform ultrasonic testing on each porosity calibration sample, collecting ultrasonic signals and performing analysis and calculation to obtain the shear wave velocity and shear wave attenuation coefficient of the ultrasonic testing of each porosity calibration sample;

[0029] S5. According to the classification of each porosity calibration sample in step S2 and the longitudinal wave velocity, longitudinal wave attenuation coefficient, shear wave velocity and shear wave attenuation coefficient of each porosity calibration sample ultrasonically detected obtained in steps S3 and S4, a mapping relationship model corresponding to the longitudinal wave velocity, longitudinal wave attenuation coefficient, shear wave velocity and shear wave attenuation coefficient and the porosity classification is established using a PSO-BP neural network;

[0030] S6. Use ultrasonic shear waves and ultrasonic longitudinal waves to perform ultrasonic testing on the metal additive manufacturing sample to be tested, collect ultrasonic signals and perform analysis and calculation to obtain the shear wave velocity, shear wave attenuation coefficient, longitudinal wave velocity and longitudinal wave attenuation coefficient of the ultrasonic testing of the metal additive manufacturing sample to be tested;

[0031] S7, inputting the shear wave velocity, shear wave attenuation coefficient, longitudinal wave velocity and longitudinal wave attenuation coefficient of the ultrasonic detection of the metal additive manufacturing sample to be tested into the corresponding mapping relationship model established in step S5 to obtain the porosity classification of the metal additive manufacturing sample to be tested.

[0032] Preferably, the metal additive manufacturing sample to be tested is an aluminum alloy component, and more preferably, the metal additive manufacturing sample to be tested is an AlSi10Mg component.

[0033] Preferably, the metal additively manufactured specimen to be tested is an aluminum alloy component manufactured by selective laser melting additive manufacturing.

[0034] Preferably, the porosity grading standards determined in step S2 are as follows: grade 1: porosity 0-1%, grade 2: porosity 1-2%, grade 3: porosity 2-5%, grade 4: porosity 5-10%, grade 5: porosity>10%.

[0035] Preferably, the specific operation of preparing a porosity calibration sample with different porosities made of the same material as the metal additive manufacturing sample to be tested by adjusting the process parameters using the same additive manufacturing method as the metal additive manufacturing sample to be tested is:

[0036] s11. Prepare additive manufacturing raw materials of the same material as the metal additive manufacturing sample to be tested, determine the optimal process parameters through preliminary experiments, and obtain the optimal laser volume energy density. The optimal process parameters include laser power, scanning speed, scanning spacing and layer thickness of selective laser melting additive manufacturing;

[0037] s12. By adjusting the laser power and scanning speed in the optimal process parameters upward or downward to change the laser volume energy density, N porosity calibration samples with different porosities are prepared, N ≥ 15.

[0038] The following takes the selective laser melting of AlSi10Mg components as an example to illustrate the specific operation of establishing a mapping relationship model of longitudinal wave velocity, longitudinal wave attenuation coefficient, shear wave velocity and shear wave attenuation coefficient and porosity classification according to the method of this embodiment:

[0039] S1. Preparation of porosity calibration samples: Using the same additive manufacturing method as the metal additive manufacturing sample to be tested, by adjusting the process parameters, prepare porosity calibration samples with different porosities and made of the same material as the metal additive manufacturing sample to be tested. The specific operations are:

[0040] s11. Prepare additive manufacturing raw materials of the same material as the metal additive manufacturing sample to be tested, determine the optimal process parameters through preliminary experiments, and obtain the optimal laser volume energy density. The optimal process parameters include laser power, scanning speed, scanning spacing and layer thickness of selective laser melting additive manufacturing;

[0041] s12. By adjusting the laser power and scanning speed in the optimal process parameters upward and downward to change the laser volume energy density, 15 porosity calibration samples with different porosities were prepared. Table 1 shows the additive manufacturing process parameters for preparing the porosity calibration samples. Then, each porosity calibration sample was polished to meet the surface requirement of a surface roughness of less than 3.2 um.

[0042] The laser volume energy density (VED) is a relationship that describes the effect of process parameters on sample molding quality and performance. VED represents the average input energy per unit volume of material during selective laser melting, where P is laser power (W), v is scanning speed (mm / s), h is scanning spacing (μm), and t is layer thickness (μm). VED is expressed as:

[0043]

[0044] Table 1 Additive manufacturing process parameters of samples

[0045]

[0046] S2. Use micro-ultrasound to detect the porosity calibration samples with different porosities to obtain the actual porosity of each porosity calibration sample. Table 2 shows the actual porosity of each porosity calibration sample. Formulate a porosity classification standard and classify each porosity calibration sample according to the actual porosity. Table 3 shows the formulated porosity classification standard and the classification of the porosity calibration samples.

[0047] Table 2 Actual porosity of porosity calibration samples

[0048]

[0049] Table 3 Porosity calibration sample classification

[0050]

[0051] S3. Use ultrasonic longitudinal waves to perform ultrasonic testing on each porosity calibration sample, collect ultrasonic signals and perform analysis and calculation to obtain the longitudinal wave velocity and longitudinal wave attenuation coefficient of each porosity calibration sample. The ultrasonic testing uses a self-generated and self-received mode to excite a 5MHz longitudinal wave probe, and the collected ultrasonic signal is a longitudinal wave waveform w L , for the longitudinal wave w L Perform noise reduction and use the cross-correlation algorithm to calculate the longitudinal wave speed V L , using the peak algorithm to calculate the longitudinal wave attenuation coefficient α L ;

[0052] S4. Use ultrasonic shear waves to perform ultrasonic testing on each porosity calibration sample, collect ultrasonic signals and perform analysis and calculation to obtain the shear wave velocity and shear wave attenuation coefficient of each porosity calibration sample. The ultrasonic testing uses a self-transmitting and self-receiving mode to excite a 5MHz shear wave probe, and the collected ultrasonic signal is a shear wave waveform w T , for the shear wave waveform w T Perform noise reduction and use the cross-correlation algorithm to calculate the shear wave speed V T , using the peak algorithm to calculate the shear wave attenuation coefficient α T ;

[0053] Figure 1 Schematic diagram of ultrasonic testing of porosity calibration samples in this embodiment, in which 1 is an ultrasonic probe, a 5MHz longitudinal wave probe for ultrasonic longitudinal wave testing, and a 5MHz shear wave probe for ultrasonic shear wave testing, and 2 is a porosity calibration sample. In order to reduce the error, 10 groups of signals are collected in 5 different areas of each porosity calibration sample (2 times for each area), and the collection positions are shown in FIG. Figure 1 Table 4 shows the transverse and longitudinal wave sound velocities and attenuation coefficients obtained by ultrasonic testing of each porosity calibration sample. The specific calculation method of the transverse and longitudinal wave sound velocities and attenuation is as follows: the thickness of each porosity calibration sample is d. During ultrasonic testing, the ultrasonic signals emitted by the transverse and longitudinal wave probes propagate in a material with a thickness of d. The transverse and longitudinal wave sound time t in the echo signal with a travel distance of 2d is collected, and the sound velocity is calculated by the following formula:

[0054]

[0055] Collect the amplitude A of the mth and nth back wall echoes of the ultrasonic signal emitted by the transverse and longitudinal wave probes m and A n , the attenuation coefficient is calculated by the following formula:

[0056]

[0057] Table 4 Transverse and longitudinal wave velocity and attenuation coefficient of the samples

[0058]

[0059] S5. According to the classification of each porosity calibration sample in step S2 and the longitudinal wave velocity, longitudinal wave attenuation coefficient, shear wave velocity and shear wave attenuation coefficient of each porosity calibration sample ultrasonically detected obtained in steps S3 and S4, a mapping relationship model corresponding to the longitudinal wave velocity, longitudinal wave attenuation coefficient, shear wave velocity and shear wave attenuation coefficient and the porosity classification is established using a PSO-BP neural network;

[0060] s51. The specific method of optimizing BP neural network by particle swarm algorithm (PSO) is:

[0061] s52, determine the particle population size N, and set the population size to 30;

[0062] s53, set the learning factor C1=C2=4 to adjust the influence of individual particle experience and group experience on particle trajectory;

[0063] s54, set the maximum speed of particle movement to [-1.0, 1.0], adjust the intensity of spatial search to avoid falling into the local optimum or flying over the global optimum; set the maximum number of model iterations to 1000, and stop calculation when the error is less than 1*10 -5 .

[0064] In order to verify the accuracy of the porosity classification prediction of the combined transverse and longitudinal wave method of the present invention compared with the single transverse wave and the single longitudinal wave, according to the PSO-BP neural network established in step S5, the longitudinal wave sound velocity, longitudinal wave attenuation coefficient and the transverse wave, sound velocity, and transverse wave attenuation coefficient are input separately, and it can be obtained that when the longitudinal wave is used alone, the accuracy of the training set is 54%, and the accuracy of the test set is 51%; when the transverse wave is used alone, the accuracy of the training set is 64%, and the accuracy of the test set is 56%; when the longitudinal wave and transverse wave data are input simultaneously without changing the neural network parameters, the accuracy of the training set and the test set can reach 98%, as shown in Table 5 below.

[0065] Table 5 Model verification accuracy

[0066]

[0067] It can be seen from Table 5 that by combining the respective advantages of shear waves and longitudinal waves, the mapping relationship between ultrasonic parameters and porosity classification is established through the particle swarm optimization (PSO) BP neural network algorithm, which greatly improves the accuracy of porosity classification prediction of selective laser melting AlSi10Mg samples.

Claims

1. A porosity classification prediction method for metal additive manufacturing samples based on ultrasonic testing, characterized in that: The following steps are involved: S1. Preparing porosity calibration samples: using the same additive manufacturing method as the metal additive manufacturing sample to be tested, by adjusting the process parameters, preparing porosity calibration samples with different porosities and made of the same material as the metal additive manufacturing sample to be tested; S2. Using micro-ultrasound to test the porosity calibration samples with different porosities, obtain the actual porosity of each porosity calibration sample, formulate a porosity classification standard and classify each porosity calibration sample according to the actual porosity; S3, using ultrasonic longitudinal waves to perform ultrasonic testing on each porosity calibration sample, collecting ultrasonic signals and performing analysis and calculation to obtain ultrasonic longitudinal wave velocity and ultrasonic longitudinal wave attenuation coefficient of ultrasonic testing of each porosity calibration sample; S4, using ultrasonic shear waves to perform ultrasonic testing on each porosity calibration sample, collecting ultrasonic signals and performing analysis and calculation to obtain ultrasonic shear wave velocity and ultrasonic shear wave attenuation coefficient of ultrasonic testing of each porosity calibration sample; S5. According to the classification of each porosity calibration sample in step S2 and the longitudinal wave velocity, longitudinal wave attenuation coefficient, shear wave velocity and shear wave attenuation coefficient of each porosity calibration sample ultrasonically detected obtained in steps S3 and S4, a corresponding mapping relationship model between the longitudinal wave velocity, longitudinal wave attenuation coefficient, shear wave velocity and shear wave attenuation coefficient and the porosity classification is established using a PSO-BP neural network; S6. Use ultrasonic shear waves and ultrasonic longitudinal waves to perform ultrasonic testing on the metal additive manufacturing sample to be tested, collect ultrasonic signals and perform analysis and calculation to obtain the shear wave velocity, shear wave attenuation coefficient, longitudinal wave velocity and longitudinal wave attenuation coefficient of the ultrasonic testing of the metal additive manufacturing sample to be tested; S7. Input the shear wave velocity, shear wave attenuation coefficient, longitudinal wave velocity and longitudinal wave attenuation coefficient of the ultrasonic detection of the metal additive manufacturing sample to be tested into the corresponding mapping relationship model established in step S4 to obtain the porosity classification of the metal additive manufacturing sample to be tested.

2. The porosity classification prediction method of metal additive manufacturing samples based on ultrasonic testing according to claim 1 is characterized by: The metal additive manufacturing sample to be tested is an aluminum alloy component.

3. The porosity classification prediction method of metal additive manufacturing samples based on ultrasonic testing according to claim 2 is characterized by: The metal additive manufacturing sample to be tested is an AlSi10Mg component.

4. The porosity classification prediction method of metal additive manufacturing samples based on ultrasonic testing according to claim 2 is characterized by: The metal additive manufacturing sample to be tested is an aluminum alloy component prepared by selective laser melting additive manufacturing.

5. The porosity classification prediction method of metal additive manufacturing samples based on ultrasonic testing according to claim 1 is characterized by: The porosity classification standards formulated in step S2 are as follows: Level 1: porosity 0-1%, Level 2: porosity 1-2%, Level 3: porosity 2-5%, Level 4: porosity 5-10%, Level 5: porosity>10%.

6. The porosity classification prediction method of metal additive manufacturing samples based on ultrasonic testing according to claim 4 is characterized by: The specific operation of preparing the porosity calibration sample with different porosity and the same material as the metal additive manufacturing sample to be tested by adjusting the process parameters using the same additive manufacturing method as the metal additive manufacturing sample to be tested is: S11, preparing additive manufacturing raw materials of the same material as the metal additive manufacturing sample to be tested, determining optimal process parameters through preliminary experiments, and obtaining optimal laser volume energy density, wherein the optimal process parameters include laser power, scanning speed, scanning spacing and layer thickness of selective laser melting additive manufacturing; S12. By adjusting the laser power and scanning speed in the optimal process parameters upward or downward to change the laser volume energy density, N porosity calibration samples with different porosities are prepared, where N≥15.

Images