Aluminum alloy component detection method and system based on LIBS and PLS

Abstract

The invention discloses an aluminum alloy component detection method and system based on LIBS and PLS, and belongs to the technical field of metal material component analysis. The method comprises the following steps: 1, carrying out deoxidation flattening treatment on an aluminum alloy sample, and focusing on the surface of the aluminum alloy sample by using an aluminum alloy component detection system to obtain an LIBS spectrum of the aluminum alloy sample; 2, carrying out abnormal spectrum elimination on the LIBS spectrum of the aluminum alloy sample through a Pauta method, then carrying out background interference elimination and spectrum normalization by adopting a window translation minimum value method and sub-channel background intensity normalization, and extracting a characteristic peak of a normalized spectrum strongly related to the components of the aluminum alloy sample; 3, using PLS to construct and train an aluminum alloy component prediction model; and 4, inputting the characteristic peak of the normalized spectrum of the aluminum alloy sample into the trained aluminum alloy component prediction model to obtain a component prediction result of the aluminum alloy sample. Compared with the prior art, non-destructive rapid high-precision second-level detection of aluminum alloy components is realized.

Description

Technical Field

[0001] This invention relates to a method and system for detecting the composition of aluminum alloys based on LIBS and PLS, belonging to the field of metal material composition analysis technology, and is applicable to the detection of aluminum alloy matrix and alloy element content in industrial and laboratory settings. Background Technology

[0002] Aluminum alloys are widely used in aerospace, automotive, and construction industries due to their low density, high strength, and corrosion resistance. Their composition, such as Cu content affecting strength and Mg content affecting corrosion resistance, directly determines the material's performance and applicable scenarios.

[0003] Existing methods for aluminum alloy composition testing have the following drawbacks: 1. Long testing cycle: Traditional methods require pretreatment such as digestion and dilution of samples, with the entire process taking 1-4 hours, which cannot meet the needs of rapid quality control or immediate analysis; 2. Destructive to samples: Some methods (such as chemical analysis) require a large amount of sample (5-10g), and the processed samples cannot be reused, making them unsuitable for valuable samples (such as failure analysis samples or R&D samples); 3. Limited application scenarios: Large instruments such as ICP-MS can only operate stably in the laboratory and cannot be adapted to the dusty and vibration environments of industrial sites; traditional LIBS testing is mostly for modeling single aluminum alloy series, with poor adaptability to different series of aluminum alloys, and the error rate often exceeds 5%; 4. Complex operation: Professional personnel are required for sample pretreatment and instrument operation, resulting in high labor costs and making it difficult to achieve automated testing.

[0004] Therefore, how to achieve non-destructive, rapid, and high-precision second-level detection of aluminum alloy composition in industrial and laboratory settings has become an urgent problem to be solved. Summary of the Invention

[0005] The purpose of this invention is to address the technical problems of long cycle time, destructive nature, limited application scenarios, and poor adaptability in existing aluminum alloy composition detection methods. It proposes an aluminum alloy composition detection method and system based on LIBS and PLS. By using laser-induced breakdown spectroscopy (LIBS) to obtain the emission spectrum of aluminum alloys and combining it with partial least squares (PLS) modeling and prediction, non-destructive, high-precision, second-level detection of different series of aluminum alloy compositions can be achieved in industrial and laboratory settings.

[0006] The objective of this invention is achieved through the following technical solution:

[0007] This invention discloses a method for detecting the composition of aluminum alloys based on LIBS and PLS, comprising the following steps:

[0008] Step 1: The aluminum alloy sample is deoxidized and leveled, and the LIBS spectrum of the aluminum alloy sample is obtained by focusing on the surface of the aluminum alloy sample using an aluminum alloy composition detection system.

[0009] Step 1.1: Remove the oxide layer from the aluminum alloy sample and smooth the sample surface;

[0010] Step 1.2: Use an aluminum alloy composition detection system to emit nanosecond pulsed laser light and focus it on the surface of the aluminum alloy sample to generate plasma, thereby obtaining the LIBS spectrum;

[0011] Step 2: After removing anomalous spectra from the LIBS spectra of the aluminum alloy samples using the Laida method, background interference was removed and the spectra were normalized using the window shift minimum method and channel background intensity normalization. Characteristic peaks of the normalized spectra that are strongly correlated with the composition of the aluminum alloy samples were extracted.

[0012] Step 2.1: Obtain the average spectral intensity and standard deviation of the LIBS spectrum, use the Laida method to remove abnormal spectra from the LIBS spectrum, retain the LIBS spectra that satisfy the equation (1), remove the mean from the retained spectra, and obtain the average spectrum.

[0013] [Mean spectral intensity ± 3*standard deviation](1)

[0014] Step 2.2: Remove background interference from the background curves obtained using the window shift minimum method in the corresponding channels of the average spectrum by channel-specific processing;

[0015] Step 2.2.1: Use the window-shifting minimum value method to obtain the background minimum value of the retained spectrum, and perform polynomial fitting on the extracted background minimum value to form a background curve;

[0016] Step 2.2.2: Subtract the background curve of the corresponding channel from the average spectrum to remove background interference by channel;

[0017] Step 2.3: Normalize the spectrum after removing background interference using the channel-specific background intensity normalization method;

[0018] Step 2.3.1: Integrate the background curve intensity for each wavelength channel;

[0019] Step 2.3.2: Divide the average spectral intensity by the integral value of the background curve intensity to obtain the normalized spectrum;

[0020] Step 2.4: Extract the characteristic peaks of the normalized spectrum of the aluminum alloy sample using the characteristic spectral line library of common elements in aluminum alloys;

[0021] Step 3: Construct and train an aluminum alloy composition prediction model using PLS;

[0022] Step 3.1: Divide the aluminum alloy samples into training, validation, and test sets according to their proportions;

[0023] Step 3.2: Use the characteristic peaks of the normalized spectrum of the aluminum alloy sample as the independent variable; use the true value of the composition of the aluminum alloy as the dependent variable;

[0024] Step 3.3: Construct an aluminum alloy composition prediction model using the multiple linear regression relationship of PLS ​​between independent and dependent variables;

[0025] Step 3.4: Train the aluminum alloy composition prediction model;

[0026] Step 3.4.1: Determine the number of principal components in the aluminum alloy composition prediction model using 10-fold cross-validation;

[0027] Step 3.4.2: Set the number of iterations and obtain the minimum value of the root mean square error of the validation set in a cyclic iterative manner until the number of iterations is reached, and obtain the trained aluminum alloy composition prediction model.

[0028] Step 4: Input the characteristic peaks of the normalized spectrum of the aluminum alloy sample into the trained aluminum alloy composition prediction model to obtain the composition prediction results of the aluminum alloy sample;

[0029] This invention discloses an aluminum alloy composition detection system based on LIBS and PLS, used to implement the above-mentioned method. The rapid aluminum alloy composition detection system based on LIBS and PLS of this invention includes a LIBS device and a computing host.

[0030] The LIBS device is used to acquire laser-induced plasma spectra of aluminum alloy samples and unknown aluminum alloy samples to be tested; it obtains LIBS spectra of aluminum alloy samples with wavelengths ranging from 180 to 950 nm and resolutions not greater than 0.05 nm by focusing a 1064 nm nanosecond pulsed laser with an energy range of 0-200 mJ onto the sample surface; and it uses the LIBS spectra of the aluminum alloy samples as input to the computing host.

[0031] Furthermore, the repetition frequency of the LIBS device is 5-15Hz;

[0032] The computing host is used to obtain the composition prediction results of the aluminum alloy sample;

[0033] The optical path of the LIBS device in this system is as follows: a 1064nm nanosecond pulsed laser emitted by a laser source passes through a dichroic mirror, is focused by a first focusing lens onto the surface of an aluminum alloy sample to generate plasma emission, is then collected by the first focusing lens, reflected by the dichroic mirror, and collected by a second focusing lens into a 1-to-4 fiber optic cable to be transmitted to a spectrometer to obtain the LIBS spectrum of the aluminum alloy sample.

[0034] Compared with existing technologies, it has the following beneficial effects:

[0035] 1. Fast detection speed: The total detection time for unknown samples is ≤3 seconds, which is more than 1000 times more efficient than traditional methods (1-4 hours), and can meet the real-time quality control needs of the production line;

[0036] 2. Non-destructive testing: Only the sample surface needs to be polished, without digestion or destruction of the sample. The sample consumption per test is ≤1mg, which is suitable for testing precious samples.

[0037] 3. High detection accuracy: PLS model prediction has an absolute error of ≤0.3% for Al, ≤0.05% for Fe, ≤0.005% for Cu, and ≤0.01% for Si. It is also based on multi-series aluminum alloy samples and has good adaptability to 1-series, 5-series, 6-series, and 7-series aluminum alloys.

[0038] 4. Low cost and easy operation: No chemical reagents are required, reducing consumable costs; after modeling is completed, the detection of unknown samples can be automated, without the need for professional personnel to operate the entire process, reducing labor costs. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0040] Figure 2 This is a schematic diagram of the LIBS device of the present invention;

[0041] Figure 3 This is a rendering of the PLS model. Detailed Implementation

[0042] To better illustrate the purpose and advantages of this invention, the invention will be further described below with reference to the accompanying drawings and examples. It should be noted that the implementation of this invention is not limited to the following embodiments, and any modifications or alterations made to this invention will fall within the scope of protection of this invention.

[0043] Example

[0044] like Figure 1 As shown in the figure, the specific implementation steps of the aluminum alloy composition detection method based on LIBS and PLS in this embodiment are as follows:

[0045] Step 1: The aluminum alloy sample is deoxidized and leveled, and the LIBS spectrum of the aluminum alloy sample is obtained by focusing on the surface of the aluminum alloy sample using an aluminum alloy composition detection system.

[0046] Step 1.1: Remove the oxide layer from the aluminum alloy sample and smooth the sample surface;

[0047] In this embodiment, the aluminum alloy samples were sanded until the surface was free of oxide layer and smooth to remove the interference of the surface oxide film on the spectrum. The sample thickness was not limited. In terms of the selection of sample quantity and type, 80 aluminum alloy samples were selected, which were cylindrical in shape with a diameter of 50 mm and a thickness of 5 mm, covering I-series (pure aluminum, 10 samples), 5-series (Al-Mg, 20 samples), 6-series (Al-Mg-Si, 30 samples), and 7-series (Al-Cu-Zn, 20 samples). The surface of each sample was sanded with 400# sandpaper and wiped with anhydrous ethanol to remove dust.

[0048] Step 1.2: Use an aluminum alloy composition detection system to emit nanosecond pulsed laser light and focus it on the surface of the aluminum alloy sample to generate plasma, thereby obtaining the LIBS spectrum;

[0049] Step 2: After removing anomalous spectra from the LIBS spectra of the aluminum alloy samples using the Laida method, background interference was removed and the spectra were normalized using the window shift minimum method and channel background intensity normalization. Characteristic peaks of the normalized spectra that are strongly correlated with the composition of the aluminum alloy samples were extracted.

[0050] Step 2.1: Obtain the average spectral intensity and standard deviation of the LIBS spectrum, use the Laida method to remove abnormal spectra from the LIBS spectrum, retain the LIBS spectra that satisfy the equation (1), remove the mean from the retained spectra, and obtain the average spectrum.

[0051] [Mean spectral intensity ± 3*standard deviation](1)

[0052] Step 2.2: Remove background interference from the background curves obtained using the window shift minimum method in the corresponding channels of the average spectrum by channel-specific processing;

[0053] Step 2.2.1: Use the window-shifting minimum value method to obtain the background minimum value of the retained spectrum, and perform polynomial fitting on the extracted background minimum value to form a background curve;

[0054] Step 2.2.2: Subtract the background curve of the corresponding channel from the average spectrum to remove background interference by channel;

[0055] Step 2.3: Normalize the spectrum after removing background interference using the channel-specific background intensity normalization method;

[0056] Step 2.3.1: Integrate the background curve intensity for each wavelength channel;

[0057] Step 2.3.2: Divide the average spectral intensity by the integral value of the background curve intensity to obtain the normalized spectrum;

[0058] Step 2.4: Extract the characteristic peaks of the normalized spectrum of the aluminum alloy sample using the characteristic spectral line library of common elements in aluminum alloys;

[0059] In this embodiment, the anomaly removal specifically involves: for each sample, collecting n≥30 spectra, calculating the average spectral intensity and standard deviation, and retaining valid spectra within the range of [average spectral intensity ± 3 * standard deviation]; the channel-specific background subtraction uses the window-shifting minimum value method, performing 3-5 order polynomial fitting on the extracted minimum value to subtract the background; the channel-specific normalization uses "channel background intensity integral normalization", that is, dividing the channel spectral intensity by the integral value of the channel background intensity; the number of 50 characteristic peaks extracted is 3 for Al, 3 for Cu, 1 for Mg, 1 for Si, 1 for Zn, 1 for Mn, and 5 for Fe.

[0060] Step 3: Construct and train an aluminum alloy composition prediction model using PLS;

[0061] Step 3.1: Divide the aluminum alloy samples into training, validation, and test sets according to their proportions;

[0062] Step 3.2: Use the characteristic peaks of the normalized spectrum of the aluminum alloy sample as the independent variable; use the true value of the composition of the aluminum alloy as the dependent variable;

[0063] In the examples, the true values ​​of the components were determined by ICP-MS (GB / T 20975.25-2020), with the following true value ranges: Cu 0.05%-6.0%, Mg 0.2%-3.5%, Si 0.1%-1.5%, and Al 92%-99.8%.

[0064] Step 3.3: Construct an aluminum alloy composition prediction model using the multiple linear regression relationship of PLS ​​between independent and dependent variables;

[0065] Step 3.4: Train the aluminum alloy composition prediction model;

[0066] Step 3.4.1: Determine the number of principal components in the aluminum alloy composition prediction model using 10-fold cross-validation;

[0067] Step 3.4.2: Set the number of iterations and obtain the minimum value of the root mean square error of the validation set in a cyclic iterative manner until the number of iterations is reached, and obtain the trained aluminum alloy composition prediction model.

[0068] In this embodiment, the training set (56 samples), validation set (16 samples), and test set (8 samples) are divided in a 7:2:1 ratio. For sample selection, several aluminum alloy samples covering different series are chosen (e.g., 1-series pure aluminum, 5-series Al-Mg, 6-series Al-Mg-Si, 7-series Al-Cu-Zn) to ensure the samples cover the concentration range of the target components (e.g., Cu content 0.1%-5%, Mg content 0.5%-3%). For data partitioning, the data is proportionally divided into a training set for model building, a validation set for parameter adjustment, and a test set for model accuracy verification. For model building, PLS is used to establish the regression relationship between independent and dependent variables to construct the component prediction model. For model optimization, 10-fold cross-validation is used to determine the number of principal components to be 8 (the training set is divided into 10 parts, with 9 parts used for training and 1 part for validation alternately). Simultaneously, after 60 iterations, the model parameters are adjusted in each iteration, such as... Figure 3 As shown, the model with the smallest root mean square error (RMSE) on the validation set was ultimately selected as the optimal prediction model. Validation results on the test set: absolute error for Cu content prediction was 0.037%, for Mg content 0.123%, for Si content 0.0077%, and for Al content 0.121%.

[0069] Step 4: Input the characteristic peaks of the normalized spectrum of the aluminum alloy sample into the trained aluminum alloy composition prediction model to obtain the composition prediction results of the aluminum alloy sample;

[0070] In this embodiment, 30 spectra were collected from 10 aluminum alloy samples not included in the modeling (2 from the I-series, 2 from the V-series, 3 from the VI-series, and 3 from the VII-series) (taking 2 seconds), preprocessed (0.5 seconds), and substituted into the optimal model for calculation (0.3 seconds), for a total time of 2.8 seconds; the predicted aluminum alloy composition was obtained, including the mass fraction of the Al matrix and the mass fractions of Cu, Mg, Si, Zn, Mn, Fe, and other trace elements;

[0071] This invention discloses an aluminum alloy composition detection system based on LIBS and PLS, used to implement the above-mentioned method. The rapid aluminum alloy composition detection system based on LIBS and PLS of this invention includes a LIBS device and a computing host.

[0072] The LIBS device is used to acquire laser-induced plasma spectra of aluminum alloy samples and unknown aluminum alloy samples to be tested; it obtains LIBS spectra of aluminum alloy samples with wavelengths ranging from 180 to 950 nm and resolutions not greater than 0.05 nm by focusing a 1064 nm nanosecond pulsed laser with an energy range of 0-200 mJ onto the sample surface; and it uses the LIBS spectra of the aluminum alloy samples as input to the computing host.

[0073] Furthermore, the repetition frequency of the LIBS device is 5-15Hz;

[0074] The computing host is used to obtain the composition prediction results of the aluminum alloy sample;

[0075] The optical path of the LIBS device in this system is as follows: a 1064nm nanosecond pulsed laser emitted by a laser source passes through a dichroic mirror, is focused by a first focusing lens onto the surface of an aluminum alloy sample to generate plasma emission, is then collected by the first focusing lens, reflected by the dichroic mirror, and collected by a second focusing lens into a 1-to-4 fiber optic cable to be transmitted to a spectrometer to obtain the LIBS spectrum of the aluminum alloy sample.

[0076] In the embodiments, such as Figure 2 As shown, the nanosecond pulse laser parameters of the LIBS device are: wavelength 1064nm, pulse energy adjustable, maximum 200mJ, repetition frequency 5-10Hz; spectral acquisition wavelength range 180-950nm, resolution ≤0.05nm.

Claims

1. A method for detecting the composition of aluminum alloys based on LIBS and PLS, characterized in that: Includes the following steps, Step 1: The aluminum alloy sample is deoxidized and leveled, and the LIBS spectrum of the aluminum alloy sample is obtained by focusing on the surface of the aluminum alloy sample using an aluminum alloy composition detection system. Step 2: After removing anomalous spectra from the LIBS spectra of the aluminum alloy samples using the Laida method, background interference was removed and the spectra were normalized using the window shift minimum method and channel background intensity normalization. Characteristic peaks of the normalized spectra that are strongly correlated with the composition of the aluminum alloy samples were extracted. Step 3: Construct and train an aluminum alloy composition prediction model using PLS; Step 3.1: Divide the aluminum alloy samples into training, validation, and test sets according to their proportions; Step 3.2: Use the characteristic peaks of the normalized spectrum of the aluminum alloy sample as the independent variable; use the true value of the composition of the aluminum alloy as the dependent variable; Step 3.3: Construct an aluminum alloy composition prediction model using the multiple linear regression relationship of PLS ​​between independent and dependent variables; Step 3.4: Train the aluminum alloy composition prediction model; Step 3.4.1: Determine the number of principal components in the aluminum alloy composition prediction model using 10-fold cross-validation; Step 3.4.2: Set the number of iterations and obtain the minimum value of the root mean square error of the validation set in a cyclic iterative manner until the number of iterations is reached, and obtain the trained aluminum alloy composition prediction model. Step 4: Input the characteristic peaks of the normalized spectrum of the aluminum alloy sample into the trained aluminum alloy composition prediction model to obtain the composition prediction results of the aluminum alloy sample.

2. The method for detecting the composition of aluminum alloys based on LIBS and PLS as described in claim 1, characterized in that: Step 1 is implemented as follows: Step 1.1: Remove the oxide layer from the aluminum alloy sample and smooth the sample surface; Step 1.2: Use an aluminum alloy composition detection system to emit nanosecond pulsed lasers and focus them on the surface of the aluminum alloy sample to generate plasma, thereby obtaining the LIBS spectrum.

3. The method for detecting the composition of aluminum alloys based on LIBS and PLS as described in claim 1, characterized in that: Step 2 is implemented as follows: Step 2.1: Obtain the average spectral intensity and standard deviation of the LIBS spectrum, use the Laida method to remove abnormal spectra from the LIBS spectrum, retain the LIBS spectra that satisfy the equation (1), remove the mean from the retained spectra, and obtain the average spectrum. [Mean spectral intensity ± 3*standard deviation](1) Step 2.2: Remove background interference from the background curves obtained using the window shift minimum method in the corresponding channels of the average spectrum by channel-specific processing; Step 2.3: Normalize the spectrum after removing background interference using the channel-specific background intensity normalization method; Step 2.4: Use the characteristic spectral library of common elements in aluminum alloys to extract the characteristic peaks of the normalized spectrum of the aluminum alloy sample.

4. The method for detecting the composition of aluminum alloys based on LIBS and PLS as described in claim 3, characterized in that: Step 2.2 is implemented as follows: Step 2.2.1: Use the window-shifting minimum value method to obtain the background minimum value of the retained spectrum, and perform polynomial fitting on the extracted background minimum value to form a background curve; Step 2.2.2: Subtract the background curve of the corresponding channel from the average spectrum to remove background interference by channel.

5. The method for detecting the composition of aluminum alloys based on LIBS and PLS as described in claim 3, characterized in that: Step 2.2 is implemented as follows: Step 2.3.1: Integrate the background curve intensity for each wavelength channel; Step 2.3.2: Divide the average spectral intensity by the integral value of the background curve intensity to obtain the normalized spectrum.

6. A rapid aluminum alloy composition detection system based on LIBS and PLS to implement the method described in claim 1, characterized in that: Including LIBS devices and computing mainframes; The LIBS device is used to acquire laser-induced plasma spectra of aluminum alloy samples and unknown aluminum alloy samples to be tested; it obtains LIBS spectra of aluminum alloy samples with wavelengths ranging from 180 to 950 nm and resolutions not greater than 0.05 nm by focusing a 1064 nm nanosecond pulsed laser with an energy range of 0-200 mJ onto the sample surface. The LIBS spectrum of the aluminum alloy sample was used as input to the computer. The computing host is used to obtain the composition prediction results of the aluminum alloy sample.

7. The rapid aluminum alloy composition detection system based on LIBS and PLS as described in claim 6, characterized in that: The repetition frequency of the LIBS device is 5-15 Hz.

Images