LIBS-based metal particle size estimation method and device, and LIBS detection system

Abstract

The application relates to the technical field of spectral detection, and discloses a metal particle size estimation method and device based on LIBS and a LIBS detection system, wherein the method comprises the following steps: acquiring LIBS spectrum data of samples with different particle sizes; performing feature variable screening on the LIBS spectrum data to obtain an optimal interval combination of wavelengths; training the optimal interval combination as an input quantity and the metal particle size as a response quantity to obtain candidate model parameters; performing k-fold cross validation on the candidate latent variable number by using MC-PLSR modeling, and combining the candidate model parameters to obtain an optimal latent variable number; the latent variable represents the relationship between spectrum information and the particle size; the MC-PLSR model is retrained by using the optimal latent variable number and the LIBS spectrum data to obtain a "particle size-LIBS signal" mapping function constructed by an optimal weight vector and a final regression coefficient vector. The method breaks the dependence of traditional particle size representation on high-resolution imaging equipment, and realizes high-precision and reliable prediction of the metal particle size in lubricating oil without intervention of a precision imager.

Description

Technical Field

[0001] This application relates to the field of spectroscopic detection technology, such as a LIBS-based method and apparatus for estimating the particle size of metal particles, and a LIBS detection system. Background Technology

[0002] Lubricating oil plays a crucial role in protecting and maintaining the operation of an engine. Under high temperature and pressure, friction inevitably occurs between internal engine components, generating micron-sized metal wear particles (such as Al, Fe, Cu, Cr, etc.) that mix into the lubricating oil. Over time, severely contaminated lubricating oil gradually turns black, a typical sign of oil aging and the suspension of a large number of metal particles.

[0003] The identification of metal particle size in lubricating oil currently relies primarily on microscopic imaging techniques, which estimate the actual particle size by analyzing the number of pixels in the particle image. However, this method is heavily dependent on image quality. When lubricating oil becomes black due to contamination, the imaging contrast decreases, and particle boundaries become blurred, easily leading to misjudgments of particle size. In such cases, repeated washing of the oil sample is necessary to remove background color interference before accurate measurement can be achieved. However, this process is cumbersome and time-consuming, making it difficult to meet the needs of rapid analysis and batch statistical analysis.

[0004] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this application, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] To provide a basic understanding of some aspects of the disclosed embodiments, a brief summary is given below. This summary is not intended as a general commentary, nor is it intended to identify key / important components or describe the scope of protection of these embodiments, but rather as a prelude to the detailed description that follows.

[0006] This disclosure provides a LIBS-based method and apparatus for estimating the particle size of metal particles, as well as a LIBS detection system, to achieve reliable analysis of the particle size of metal particles in contaminated oil.

[0007] In some embodiments, the LIBS-based metal particle size estimation method includes: a data acquisition step: acquiring LIBS spectral data of samples with different particle sizes; a screening step: screening the LIBS spectral data for characteristic variables to obtain the optimal wavelength range combination for each sample; an initial training step: training the model using the optimal range combination as input and the metal particle size as response to obtain candidate model parameters; a latent variable number determination step: performing k-fold cross-validation on the candidate latent variable number using MC-PLSR modeling, and combining the candidate model parameters to obtain the optimal latent variable number; wherein, the latent variable represents the relationship between spectral information and particle size; and a final training step: retraining the MC-PLSR model using the optimal latent variable number and the LIBS spectral data to obtain a "particle size-LIBS signal" mapping function constructed from the final weight vector and the final regression coefficient vector.

[0008] In some embodiments, the LIBS-based metal particle size estimation apparatus includes a processor and a memory storing program instructions, the processor being configured to execute the aforementioned LIBS-based metal particle size estimation method when the program instructions are executed.

[0009] In some embodiments, the LIBS detection system includes: a laser for providing a high-energy pulse required to excite a sample; an optical path system for receiving and adjusting the laser output from the laser to focus the laser onto the surface of the sample, thereby exciting the sample to generate plasma; a spectral acquisition system for receiving the emitted light of the plasma output from the optical path system and analyzing the emitted light to obtain spectral intensity; and a LIBS-based metal particle size estimation device as described above, which is communicatively connected to the spectral acquisition system.

[0010] In some embodiments, the computer-readable storage medium stores program instructions that, when executed, perform the aforementioned LIBS-based method for estimating the particle size of metal particles.

[0011] The LIBS-based metal particle size estimation method and apparatus, and LIBS detection system provided in this disclosure can achieve the following technical effects:

[0012] By screening characteristic variables from LIBS spectral data of samples with different particle sizes, multiple combinations of characteristic spectral line intervals with the highest correlation to particle size were obtained. The screened data were used as input and metal particle size as the response for training, yielding candidate model parameters. MC-PLSR modeling and cross-validation were used to determine the optimal number of latent variables to ensure the model's fitting ability and generalization. Finally, the MC-PLSR model was retrained using the optimal number of latent variables and LIBS spectral data to obtain the final "particle size-LIBS signal" mapping function. Thus, through systematic iteration on samples of different particle sizes, a precise quantitative mapping model of metal particle size-spectral response intensity was constructed using MC-PLSR, establishing the theoretical feasibility of quantitative characterization of microscale particles in the field of laser-induced spectroscopy. This is particularly suitable for "black" highly contaminated lubricating oil, which is difficult to process using traditional visual methods. It enables rapid and non-destructive particle size analysis of metal wear particles in the critical particle size range of 1 μm to 60 μm without clear preprocessing. This breakthrough eliminates the dependence of traditional particle size characterization on high-resolution imaging equipment, achieving high-precision and reliable prediction of metal particle size in lubricating oil without the intervention of a precision imager.

[0013] The above general description and the description below are exemplary and illustrative only and are not intended to limit this application. Attached Figure Description

[0014] One or more embodiments are illustrated by way of example with reference to the accompanying drawings. These illustrations and drawings do not constitute a limitation on the embodiments. Elements having the same reference numerals in the drawings are shown as similar elements. The drawings are not to be scaled. And wherein:

[0015] Figure 1 This is a schematic diagram of the LIBS detection system provided in this embodiment;

[0016] Figure 2 This is a schematic diagram of a LIBS-based method for estimating the particle size of metal particles provided in an embodiment of this disclosure;

[0017] Figure 3 This is a schematic diagram of a method for filtering feature variables from LIBS spectral data to obtain the optimal combination of wavelength ranges, provided in an embodiment of this disclosure.

[0018] Figure 4 This is a schematic diagram of a method provided in this disclosure for training a candidate model by using the optimal interval combination as input and the metal particle size as response, thereby obtaining the model parameters.

[0019] Figure 5 This is a schematic diagram of a method provided in this disclosure for using MC-PLSR modeling to perform k-fold cross-validation on the number of candidate latent variables and combining it with candidate model parameters to obtain the optimal number of latent variables;

[0020] Figure 6 This is a schematic diagram of another LIBS-based metal particle size estimation method provided in an embodiment of this disclosure;

[0021] Figure 7 This is a PCA load diagram obtained after extracting data features from standard samples of aluminum bronze, carbon steel, and aluminum-silicon alloy metal particles provided in this embodiment of the disclosure;

[0022] Figure 8 This is a score graph obtained by processing the LIBS spectral data of standard samples of three metal particles—aluminum bronze, carbon steel, and aluminum-silicon alloy—based on PCA, according to an embodiment of this disclosure.

[0023] Figure 9 This is an MC-PLSR particle size prediction diagram of aluminum bronze particles provided in the embodiments of this disclosure;

[0024] Figure 10 This is the metal particle detection result provided in the embodiments of this disclosure;

[0025] Figure 11 This is a schematic diagram of a LIBS-based metal particle size estimation device provided in an embodiment of this disclosure;

[0026] Figure 12 This is a schematic diagram of another LIBS-based device for estimating the particle size of metal particles provided in an embodiment of this disclosure.

[0027] Figure label:

[0028] 10. Laser; 20. Optical path system; 21. Mirror; 22. Dichroic mirror; 23. Achromatic objective lens; 30. Spectral acquisition system; 31. Spectral signal receiving component; 311. Lens; 312. Fiber optic probe; 32. Spectrometer; 40. Sample; 50. Three-dimensional displacement platform; 60. Computer; 70. PIN tube;

[0029] 110. LIBS-based metal particle size estimation device; 111. Data acquisition module; 112. Screening module; 113. Initial training module; 114. Latent variable number determination module; 115. Final training module;

[0030] 120. LIBS-based metal particle size estimation device; 121. Processor; 122. Memory; 123. Communication interface; 124. Bus. Detailed Implementation

[0031] To provide a more detailed understanding of the features and technical content of the embodiments of this disclosure, the implementation of the embodiments of this disclosure will be described in detail below with reference to the accompanying drawings. The accompanying drawings are for illustrative purposes only and are not intended to limit the embodiments of this disclosure. In the following technical description, for ease of explanation, several details are used to provide a full understanding of the disclosed embodiments. However, one or more embodiments may still be implemented without these details. In other cases, well-known structures and devices may be simplified in their depiction to simplify the drawings.

[0032] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this disclosure described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion.

[0033] Unless otherwise stated, the term "multiple" means two or more.

[0034] In this embodiment of the disclosure, the character " / " indicates that the objects before and after it are in an "or" relationship. For example, A / B means: A or B.

[0035] The term "and / or" describes an association between objects, indicating that three relationships can exist. For example, A and / or B means: A or B, or A and B.

[0036] The term "correspondence" can refer to an association or binding relationship. The correspondence between A and B means that there is an association or binding relationship between A and B.

[0037] Combination Figure 1As shown, this disclosure provides a LIBS (Laser-Induced Breakdown Spectroscopy) detection system, including a laser 10, an optical path system 20, and a spectral acquisition system 30. The laser 10, once activated, provides high-energy pulses to a sample 40, exciting the sample 40 to generate high-temperature plasma. Optionally, the laser 10 is a Q-switched Nd:YAG laser with a pulse repetition frequency of 1 Hz and a laser energy of 50 mJ. The optical path system 20 is positioned between the laser 10 and the sample 40. The laser emitted by the laser 10 is received by the receiving side of the optical path system 20, adjusted by the optical path system 20, and output through the first output side of the optical path system 20, focusing onto the surface of the sample 40, thereby exciting a localized area of ​​the sample 40 to generate high-temperature plasma. The emitted light from the high-temperature plasma is output through the second output side of the optical path system 20 and received by the spectral acquisition system 30. The spectral acquisition system 30 analyzes the emitted light to obtain the spectral intensity.

[0038] Optionally, sample 40 is obtained by ultrasonically dispersing standard metal particles and lubricating oil using an ultrasonic cleaner to obtain a metal particle lubricating oil sample. A certain amount of sample 40 is dropped onto filter paper, and after sample 40 forms an oil film, the filter paper is fixed on a three-dimensional displacement platform 50.

[0039] Optionally, the optical path system 20 includes a reflecting mirror 21, a dichroic mirror 22, and an achromatic objective lens 23 sequentially arranged along a preset optical path. The achromatic objective lens 23 serves as the first output side of the optical path system 20, and the dichroic mirror 22 serves as the second output side. The computer 60 is communicatively connected to the three-dimensional displacement platform 50. By adjusting the Z-axis of the three-dimensional displacement platform 50, the sample 40 is positioned at the focal point of the achromatic objective lens 23. The laser emitted by the laser 10 is first reflected by the reflecting mirror 21, causing the laser to enter the dichroic mirror 22. The laser is transmitted through the dichroic mirror 22 and focused onto the surface of the sample 40 by the achromatic objective lens 23.

[0040] Optionally, the spectral acquisition system 30 includes a spectral signal receiving component 31 and a spectrometer 32. The spectral signal receiving component 31 is located on the second output side of the optical path system 20, i.e., on one side of the dichroic mirror 22, to collect spectral signals backward. The spectrometer 32 is communicatively connected to the spectral signal receiving component 31 and receives the emitted light from the plasma. Subsequently, the spectrometer 32 disperses the light, and the ICCD detector of the spectrometer 32 amplifies the signal, completes photoelectric conversion, and acquires the spectral signal to obtain the spectral intensity.

[0041] The spectrometer 32 is also connected in communication with the computer 60 to send spectral intensity to the computer 60.

[0042] Optionally, the spectral signal receiving component 31 includes a lens 311 and an optical fiber probe 312. The lens 311 is disposed on one side of the dichroic mirror 22, and the optical fiber probe 312 is disposed on one side of the lens 311 and is communicatively connected to the first input channel of the spectrometer 32 via an optical fiber.

[0043] Optionally, the LIBS detection system also includes a PIN photodiode 70, disposed on one side of the reflector 21. The second input channel of the spectrometer 32 is communicatively connected to the PIN photodiode 70. When the laser emitted by the laser 10 reaches the PIN photodiode 70 through the reflector 21, the PIN photodiode 70 responds and triggers the spectrometer 32 to acquire the spectral signal.

[0044] Combination Figure 2 As shown, this disclosure provides a LIBS-based method for estimating the particle size of metal particles, including:

[0045] S101, obtain LIBS spectral data of samples with different particle sizes;

[0046] S102, Select characteristic variables from LIBS spectral data to obtain the optimal combination of wavelength ranges for each sample;

[0047] S103, use the optimal interval combination as the input and the metal particle size as the response to train the candidate model parameters.

[0048] S104, using MC-PLSR modeling to perform k-fold cross-validation on the number of candidate latent variables, and combining the candidate model parameters to obtain the optimal number of latent variables; where latent variables represent the relationship between spectral information and particle size;

[0049] S105, the MC-PLSR model is retrained using the optimal number of latent variables and LIBS spectral data to obtain the "particle size-LIBS signal" mapping function constructed from the final weight vector and the final regression coefficient vector.

[0050] use Figure 1 The LIBS detection system shown ablates samples of different particle sizes to obtain LIBS spectral data for each sample, forming a full-spectrum matrix for multiple samples (e.g., a matrix of 3 samples × 800 wavelength points). The particle size of the samples is known. Optionally, the samples are metal particles with particle sizes ranging from 1 μm to 60 μm. Using siPLS (Combined Interval PLS), the system automatically filters for "combinations of multiple characteristic spectral line intervals with the highest correlation to particle size" within the LIBS full-spectrum matrix, i.e., it filters characteristic variables to obtain the optimal wavelength interval combination for each sample.

[0051] The optimal combination of wavelength intervals is used as the input (input matrix). The metal grain size is used as the response quantity to train the model, and candidate model parameters are obtained by solving the model. These candidate model parameters can then be used to construct a mapping function that satisfies physical laws. ,in, For particle size, This represents the regression coefficient vector of the model. This is a matrix composed of the intensities of multiple highly correlated characteristic spectral lines selected by siPLS. The constructed "particle size-LIBS signal" mapping function is also shown. Strictly satisfying the condition that an increase in the intensity of any feature spectral line will lead to a monotonically increasing predicted particle size value, thus ensuring that the model satisfies both the statistical learning objective and the physical laws.

[0052] MC-PLSR modeling was used to perform k-fold cross-validation on the number of candidate latent variables, where there were multiple candidate latent variables. Within the Partial Least Squares (PLS) framework, the training data was processed through a series of latent variables... LV This is used to characterize the relationship between spectral information and particle size. LV This represents the number of latent factors extracted from the original high-dimensional spectral variables (hundreds or thousands of wavelengths). In PLS, latent variables are linear combinations obtained by maximizing the covariance of the independent variable X and the dependent variable y. ,in, t It is a latent variable. 𝑤 It is a weight vector. Therefore LV This indicates the number of such latent variables. These latent variables are retained for modeling. The choice of the number of latent variables directly affects the model's fitting ability and generalization performance; therefore, cross-validation is needed to determine the optimal number of latent variables. LV *

[0053] The MC-PLSR model was retrained using the optimal number of latent variables and LIBS spectral data, and the final weight vector was obtained by solving the problem. and the final regression coefficient vector and based on and Construct the final "particle size-LIBS signal" mapping function: ,in, Used to construct latent variables The final particle size estimation model is based on regression coefficients. With input matrix Directly establish, that is This leads to the particle size estimation model.

[0054] The aforementioned mapping function can be used to predict the particle size of unknown samples. Under the same conditions, the LIBS spectra of metal particles at the location are collected using the aforementioned LIBS detection system. After identifying the metal type, the pre-constructed MC-PLSR particle size estimation model corresponding to that metal type is invoked. The unknown spectrum is input into the model, and the model can output the predicted value of its particle size.

[0055] This embodiment employs a LIBS-based metal particle size estimation method. By screening characteristic variables from LIBS spectral data of samples with different particle sizes, multiple combinations of characteristic spectral line intervals with the highest correlation to particle size are obtained. The screened data are used as input and metal particle size as the response for training, yielding candidate model parameters. MC-PLSR modeling and cross-validation are used to determine the optimal number of latent variables to ensure the model's fitting ability and generalization. Finally, the MC-PLSR model is retrained using the optimal number of latent variables and LIBS spectral data to obtain the final "particle size-LIBS signal" mapping function. Thus, through systematic iteration on samples of different particle sizes, a precise quantitative mapping model of metal particle size-spectral response intensity is constructed using MC-PLSR, establishing the theoretical feasibility of quantitative characterization of microscale particles in the field of laser-induced spectroscopy. This method is particularly suitable for "black" highly contaminated lubricating oil, which is difficult to process using traditional visual methods, enabling rapid and non-destructive particle size analysis of metal wear particles within the critical particle size range of 1 μm to 60 μm without clear preprocessing. Breakthrough: This technology breaks free from the reliance on high-resolution imaging equipment for traditional particle size characterization, enabling high-precision and reliable prediction of the particle size of metal particles in lubricating oil without the intervention of a precision imager.

[0056] Optionally, the coefficient of determination (R²) is used for the final constructed "particle size-LIBS signal" mapping function. 2 The prediction accuracy and reliability of the model are evaluated. If R... 2 If the value is greater than 0.9, then the particle size of the metal particles is linearly related to the intensity of the LIBS spectrum. By combining the LIBS detection system with the MC-PLSR data processing method, an effective technical means can be provided for the analysis of the particle size of metal particles.

[0057] Optionally, combined Figure 3 As shown in S102, characteristic variable screening is performed on the LIBS spectral data to obtain the optimal wavelength range combination for each sample, including:

[0058] S112 divides the LIBS spectral data into multiple continuous wavelength ranges;

[0059] S122, optimizes each wavelength range and different combinations of each wavelength range;

[0060] S132 outputs the physical wavelength range of the optimal interval combination based on minimizing cross-validation error and maximizing the coefficient of determination.

[0061] LIBS spectral data was divided into multiple continuous wavelength intervals using siPLS, and optimization was attempted at the interval level for different wavelength intervals and their combinations. The goal was to minimize the cross-validation error (RMSECV) and the coefficient of determination (R²). 2 The criterion is to maximize the range of intervals that significantly improve prediction performance and have statistical significance. The final output is the physical wavelength range of the optimal interval combination. And its index position in the input matrix, the index number is: That is, the range of column numbers corresponding to this interval in the spectral matrix, which provides a clear set of input variables for subsequent regression MC-PLSR (PLS with monotonic constraints) modeling.

[0062] Optionally, combined Figure 4 As shown in S103, the optimal interval combination is used as the input and the metal particle size is used as the response to train the candidate model parameters, including:

[0063] S113, mean centering of the input and response quantities to obtain the centering matrix;

[0064] S123, construct a constrained optimization problem using a centered matrix and apply monotonic constraints, iteratively updating the model's weight vector and regression coefficient vector;

[0065] S133 outputs a weight vector and regression coefficient vector that satisfy the monotonicity constraint when the iteration terminates, as candidate model parameters.

[0066] First, model training with monotonicity constraints applied:

[0067] (1) Input sample intensity matrix 𝑋 (i.e., optimal interval combination) and response vector 𝑦. Perform mean centering on 𝑋 and 𝑦 to obtain the centered matrix. .in, This represents an n-order identity matrix. This represents a vector of all 1s with dimension n. This represents an n×n matrix consisting entirely of 1s.

[0068] Secondly, using a centralized matrix A constrained optimization problem is constructed, comprising an objective function, an optimization objective, and constraints. Monotonicity constraints are imposed on the objective function. Based on the constraints and the optimization objective, the model's weight vector and regression coefficient vector are iteratively updated. After iteration, the weight vector and regression coefficient vector satisfying the monotonicity constraints are output and used as candidate parameters for the model. Thus, by constructing a constrained optimization problem and imposing monotonicity constraints on the objective function, the model's weight vector and regression coefficient vector are iteratively updated, ultimately yielding weight vectors and regression coefficient vectors satisfying the monotonicity constraints. A mapping function is constructed based on this. It can ensure that the model meets the statistical learning objectives and conforms to the laws of physics.

[0069] Specifically, it includes:

[0070] (2) Iterate using the NIPALS algorithm. In each step of the iteration, the solution for the weight vector 𝑤 is transformed into a constrained optimization problem, including:

[0071] Objective function: ;

[0072] Optimization goal: ;

[0073] Constraints st ;

[0074] ≥0;

[0075] in This is the vector of regression coefficients.

[0076] (3) Apply monotonic constraints: After each iteration of the conventional PLS algorithm updates 𝑤, apply monotonic constraints to the objective function and to the regression coefficient vector corresponding to the updated 𝑤. Perform symbol detection, Project onto a non-negative orthogonal space, i.e., take all non-negative elements.

[0077] (4) Iteratively update the score vector: based on the corrected weight vector 𝑤 Calculate the score vector t=X 𝑤. Use t right and y Perform regression and calculate the residual matrix. Repeat the above process until the preset convergence condition is met: iteration error. <10 -5 The iteration terminates at this point. The iteration error is the relative difference between the weight vectors obtained in two consecutive iterations.

[0078] (5) Output monotonicity constraint model: Obtain a set of final weight vectors that meet the monotonicity constraints. 𝑤 With regression coefficient vector As candidate model parameters, this demonstrates that the monotonicity constraint mechanism is feasible and can generate a mapping function that satisfies physical laws. .

[0079] Optionally, combined Figure 5 As shown in S104, k-fold cross-validation of the candidate latent variables is performed using MC-PLSR modeling, and the optimal number of latent variables is obtained by combining the candidate model parameters, including:

[0080] S114, divide the LIBS spectral data into a training set and a validation set;

[0081] S124, perform MC-PLSR modeling for each fold and apply monotonic constraints, use the number of candidate latent variables to perform k-fold cross-validation, and obtain the root mean square error of cross-validation for each candidate latent variable.

[0082] S134: The number of candidate latent variables corresponding to the average of the smallest cross-validation root mean square error is taken as the optimal number of latent variables. Modeling is performed for each fold and monotonic constraints are applied. K-fold cross-validation is then conducted using the number of candidate latent variables to obtain the cross-validation root mean square error corresponding to each candidate latent variable number.

[0083] k-fold cross-validation determines the number of latent variables in MC-PLSR. LV:

[0084] (1) Divide the training-validation set: k Cross-validation of fold count k The value of is set to 5, and the total number of samples n is randomly divided into k Approximate sets D1, D2, ..., D k Used in each iteration k -1 subset is used as the training set, and the remaining 1 subset is used as the validation set, alternating between the two. k Second-rate.

[0085] (2) Try different numbers of latent variables in a loop. LV Number of candidate latent variables LV =1,2,..., LV Maximize. Model MC-PLSR for each fold, apply monotonicity constraints, and ensure that the regression coefficient vector... Consistent signs. For each group LV The prediction function is obtained on the training set: ,in, For the first The score vector of each latent variable. The corresponding regression weights are given.

[0086] (3) Predict on the validation set: using the candidate model parameters obtained during training Make predictions and calculate the root mean square error of cross-validation. ) and the coefficient of determination R 2 .

[0087] (4) Summary k Folding results: For each candidate LV Take the average RMSECV and R 2 The overall cross-validation metric was obtained.

[0088] (5) Selecting the optimal number of latent variables LV* Choose the one that minimizes the average RMSECV. LV As the optimal number of latent variables LV *

[0089] Combination Figure 6 As shown, this disclosure provides another LIBS-based method for estimating the particle size of metal particles, including:

[0090] S101, obtain LIBS spectral data of samples with different particle sizes;

[0091] S106, constructing a sample intensity matrix using LIBS spectral data;

[0092] S107 uses the PCA algorithm to analyze the sample intensity matrix and extract key feature spectral lines to distinguish the types of metal particles.

[0093] S108, Construct a PCA metal classification model based on key feature spectral lines;

[0094] S102, Select characteristic variables from LIBS spectral data to obtain the optimal combination of wavelength ranges for each sample;

[0095] S103, use the optimal interval combination as the input and the metal particle size as the response to train the candidate model parameters.

[0096] S104, using MC-PLSR modeling to perform k-fold cross-validation on the number of candidate latent variables, and combining the candidate model parameters to obtain the optimal number of latent variables; where latent variables represent the relationship between spectral information and particle size;

[0097] S105, the MC-PLSR model is retrained using the optimal number of latent variables and LIBS spectral data to obtain the "particle size-LIBS signal" mapping function constructed from the final weight vector and the final regression coefficient vector.

[0098] After obtaining the LIBS spectral data of the samples, a sample intensity matrix (e.g., a matrix of 3 samples × 800 wavelength points) is constructed using the LIBS spectral data. This matrix is ​​then input into the PCA algorithm. The PCA algorithm uses linear transformation to find a new comprehensive variable—the principal component (PC)—that best preserves the variance of the original data. The score plot is analyzed to observe the distribution of the samples in the two-dimensional space formed by PC1 and PC2. Samples of the same metal will cluster together to form independent groups, while clusters of different metals will be clearly distinguished. This achieves non-destructive identification of metal species.

[0099] Analyzing the loading plots, we interpret the physical meaning of the principal components. Wavelengths with high absolute loading values ​​are the characteristic spectral lines that contribute most to distinguishing different metal species. These lines correspond to the major or characteristic elements of each metal, and their intensity has the strongest potential correlation with grain size; they are considered "key spectral lines." This provides a basis for a discriminative model for the classification of subsequent unknown samples. A PCA metal classification model is constructed using these key characteristic spectral lines. This approach, utilizing the PCA algorithm, automatically and objectively selects the key characteristic wavelengths for distinguishing metal species, avoiding the complexity and subjectivity of manual selection.

[0100] When detecting the metal particles to be tested, the spectral information of the particles is first collected using a LIBS detection system and input into a pre-constructed PCA classification model to identify the metal type. After the type is determined, the spectral data is then imported into the corresponding MC-PLSR particle size prediction model, and the particle size estimate of the metal particles is calculated based on the established linear relationship.

[0101] Through systematic iterative experiments with multi-scale particle size samples, combined with the PCA algorithm feature selection mechanism, MC-PLSR constructed a precise quantitative mapping model between metal particle size and spectral response intensity. Experimental verification showed that the model's coefficient of determination (R²) was [value missing]. 2 >0.9) is statistically significant. This method achieves dual-dimensional diagnostic capabilities of simultaneously analyzing the chemical composition and accurately assessing the particle size of 1-60 μm metal particles in lubricating oil without the intervention of a precision imaging instrument, providing an integrated analytical dimension for the wear condition of industrial equipment.

[0102] In practical applications, such as Figure 7 The PCA load diagram shown illustrates the load distribution of the first principal component (PC1) obtained after extracting data features from standard samples of aluminum bronze, carbon steel, and aluminum-silicon alloy metal particles based on laser-induced breakdown spectroscopy (LIBS) combined with principal component analysis (PCA). Figure 7The upper-middle section represents the original LIBS spectrum, showing the effect of superimposing the original LIBS spectra of multiple standard samples in the wavelength range of 250 nm to 330 nm. Due to the differences in spectral signal intensity between particles of different sizes, as well as the presence of background noise and spectral line overlap, it is difficult to directly identify which specific spectral lines are most correlated with changes in particle size from the original spectra with the naked eye. Figure 7 The lower half of the image shows the PC1 loading coefficient distribution. This region represents the core result after PCA algorithm processing, displaying the loading coefficients of the first principal component (PC1) at each wavelength. PC1 represents the direction of maximum variance variation in the original spectral data, typically capturing the most significant and dominant influencing factors. The vertical axis (loading coefficients) reflects the sensitivity and contribution of the spectral intensity at the corresponding wavelength to changes in particle size. Larger absolute values ​​indicate more drastic changes in the spectral signal due to particle type, and a stronger correlation with particle type. The horizontal axis (wavelength) corresponds to the wavelength range of the original spectrum in the upper half.

[0103] See Figure 8 This figure shows the score plots after processing the LIBS spectral data of standard samples of three metal particles—aluminum bronze, carbon steel, and aluminum-silicon alloy—based on Principal Component Analysis (PCA). The plot visually demonstrates how the PCA model, after dimensionality reduction, effectively distinguishes and clusters the spectral data of different particle types in a two-dimensional space spanned by the first principal component (PC1) and the second principal component (PC2). The horizontal axis (PC1) represents the first principal component, with a contribution rate of 44.8%. This means that the difference along the PC1 direction can explain 44.8% of the variation information in the original spectral data. The vertical axis (PC2) represents the second principal component, with a contribution rate of 34.7%. PC2 captures the largest variance information besides PC1. The cumulative contribution rate of PC1 and PC2 is as high as 79.5%, indicating that this two-dimensional plane has preserved most of the core features of the original high-dimensional spectral data, sufficient for reliable pattern recognition. Figure 8 Each point represents the projected position of a single LIBS measurement spectrum of a standard sample in the space defined by PC1 and PC2. For example... Figure 8 As shown, the data points of three different types of metal particles form three distinct clusters.

[0104] See Figure 9 This paper presents the MC-PLSR particle size prediction map of aluminum bronze particles. The LIBS spectra and corresponding particle sizes of the PCA-classified aluminum bronze particles are imported into the MC-PLSR algorithm. siPLS automatically selects the combination of feature spectral intervals with the highest correlation to particle size from the full spectrum, and then performs k-fold cross-extraction. and YThe latent variables in the model are used to establish a linear regression model between them under the imposition of monotonic constraints on the physical model. The horizontal axis represents the actual particle size of the standard particle, serving as a benchmark for evaluating the accuracy of the MC-PLSR prediction results. The vertical axis represents the particle size predicted by the MC-PLSR model based on LIBS spectral data. Figure 9 Each scatter point represents an independent aluminum bronze particle standard sample. It can be observed that all data points are very closely distributed near a straight line close to the ideal 45° diagonal. This indicates that the model's predictions are in high agreement with the actual values. For the MC-PLSR particle size prediction curve of aluminum bronze particles, R0... 2 The coefficient of determination (COD) of 0.92 indicates that this MC-PLSR particle size prediction model can accurately estimate the particle size of aluminum bronze particles. This represents a very high prediction accuracy in practical industrial applications, fully demonstrating the model's reliability. The figure objectively demonstrates that "a strong functional relationship exists between LIBS spectral intensity and metal particle size, which can be explored and utilized." By fully utilizing LIBS spectral information and combining it with chemometric algorithms (PCA+MC-PLSR), it is entirely possible to develop a highly efficient online particle size analysis technique.

[0105] Figure 10 This image displays the detection results of metal particles, specifically a 3D image of aluminum bronze particles obtained after detection by the LIBS system and processing using the PCA and MC-PLSRS algorithms. In the image, red spheres represent metal particles. The X and Y axes represent the particle location information (unit: mm), obtained through rapid scanning of the sample using a 3D displacement platform. The Z-axis represents the particle size information (unit: μm), derived from a previously established MC-PLSR particle size estimation model. The results provide a clear visual representation of the location and size of the metal particles.

[0106] Combination Figure 11As shown in the figure, this disclosure provides a LIBS-based metal particle size estimation device 110, including: a data acquisition module 111, a screening module 112, an initial training module 113, a latent variable number determination module 114, and a final training module 115. The data acquisition module 111 is configured to acquire LIBS spectral data of samples with different particle sizes. The screening module 112 is configured to screen the LIBS spectral data for characteristic variables to obtain the optimal wavelength range combination for each sample. The initial training module 113 is configured to train the model using the optimal range combination as input and the metal particle size as response to obtain candidate model parameters. The latent variable number determination module 114 is configured to perform k-fold cross-validation on the candidate latent variable number using MC-PLSR modeling, and combine the candidate model parameters to obtain the optimal latent variable number; wherein, the latent variables characterize the relationship between spectral information and particle size. The final training module 115 is configured to retrain the MC-PLSR model using the optimal number of latent variables and LIBS spectral data, resulting in a "particle size-LIBS signal" mapping function constructed from the final weight vector and the final regression coefficient vector.

[0107] The LIBS-based metal particle size estimation device 110 provided in this disclosure uses LIBS spectral data of samples with different particle sizes to screen characteristic variables, obtaining multiple combinations of characteristic spectral line intervals with the highest correlation to particle size. The screened data is used as input and metal particle size as response quantity for training, yielding candidate model parameters. The optimal number of latent variables is determined through MC-PLSR modeling and cross-validation to ensure the model's fitting ability and generalization. Finally, the MC-PLSR model is retrained using the optimal number of latent variables and LIBS spectral data to obtain the final "particle size-LIBS signal" mapping function. Thus, through systematic iteration on samples of different particle sizes, a precise quantitative mapping model of metal particle size-spectral response intensity is constructed using MC-PLSR, establishing the theoretical feasibility of quantitative characterization of microscale particles in the field of laser-induced spectroscopy. This is particularly suitable for "black" highly contaminated lubricating oil, which is difficult to process using traditional visual methods, enabling rapid and non-destructive particle size analysis of metal wear particles in the critical particle size range of 1 μm to 60 μm without clear preprocessing. Breakthrough: This technology breaks free from the reliance on high-resolution imaging equipment for traditional particle size characterization, enabling high-precision and reliable prediction of the particle size of metal particles in lubricating oil without the intervention of a precision imager.

[0108] Combination Figure 12As shown, this disclosure provides a LIBS-based metal particle size estimation device 120, including a processor 121 and a memory 122. Optionally, the device 120 may further include a communication interface 123 and a bus 124. The processor 121, communication interface 123, and memory 122 can communicate with each other via the bus 124. The communication interface 123 can be used for information transmission. The processor 121 can call logical instructions in the memory 122 to execute the LIBS-based metal particle size estimation method of the above embodiment.

[0109] Furthermore, the logic instructions in the aforementioned memory 122 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium.

[0110] The memory 122, as a computer-readable storage medium, can be used to store software programs and computer-executable programs, such as program instructions / modules corresponding to the methods in the embodiments of this disclosure. The processor 121 executes functional applications and data processing by running the program instructions / modules stored in the memory 122, that is, it implements the LIBS-based metal particle size estimation method in the above embodiments.

[0111] The memory 122 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the terminal device. Furthermore, the memory 122 may include high-speed random access memory and may also include non-volatile memory.

[0112] This disclosure provides another LIBS detection system, including a laser 10, an optical path system 20, and a spectral acquisition system 30, as well as the aforementioned LIBS metal particle size estimation device 110 (120). The LIBS metal particle size estimation device 110 (120) is installed in a computer. The installation relationship described herein is not limited to placement inside the computer, but also includes installation connections with other components of the LIBS detection system, including but not limited to physical connections, electrical connections, or signal transmission connections. Those skilled in the art will understand that the LIBS metal particle size estimation device 110 (120) can be adapted to feasible product bodies to achieve other feasible embodiments.

[0113] For the specific structure and connection relationship of laser 10, optical path system 20 and spectral acquisition system 30, please refer to Figure 1 This will not be elaborated upon here.

[0114] This disclosure provides a computer-readable storage medium storing computer-executable instructions configured to perform the above-described LIBS-based metal particle size estimation method.

[0115] The technical solutions of this disclosure can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes one or more instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in this disclosure. The aforementioned storage medium can be a non-transitory storage medium, such as a USB flash drive, external hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, etc., and other media capable of storing program code.

[0116] The foregoing description and accompanying drawings fully illustrate embodiments of this disclosure to enable those skilled in the art to practice them. Other embodiments may include structural, logical, electrical, procedural, and other changes. The embodiments represent only possible variations. Individual components and functions are optional unless explicitly required, and the order of operation may vary. Parts and features of some embodiments may be included in or replace parts and features of other embodiments. Moreover, the terminology used in this application is for describing embodiments only and is not intended to limit the claims. As used in the description of embodiments and claims, the singular forms “a,” “an,” and “the” are intended to equally include the plural forms unless the context clearly indicates otherwise. Similarly, the term “and / or” as used in this application means including one or more of the associated listed items and all possible combinations thereof. Additionally, when used in this application, the term "comprise" and its variations "comprises" and / or "comprising" refer to the presence of stated features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof. Without further limitations, an element defined by the phrase "comprises a..." does not exclude the presence of other identical elements in the process, method, or apparatus that includes said element. In this document, each embodiment may focus on the differences from other embodiments, and similar or identical parts between embodiments can be referred to mutually. For methods, products, etc., disclosed in the embodiments, if they correspond to the method section disclosed in the embodiments, the relevant parts can be referred to the description of the method section.

[0117] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the embodiments of this disclosure. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0118] The methods and products (including but not limited to devices and equipment) disclosed in the embodiments herein can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units may be merely a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the shown or discussed units may be through some interfaces, and the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected to implement this embodiment according to actual needs. Furthermore, the functional units in the embodiments of this disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0119] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than that shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur in a different order than disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. Each block in a block diagram and / or flowchart, and combinations of blocks in a block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

Claims

1. A LIBS-based method for estimating the particle size of metal particles, characterized in that, include: S1, Data acquisition steps: Acquire LIBS spectral data of samples with different metal particle sizes; S2, Screening Step: Screening the LIBS spectral data for characteristic variables to obtain the optimal wavelength range combination for each sample; wherein, screening the LIBS spectral data for characteristic variables to obtain the optimal wavelength range combination for each sample includes: dividing the LIBS spectral data into multiple continuous wavelength ranges; optimizing each wavelength range and different combinations of each wavelength range; and outputting the physical wavelength range of the optimal range combination based on minimizing cross-validation error and maximizing the coefficient of determination. S3, Initial Training Steps: The optimal interval combination is used as the input and the metal particle size as the response to train the model and obtain candidate model parameters. This process includes: centering the input and response values ​​to obtain a centering matrix; constructing a constrained optimization problem using the centering matrix and applying monotonicity constraints; iteratively updating the model's weight vector and regression coefficient vector; and outputting the weight vector and regression coefficient vector that satisfy the monotonicity constraints when the iteration terminates, as the candidate model parameters. S4, Latent Variable Number Determination Step: The candidate latent variable numbers are cross-validated using the MC-PLSR model with k-fold cross-validation, and the optimal number of latent variables is obtained by combining the candidate model parameters. The latent variable number represents the relationship between the spectral data and the particle size of the metal particles. The step of using the MC-PLSR model to perform k-fold cross-validation on the candidate latent variables and obtaining the optimal number of latent variables by combining the candidate model parameters includes: dividing the LIBS spectral data into a training set and a validation set; modeling each fold using the MC-PLSR model and applying monotonic constraints; performing k-fold cross-validation on the candidate latent variables to obtain the root mean square error of cross-validation for each candidate latent variable number; and taking the candidate latent variable number corresponding to the average of the smallest root mean square error of cross-validation as the optimal number of latent variables. S5, Final training step: Retrain the MC-PLSR model using the optimal number of latent variables and the LIBS spectral data to obtain the "particle size-LIBS signal" mapping function constructed from the final weight vector and the final regression coefficient vector.

2. The LIBS-based method for estimating metal particle size according to claim 1, characterized in that, The step of constructing a constrained optimization problem using the centered matrix and applying monotonic constraints includes: The constrained optimization problem of the NIPALS algorithm is constructed using the centralized matrix, and the weight vector and regression coefficient vector of the model are iteratively updated; wherein, the constrained optimization problem includes: objective function, optimization objective and constraint conditions; After updating the weight vector in each iteration, a monotonicity constraint is applied, and the regression coefficient vector corresponding to the model is projected into a non-negative orthogonal space.

3. The LIBS-based method for estimating metal particle size according to claim 1, characterized in that, The iterative update model's weight vector and regression coefficient vector include: Calculate the score vector based on the iteratively updated weight vector; The score vector is used to regress the input and output quantities and the residual matrix is ​​calculated. The iteration calculates the score vector and the residual matrix iteratively until the preset convergence condition is met, at which point the iteration terminates.

4. The LIBS-based method for estimating metal particle size according to any one of claims 1 to 3, characterized in that, After acquiring LIBS spectral data of samples with different particle sizes, the method further includes: A sample intensity matrix was constructed using the LIBS spectral data; The PCA algorithm was used to analyze the intensity matrix of the samples and extract key feature spectral lines that distinguish the types of metal particles. A PCA metal classification model is constructed based on the key feature spectral lines.

5. A LIBS-based metal particle size estimation device, comprising a processor and a memory storing program instructions, characterized in that, The processor is configured to execute the LIBS-based metal particle size estimation method as described in any one of claims 1 to 4 when running the program instructions.

6. A LIBS detection system, characterized in that, include: A laser, used to provide the laser light required to excite a sample; An optical path system is used to receive and adjust the laser output from the laser, so that the laser is focused on the surface of the sample and excites the sample to generate plasma; A spectral acquisition system is used to receive the emitted light of the plasma output from the optical path system and analyze the emitted light to obtain the spectral intensity; The LIBS-based metal particle size estimation device as described in claim 5 is communicatively connected to the spectral acquisition system.

7. A computer-readable storage medium storing program instructions, characterized in that, When the program instructions are executed, they cause the computer to perform the LIBS-based metal particle size estimation method as described in any one of claims 1 to 4.

Images