Prediction method of optimal acquisition parameters of LIBS system and related equipment

Abstract

This disclosure provides a method, apparatus, and system for predicting optimal acquisition parameters for a LIBS system, belonging to the field of data processing technology. Specifically, it includes: Step 1, acquiring initial detection data of the sample to be tested; Step 2, obtaining a correlation model corresponding to the initial detection data; Step 3, predicting the acquisition parameters that optimize the preset spectral quality index based on the correlation model; Step 4, outputting the optimal acquisition parameters. This disclosure improves prediction efficiency, accuracy, and adaptability.

Description

Technical Field

[0001] This disclosure relates to the field of laser-induced breakdown spectroscopy detection and data processing technology, and in particular to a method, apparatus and system for predicting optimal acquisition parameters for a LIBS system. Background Technology

[0002] Currently, laser-induced breakdown spectroscopy (LIBS) elemental detection technology boasts advantages such as rapid response, low sample pretreatment requirements, and support for non-contact in-situ online detection, making it a promising technology for industrial process monitoring, geological exploration, and environmental monitoring. During LIBS analysis, spectral quality is closely related to acquisition parameters, which typically include delay time, integration time, and gain settings. Inappropriate acquisition parameter settings can easily lead to problems such as excessively strong continuous background, attenuation of characteristic spectral lines, or spectral saturation, thereby reducing the signal-to-noise ratio, signal-to-background ratio, and spectral intensity, affecting the accuracy, precision, and stability of elemental quantitative analysis. Furthermore, the high-quality spectral acquisition conditions of LIBS equipment are easily affected by various unstable factors during excitation and sampling, causing drift. For example, laser energy fluctuations can alter ablation efficiency and the initial plasma state; surface heterogeneity can lead to inconsistent ablation yield and plasma evolution; and fluctuations in ambient pressure, focusing lens contamination, and changes in sample surface morphology can also affect plasma formation and light collection efficiency, rendering previously optimal acquisition parameters inapplicable.

[0003] It is evident that there is an urgent need for a method to predict the optimal acquisition parameters of LIBS systems with high prediction efficiency, accuracy, and adaptability. Summary of the Invention

[0004] In view of this, the present disclosure provides a method, apparatus and system for predicting optimal acquisition parameters of a LIBS system, which at least partially solves the problems of poor prediction efficiency, accuracy and adaptability in the prior art.

[0005] In a first aspect, embodiments of this disclosure provide a method for predicting optimal acquisition parameters for a LIBS system, including: Step 1: Obtain the initial detection data of the sample to be tested; Step 2: Obtain the association model corresponding to the initial detection data; Step 3: Based on the correlation model, predict the acquisition parameters that will optimize the preset spectral quality index. Step 4: Output the predicted optimal acquisition parameters.

[0006] Secondly, embodiments of this disclosure provide an optimal acquisition parameter prediction device for a LIBS system, including a data module, a model module, a prediction module, and an output module.

[0007] Thirdly, embodiments of this disclosure provide a laser-induced breakdown spectroscopy detection system, including an excitation unit, a control unit, an acquisition unit, a prediction unit, and an output unit.

[0008] Fourthly, embodiments of this disclosure provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method for predicting optimal acquisition parameters for the system.

[0009] Fifthly, embodiments of this disclosure provide a computer program product, including a computer program or instructions, which, when executed by a processor, implement the above-described method for predicting optimal acquisition parameters for a system.

[0010] The beneficial effects of the embodiments disclosed herein are as follows: (1) The time spent on parameter optimization is greatly reduced: The traditional "exhaustive search" logic that relies on hardware stepping and physical excitation is reconstructed into an "algorithm reasoning" process based on an association model. The time spent determining the optimal acquisition parameters is changed from depending on the spectrometer acquisition cycle and the mechanical movement frequency to the millisecond-level processing cycle of the computing unit. Therefore, the system only needs a very short computational response to complete the optimization, reducing the total time from minutes / hours to less than seconds.

[0011] (2) Significantly reduced sample loss: The correlation model establishes a data mapping mechanism with "high information density". Only a single or single-digit number of transient signals generated by initial excitation are needed to extract characteristic information representing the physical evolution trend. This paradigm eliminates the need for physical scanning of the global parameter space and minimizes the physical ablation requirement of the sample in the optimization process, thereby significantly reducing the loss of trace or precious samples.

[0012] (3) Accurate prediction of optimal acquisition parameters: There is an essential dynamic mapping relationship between the initial detection data and the optimal acquisition parameters. By learning the plasma cooling evolution law, the correlation model can still capture the changes in data characteristics and map out the matching parameter correction amount in real time even when the environmental conditions deviate, ensuring that the acquisition state is always locked in the physical optimal range, so that the prediction index is highly close to the actual theoretical optimal value. Attached Figure Description

[0013] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a flowchart illustrating a method for predicting optimal acquisition parameters in a LIBS system, as provided in an embodiment of this disclosure. Detailed Implementation

[0015] The embodiments of this disclosure will now be described in detail with reference to the accompanying drawings.

[0016] The following specific examples illustrate the implementation of this disclosure. Those skilled in the art can easily understand other advantages and effects of this disclosure from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. This disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this disclosure. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0017] It should be noted that various aspects of embodiments within the scope of the appended claims are described below. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and any particular structure and / or function described herein is merely illustrative. Based on this disclosure, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects set forth herein can be used to implement the device and / or practice the method. Additionally, this device and / or method can be implemented using structures and / or functionalities other than one or more of the aspects set forth herein.

[0018] It should also be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this disclosure. The illustrations only show the components related to this disclosure and are not drawn according to the number, shape and size of the components in actual implementation. In actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0019] Furthermore, specific details are provided in the following description to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that the described aspects can be practiced without these specific details.

[0020] Currently, the closest existing technology to this disclosure is mainly a LIBS acquisition parameter optimization method based on parameter search, statistical experiment screening, or empirical setting. For example, it involves pre-setting the search range and step size for acquisition parameters such as delay time, integration time, gain, or gate width, performing spectral acquisition point by point, and selecting optimal parameter values ​​or combinations based on spectral quality indicators; or using orthogonal experiments, response surface analysis, etc., to screen for optimal operating conditions while reducing the number of experiments; or directly setting acquisition parameters based on historical experimental results and empirical rules. While these methods can improve spectral acquisition performance to some extent, they still rely on pre-setting, screening, comparison, or empirical setting of candidate parameters, making it difficult to quickly determine the optimal acquisition parameters based on only a small amount of initial detection data under changing operating conditions.

[0021] The existing technical solutions have the following main shortcomings: (1) It is time-consuming and inefficient. Parameter optimization usually requires the collection and comparison of multiple candidate parameters or parameter combinations. When parameters such as delay time and integration time are optimized together, the number of experiments increases exponentially, which makes it difficult to meet the needs of online detection and rapid parameter adjustment.

[0022] (2) The sample consumption is relatively large. Each laser excitation will cause ablation on the sample surface. A large number of repeated experiments will significantly increase the sample consumption, which is especially unfavorable for the detection of precious samples, trace samples or limited test areas.

[0023] (3) Traditional improvement methods have limited optimization accuracy. To avoid the time-consuming problem of full scanning, existing technologies often use low-density sampling, statistical screening, or empirical setting. However, due to insufficient coverage of candidate parameters, limited sampling density, or strong rule dependence, it is easy to end up selecting suboptimal parameters.

[0024] Therefore, how to quickly and accurately determine the optimal acquisition parameters of the LIBS system under the current operating conditions based on a small amount of initial detection data, and realize the automatic adjustment and / or parameter guidance of subsequent acquisition parameters, has become a technical problem that urgently needs to be solved in this field.

[0025] This disclosure provides a method for predicting optimal acquisition parameters for a LIBS system, which can be applied to element detection processes in scenarios such as industrial process monitoring, geological exploration, and environmental monitoring.

[0026] See Figure 1 This is a flowchart illustrating a method for predicting optimal acquisition parameters in a LIBS system according to an embodiment of this disclosure. Figure 1 As shown, the method mainly includes the following steps: Step 1: Obtain the initial detection data of the sample to be tested.

[0027] In practice, the initial detection data can be acquired in real time by the detection device or retrieved from storage devices, databases, or other data sources. The initial detection data includes spectral data and / or non-spectral observation data. The spectral data may include spectra obtained from single or multiple detections, spectra obtained at different acquisition times, and / or spectral sequences; the non-spectral observation data may include plasma radiation time-domain data and other data capable of characterizing the plasma state or evolution process. For example, the initial detection data may include one or more of the following: early spectra, sparse delayed spectra, continuously or discretely acquired spectral sequences, and plasma radiation time-domain curves, but is not limited to these.

[0028] It should be noted that, when establishing and / or updating the correlation model, the initial detection data may also include the optimal acquisition parameters corresponding to the spectral data and / or non-spectral observation data. The initial detection data obtained during the online prediction phase of the sample to be tested does not necessarily require the inclusion of the optimal acquisition parameters.

[0029] Step 2: Obtain the association model corresponding to the initial detection data.

[0030] In specific implementation, the association model includes the method for constructing the feature information corresponding to the initial detection data, and the association relationship between the feature information and the optimal acquisition parameters. The association model can be pre-built based on sample data and called during prediction, or it can be selected from multiple pre-built association models based on the initial detection data. Alternatively, it can be built based on the initial detection data or the pre-built association model can be updated.

[0031] The feature information may consist of the initial detection data itself, or it may include signal quality features, physical features, temporal evolution features, dimensionality reduction representation features, and / or derived features obtained from the initial detection data. The features obtained from the initial detection data can be acquired through explicit feature extraction and / or implicit feature learning. In some embodiments, the feature information may include signal strength, background strength, signal-to-noise ratio, signal-to-background ratio, temporal evolution parameters, electron temperature characterization, electron density characterization, etc., but is not limited thereto.

[0032] The relationships can be established through methods such as data fitting, data mapping, pattern recognition, instance matching, and / or state determination. The model used to establish these relationships can be based on regression, kernel methods, tree structures, neural networks, instance matching, classification mapping, lookup tables, or state machines, or other models capable of establishing these relationships. The model used to establish these relationships can include a single model or multiple models operating in a predetermined order, hierarchy, or collaborative manner. The training and / or updating methods for the model used to establish these relationships can include offline training, online incremental learning, reinforcement learning, and / or feedback adjustment.

[0033] The sample dataset used to establish and / or update the association model can cover one or more detection conditions, including changes in sample type, changes in sample state, changes in equipment state, and changes in detection environment; the sample type can include metals, ores, soil, coal, and other samples suitable for detection using LIBS technology.

[0034] Step 3: Based on the correlation model, predict the acquisition parameters that will enable the preset spectral quality index to reach the optimal level.

[0035] In specific implementation, feature information is obtained based on the association model corresponding to the initial detection data, and calculations, mappings, matching, state determinations, and / or feedback corrections are performed based on the correlation between the feature information and the optimal acquisition parameters to determine the optimal acquisition parameters corresponding to the current detection condition. The prediction method may include direct prediction, hierarchical prediction, mapping prediction, and / or feedback correction prediction.

[0036] The preset spectral quality indicators may include characteristic peak signal intensity, background interference index, signal-to-noise ratio, signal-to-background ratio, analytical accuracy, and / or multi-objective comprehensive evaluation index. The optimal spectral quality index can be achieved by achieving the optimal value of a single spectral quality index, the optimal value of a comprehensive evaluation index of multiple spectral quality indicators, or by achieving the optimal value under preset parameter ranges, equipment operating conditions, detection time, or other constraints.

[0037] The optimal acquisition parameters may include delay time, integration time, gain, and / or gate width, as well as other parameters that affect the acquisition status of the detection data. The prediction result can be expressed as acquisition parameter values, acquisition parameter ranges, and / or parameter corrections relative to the current acquisition parameters. When multiple acquisition parameters need to be determined, each acquisition parameter can be predicted separately, or multiple acquisition parameters can be predicted jointly.

[0038] In this disclosure, the optimal acquisition parameters refer to the preferred acquisition parameters determined according to preset spectral quality indicators, evaluation rules and applicable constraints, including global optimal parameters, local optimal parameters, comprehensive evaluation optimal parameters or acquisition parameters that meet preset selection conditions.

[0039] Step 4: Output the optimal acquisition parameters.

[0040] In practice, the optimal acquisition parameters can be output in the form of control output, data output, and / or guidance output. For example, the predicted optimal acquisition parameters can be sent to the spectral acquisition device, delay control device, or other control device to adjust the acquisition state of subsequent detection; they can also be sent to local or remote devices, systems, programs, or storage media for other devices, systems, or programs to call; guidance output can also be provided through display, prompts, or generation of parameter configuration suggestions.

[0041] Correspondingly, this disclosure also provides a prediction device and a laser-induced breakdown spectroscopy detection system for implementing the above-described method. The prediction device can be implemented through software, firmware, hardware logic, or a combination thereof, and can be deployed in local computing devices, servers, cloud platforms, embedded or edge computing devices, and programmable or dedicated hardware platforms. The laser-induced breakdown spectroscopy detection system includes a prediction unit for executing the above-described prediction logic. The prediction unit can predict optimal acquisition parameters based on the initial detection data of the sample to be tested and output the optimal acquisition parameters. In some embodiments, the prediction unit can be linked with a control unit, which can adjust the acquisition state of subsequent detections based on the predicted optimal acquisition parameters to achieve adaptive adjustment of the acquisition parameters.

[0042] Optionally, the embodiments of this disclosure are based on the acquisition parameter prediction process of model training and online software forward inference. The specific implementation logic is as follows: a two-stage software computing architecture of "model training - online invocation" is adopted. Feature information is constructed through the full amount of experimental data, and a continuous mapping regression model of "feature information and optimal acquisition parameters" is established. In the online state, the computer or main control software directly predicts and explicitly outputs the optimal acquisition parameter value under the current working condition through forward inference based on the very small amount of initial detection data features extracted in real time.

[0043] Alternative Solution 1: Implicit Closed-Loop Control Architecture Based on Hardware Instructions Implementation logic: The software-based "feature extraction" and "model inference" processes are transformed into underlying hardware logic. This involves compiling and programming the pre-trained correlation model into devices such as Field Programmable Gate Arrays (FPGAs) or Application-Specific Integrated Circuits (ASICs).

[0044] Equivalent point explanation: This scheme breaks down the software interaction boundary between prediction and input, directly acquiring photoelectric signals at the hardware level and instantly completing logical operations, directly overwriting the system timing control pins through registers. While it does not involve explicit model calculations and parameter output instructions, it essentially utilizes the same correlation principles to achieve an equivalent, extremely fast automatic optimization control effect.

[0045] Alternative Solution 2: Optimization and Dimensionality Reduction Architecture Based on Discrete State Machines and Mapping Tables Implementation logic: Instead of constructing a regression model for continuous numerical prediction, a discrete state machine (classifier) ​​containing a finite number of typical operating conditions is built in the offline phase, and the corresponding optimal acquisition parameters are fixed into a static lookup table. In the online phase, only the initial detection data is used to evaluate the current "classification state" of the system, and then the corresponding parameter set is directly mapped and retrieved.

[0046] Equivalent point explanation: This scheme reduces high-dimensional mathematical calculations to a "state classification + condition triggering" logic. Although it sacrifices the prediction accuracy of the continuous parameter space, it also bypasses the traditional physical traversal scanning, achieving the core objectives of reducing optimization time and minimizing sample loss.

[0047] Alternative Solution 3: Online Adaptive Optimization Architecture Based on Real-Time Interactive Feedback Implementation logic: The boundary between "model training and online invocation" is weakened, and an online incremental learning or reinforcement learning paradigm is adopted. The system has built-in basic model rules. After acquiring the initial detection data, the system combines the quality feedback from a very small number (e.g., 1-3) of online exploratory excitations to iterate and converge the model in real time, thereby directly locking in the optimal acquisition parameters.

[0048] Equivalent point explanation: This scheme integrates "offline training and modeling" with "online parameter application" in terms of execution timing. Although it adds a very small number of exploratory activation steps, it still falls within the category of inference optimization using data features, and its optimization efficiency and sample protection effect are equivalent to the main scheme.

[0049] The optimal acquisition parameter prediction method for the LIBS system provided in this embodiment reconstructs the traditional "exhaustive search" logic, which relies on hardware stepping and physical excitation, into an "algorithm reasoning" process based on an association model. The time required to determine the optimal acquisition parameters changes from depending on the spectrometer acquisition cycle and mechanical movement frequency to a millisecond-level processing cycle of the computing unit. Therefore, the system requires only a very short computational response to complete the optimization, reducing the total time from minutes / hours to less than seconds. The association model establishes a "high information density" data mapping mechanism, requiring only a single or single-digit number of transient signals generated by initial excitations to extract characteristic information representing the physical evolution trend. This paradigm eliminates the need for physical scanning of the global parameter space, minimizing the physical ablation requirements of the sample during the optimization process, thereby significantly reducing the loss of trace or precious samples. There is an essential dynamic mapping relationship between the initial detection data and the optimal acquisition parameters. By learning the evolution of plasma cooling, the correlation model can capture changes in data characteristics and map out matching parameter corrections in real time, even when environmental conditions deviate. This ensures that the acquisition status is always locked in the physical optimal range, making the predicted indicators highly close to the actual theoretical optimal values.

[0050] The method of this disclosure will be further described below with reference to several embodiments: Example 1: A method, apparatus, and system for predicting aluminum alloy acquisition parameters based on early and sparse delayed spectra. 1. Experimental Apparatus and System The detection system includes a 1064 nm Q-switched Nd:YAG pulsed laser (single pulse energy 32 mJ to 42 mJ, pulse width 6 ns, repetition rate 10 Hz), a 100 mm focal length focusing lens, a three-dimensional displacement platform, a gated spectrometer (wavelength range 190 nm to 1000 nm, resolution 0.08 nm, delay time resolution 1 ns), a digital delay generator, and a local industrial control computer. The integration time is 1.00 μs.

[0051] 2. Specific operating steps Step S1: Obtain initial detection data for the sample to be tested. Training sample data collection: The training samples consisted of standard aluminum alloy blocks (30 mm × 30 mm × 5 mm) of three alloys: 5A06, 6061, and 7075. The surfaces were polished with 800-grit sandpaper and then cleaned with anhydrous ethanol. During the training phase, laser energies were set to 32 mJ, 34 mJ, 36 mJ, 38 mJ, 40 mJ, and 42 mJ, with defocusing amounts of -0.8 mm, 0 mm, and +0.8 mm. Ten spectral acquisitions were performed for each sample and each set of experimental parameters. Stepped delay scans were performed for each set of experimental parameters, with delay times ranging from 0.10 μs to 3.50 μs in 0.05 μs increments. Target spectral lines were selected as Mg II 279.55 nm, Mg II 280.27 nm, and Si I 288.16 nm. Three spectra were acquired at each delay time, and their average value was calculated.

[0052] Online initial detection data acquisition: First, a 100 ns early spectrum of the sample to be tested is acquired; then, based on the upper limit of the predicted effective observation time, five sparse delayed spectra (200 ns, 500 ns, 1200 ns, 2000 ns, 3000 ns) are acquired at five sampling points within this upper limit in a denser-then-sparser manner.

[0053] Step S2: Obtain the association model corresponding to the initial detection data. Tag building: (1) Optimal delay time label: Take the delay time that is optimal for the signal-to-background ratio of the three spectral lines. If there are multiple candidate points, select the delay point that maximizes the minimum signal-to-background ratio among the three.

[0054] (2) Effective observation time upper limit label: the earliest delay time to reduce the signal-to-background ratio of all three spectral lines to below 80% of their respective peak values.

[0055] Feature extraction rules and feature space construction: (1) Signal quality characteristics: signal-to-noise ratio and signal-to-background ratio of the three target spectral lines in the early spectrum.

[0056] (2) Macroscopic characteristics: full spectrum integral intensity, 200 nm to 300 nm band integral intensity, 300 nm to 450 nm band integral intensity, and continuous background average intensity.

[0057] (3) Physical characteristics: The initial electron temperature and electron density characterization based on Al atomic lines and ionic lines constructed by Boltzmann diagram method and Saha-Boltzmann relation.

[0058] (4) Derived temporal evolution characteristics: The fast and slow decay time constants, weights, and other features are extracted from the spectral intensity, temperature, and density sequences at the 5 sparse nodes through double exponential fitting.

[0059] The association model yielded: The association between the feature information and the corresponding labels is established using a two-level random forest regression model. The first level takes early (100 ns delay) spectral features as input and the upper limit label of the effective observation time as output; the second level takes the constructed multidimensional feature vector as input and the optimal delay time label as output. The model is saved after training.

[0060] Step S3: Based on the correlation model, predict the acquisition parameters that will optimize the preset spectral quality index. During the online phase, early spectral features (100 ns delay time) are input into the first-level model to predict the upper limit of the effective observation time. After acquiring sparse spectra, the constructed multidimensional feature vector is input into the second-level model to calculate and predict the optimal delay time so that the signal-to-background ratio of the spectral feature peaks reaches the optimal level.

[0061] Step S4: Output the predicted optimal acquisition parameters The system outputs the predicted optimal delay time through the communication interface and writes it into the digital delay generator to adjust the working status of subsequent formal data acquisition.

[0062] 3. Comparative Verification (1) Time spent on parameter acquisition and optimization: Traditional full-scan strategies adapt to 18 typical operating conditions, with a total time consumption of approximately 1.5 to 2 hours. In this embodiment, online parameter updates require only 6 triggers plus model inference, with the time consumption controlled within 10 seconds, reducing the total time consumption by more than 95%.

[0063] (2) Sample loss: Traditional traversal optimization requires 69 delay points × 3 pulses, totaling 207 excitations; this embodiment only requires 6 excitations to complete the prediction, reducing the sample ablation amount by more than 97%.

[0064] (3) Prediction of optimal acquisition parameters: In the test of sample 6061, the predicted optimal delay was 1.60 μs, and the true optimal delay for the full scan was 1.55 μs. The deviation between the spectral quality index of the predicted value and the true optimal index was only 3.2%.

[0065] Example 2: Ore Detection Based on Early Spectra and Total Radiation Time-Domain Curves 1. Experimental conditions The detection system includes a 355 nm frequency-triple Nd:YAG pulsed laser, a focusing optical system, a beam splitting assembly, a gated spectral acquisition unit, a high-speed silicon photodiode, a high-speed acquisition card (1 GS / s), a digital delay generator, and an edge computing workstation.

[0066] 2. Specific operating steps Step S1: Obtain initial detection data for the sample to be tested. Training sample data collection: Twelve types of phosphate rock were used, with laser single-pulse energy ranging from 10 to 20 mJ and defocusing from -2 to +2 mm. Two-dimensional scanning was performed with a delay of 0.10–2.50 μs (step size 0.05 μs) and an integration of 0.50–3.00 μs (step size 0.25 μs). Target spectral lines were PI 213.62 nm and Ca II 393.37 / 396.85 nm. The total plasma intensity radiation time-domain curves were acquired simultaneously with a delay of 200 ns and a gate width of 3000 ns.

[0067] Online initial detection data acquisition: An initial observation was performed to obtain an early spectrum with a delay of 200 ns and a gate width of 3000 ns, and a time-domain curve of total plasma luminescence recorded simultaneously.

[0068] Step S2: Obtain the association model corresponding to the initial detection data. Tag building: The delay time and integration time labels that have the best signal-to-noise ratio among the three spectral lines are selected as the optimal acquisition parameters. If multiple candidate points exist, the delay time and integration time that maximize the minimum signal-to-noise ratio among the three are chosen.

[0069] Feature extraction rules and feature space construction: (1) Signal quality characteristics: target spectral line signal-to-noise ratio and signal-to-background ratio in the early spectrum with a delay time of 200 ns and a gate width of 3000 ns.

[0070] (2) Macroscopic characteristics: the integrated intensity of each band of the early spectrum with a delay time of 200 ns and a gate width of 3000 ns, and the peak intensity and full width at half maximum of the total light intensity radiation time domain curve.

[0071] (3) Physical characteristics: Combining Boltzmann and Saha-Boltzmann equations, the initial electron temperature and electron density characterizations are derived from the Ca spectral line inversion based on the early spectrum with a delay time of 200 ns and a gate width of 3000 ns.

[0072] (4) Derived temporal evolution characteristics: The fast and slow decay amplitudes, time constants and their ratios and reciprocals are extracted by double exponential fitting of the total light intensity radiation time domain curve.

[0073] The association model yielded: The constructed multidimensional feature vector is used as input, and the delay time and integral time with the best signal-to-noise ratio of the three spectral lines are used as output. The delay time regressor and integral time regressor are trained using an extreme random tree regression model.

[0074] Step S3: Based on the correlation model, predict the acquisition parameters that will optimize the preset spectral quality index. The multidimensional feature vector constructed under the current operating conditions is input into two regressors to predict the optimal delay time and the optimal integration time.

[0075] Step S4: Output the predicted optimal acquisition parameters The predicted optimal delay time and integration time are output and sent to the digital delay generator and the gated spectrum acquisition unit.

[0076] 3. Comparative Verification (1) Time spent on parameter acquisition and optimization: Traditional 2D full-scan adaptation to the conditions described in this embodiment takes approximately 10 hours. This embodiment updates parameters online in under 8 seconds, resulting in a total time reduction of over 99%.

[0077] (2) Sample loss: Traditional two-dimensional scanning requires hundreds of excitations; this embodiment only requires one synchronous observation (one excitation) to complete the prediction, reducing the sample ablation amount by more than 99%.

[0078] (3) Prediction of optimal acquisition parameters: In the +1.5 mm defocus test, the predicted optimal delay was 0.90 μs and the integral delay was 1.73 μs. The nominal optimal delays for the full scan were 0.85 μs and 1.75 μs, respectively, with a spectral quality index deviation of less than 4.6%.

[0079] Example 3: Soil Heavy Metal Detection Based on Local Loading of Pre-trained Model 1. Experimental conditions A portable, integrated LIBS device includes a 1064 nm pulsed laser source (25-35 mJ), a small gated spectral acquisition unit, an embedded main control board, and an internal timing control module. Target spectral lines: PbI 405.78 nm, CrI 425.43 nm, FeI 404.58 nm.

[0080] 2. Specific operating steps Step S1: Obtain initial detection data for the sample to be tested. Training sample data collection: Ten soil types were studied, with moisture contents ranging from 2% to 15% and different compaction degrees. The time delay scan ranged from 0.20 to 2.20 μs (step size 0.05 μs).

[0081] Online initial detection data acquisition: One 200 ns early spectrum was acquired, along with seven sparse delayed spectra acquired within the upper limit of the predicted effective observation time.

[0082] Step S2: Obtain the association model corresponding to the initial detection data. Tag building: Similar to Example 1, we construct an effective observation time upper limit label and an optimal delay time label.

[0083] Feature extraction rules and feature space construction: (1) Signal quality characteristics: initial signal-to-noise ratio and signal-to-background ratio of the three spectral lines.

[0084] (2) Macroscopic characteristics: full spectrum / band integral intensity, continuous background average value.

[0085] (3) Physical characteristics: The initial electron temperature and electron density characterization based on the Cr and Fe spectral lines are constructed by combining the Boltzmann and Saha-Boltzmann equations.

[0086] (4) Derived temporal evolution characteristics: The temperature and density sequences at the 7 sparse nodes were fitted using a single exponential model to extract features such as temperature and density decay time constants and their ratios.

[0087] The association model yielded: The first stage uses a random forest regression model; the second stage uses a radial basis function (RBF) kernel support vector regression model. The first stage takes early (200 ns delay) spectral features as input and an upper limit label for the effective observation time as output; the second stage takes a constructed multidimensional feature vector as input and an optimal delay time label as output. The first and second stage models establish the correlation between the corresponding feature information and the corresponding labels. The model is saved after training.

[0088] Step S3: Based on the correlation model, predict the acquisition parameters that will optimize the preset spectral quality index. The constructed multidimensional feature vector is input into the association model to predict the optimal delay time.

[0089] Step S4: Output the predicted optimal acquisition parameters The predicted optimal delay time is output to the device's internal timing control module.

[0090] 3. Comparative Verification (1) Time spent on parameter acquisition and optimization: Traditional scanning to adapt to various soil conditions takes several hours, while this embodiment controls online updates to within 10 seconds, reducing the total time consumption by more than 99%.

[0091] (2) Sample loss: Traditional optimization requires hundreds of excitations, while this embodiment only requires 8 excitations (1 early stage + 7 sparse stages), reducing the sample ablation by more than 90%.

[0092] (3) Prediction of optimal acquisition parameters: In the test with a moisture content of 12%, the predicted optimal delay was 1.05 μs, the actual optimal delay for dense scanning was 1.00 μs, and the deviation between the spectral quality index under the predicted value and the actual optimal value was controlled within 5%.

[0093] Experiments have shown that the beneficial effects of the method disclosed herein are as follows: (1) The time spent on parameter optimization has been greatly reduced: the time spent on parameter optimization has been shortened from minutes / hours in the traditional full parameter scan to less than seconds, and the total time has been reduced by more than 95%.

[0094] (2) Sample loss is significantly reduced: the number of laser excitations required for parameter optimization is reduced from hundreds to single digits, and the sample ablation is reduced by more than 90%.

[0095] (3) The optimal acquisition parameters are accurately predicted: the spectral quality index acquired under the predicted optimal acquisition parameters has a deviation of less than 5% compared with the actual objective situation.

[0096] Corresponding to the above method embodiments, this disclosure also provides an optimal acquisition parameter prediction device for a LIBS system, comprising: The data module is used to acquire the initial detection data of the sample to be tested; The model module is used to obtain the association model corresponding to the initial detection data; The prediction module is used to predict the acquisition parameters that will enable the preset spectral quality index to reach the optimal level, based on the correlation model. The output module is used to output the optimal acquisition parameters.

[0097] The device operates in the form of software scripts, algorithm plugins, desktop programs, web application services, firmware instructions, hardware logic, or combinations thereof, and supports deployment and execution in computers, servers, cloud platforms, embedded main control boards, edge computing devices, field-programmable gate arrays, and application-specific integrated circuits. The device can execute the contents of the above method embodiments accordingly. For parts not described in detail in this embodiment, refer to the contents recorded in the above method embodiments, and will not be repeated here.

[0098] Corresponding to the above method embodiments, this disclosure also provides a laser-induced breakdown spectroscopy detection system, including: The excitation unit is used to excite the sample under test to generate a plasma signal. The control unit is used to adjust the system's operating status based on the collected parameters; The acquisition unit is used to acquire and output the initial detection data of the sample to be tested; The prediction unit is used to obtain the association model corresponding to the initial detection data, and predict the acquisition parameters that make the preset spectral quality index reach the optimal based on the association model. The output unit is used to output the optimal acquisition parameters.

[0099] The system can execute the contents of the above method embodiments. For the parts not described in detail in this embodiment, please refer to the contents recorded in the above method embodiments, and will not be repeated here.

[0100] This disclosure also provides an electronic device, comprising: at least one processor and a memory communicatively connected to the at least one processor. The memory stores instructions executable by the at least one processor, which, when executed, enable the at least one processor to perform the LIBS system optimal acquisition parameter prediction method described in the foregoing method embodiments.

[0101] This disclosure also provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute the LIBS system optimal acquisition parameter prediction method in the foregoing method embodiments.

[0102] This disclosure also provides a computer program product, which includes a computing program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions that, when executed by a computer, cause the computer to perform the LIBS system optimal acquisition parameter prediction method in the foregoing method embodiments.

[0103] It should be understood that the various parts of this disclosure can be implemented in hardware, software, firmware, or a combination thereof.

[0104] The above description is merely a specific embodiment of this disclosure, but the scope of protection of this disclosure is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this disclosure should be included within the scope of protection of this disclosure. Therefore, the scope of protection of this disclosure should be determined by the scope of the claims.

Claims

1. A method for predicting optimal acquisition parameters in a LIBS system, characterized in that, include: Step 1: Obtain the initial detection data of the sample to be tested; Step 2: Obtain the association model corresponding to the initial detection data; Step 3: Based on the correlation model, predict the acquisition parameters that will optimize the preset spectral quality index. Step 4: Output the optimal acquisition parameters.

2. The method according to claim 1, characterized in that, The initial detection data in step 1 includes spectral data and / or non-spectral observation data; when the initial detection data is used to establish and / or update the correlation model, it also includes the optimal acquisition parameters corresponding to the spectral data and / or non-spectral observation data. Spectral data and / or non-spectral observation data include at least one of real-time acquired data and historical data; The optimal acquisition parameters are the result of selecting at least one acquisition control quantity among delay time, integration time, gain, and gate width based on a preset spectral quality index. The preset spectral quality indicators include at least one of the following: characteristic peak signal intensity, background interference index, signal-to-noise ratio, signal-to-background ratio, analysis accuracy, and multi-objective comprehensive evaluation index; The parameter results may take the form of at least one of parameter values, parameter ranges, and parameter correction amounts.

3. The method according to claim 1, characterized in that, The association model in step 2 includes the method for constructing the feature information corresponding to the initial detection data, and the association relationship between the feature information and the optimal acquisition parameters; The method of obtaining the association model includes at least one of the following: establishing the association model based on the initial detection data, matching and selecting the association model corresponding to the initial detection data from the pre-established association models, and updating the pre-established association model based on the initial detection data.

4. The method according to claim 3, characterized in that, The feature information includes at least one of the initial detection data itself and features obtained based on the initial detection data; The features obtained based on the initial detection data include at least one of signal quality features, physical features, temporal evolution features, dimensionality reduction representation features, and derived features obtained from the signal quality features, physical features, temporal evolution features, and dimensionality reduction representation features; The methods for obtaining features based on the initial detection data include at least one of explicit feature extraction and implicit feature learning.

5. The method according to claim 3, characterized in that, The association is established through at least one of the following methods: data fitting, data mapping, pattern recognition, instance matching, and state determination. The model used to establish the association includes at least one of the following: regression, kernel method, tree structure, neural network, instance matching, classification mapping, lookup table, and state machine. The training and / or updating methods for the model used to establish the association include at least one of offline training, online incremental learning, reinforcement learning, and feedback adjustment; The detection conditions covered by the sample dataset used to establish and / or update the association model include at least one of the following: sample type change, sample state change, equipment state change, and detection environment change; the sample type includes at least one of metal, ore, soil, and coal.

6. The method according to claim 3, characterized in that, The prediction in step 3 includes obtaining feature information based on the initial detection data, and determining the optimal acquisition parameters according to the correlation relationship between the feature information and the optimal acquisition parameters in the association model; The scenario in which the preset spectral quality index reaches its optimal state includes at least one of the following: a single spectral quality index reaches its optimal state, or a comprehensive evaluation index of multiple spectral quality indices reaches its optimal state. The prediction method includes at least one of direct prediction, hierarchical prediction, mapping prediction, and feedback correction prediction.

7. The method according to claim 1, characterized in that, The output in step 4 includes at least one of control output, data output, and guidance output.

8. A device for predicting optimal acquisition parameters for a LIBS system, used to execute the prediction method according to any one of claims 1 to 7, characterized in that, include: The data module is used to acquire the initial detection data of the sample to be tested; The model module is used to obtain the association model corresponding to the initial detection data; The prediction module is used to predict the acquisition parameters that will enable the preset spectral quality index to reach the optimal level, based on the correlation model. The output module is used to output the optimal acquisition parameters.

9. The apparatus according to claim 8, characterized in that, The device operates in the form of software scripts, algorithm plugins, desktop programs, web application services, firmware instructions, hardware logic, or combinations thereof, and supports deployment and execution in computers, servers, cloud platforms, embedded main control boards, edge computing devices, field-programmable gate arrays, and application-specific integrated circuits.

10. A laser-induced breakdown spectroscopy detection system for performing the prediction method according to any one of claims 1 to 7, characterized in that, include: The excitation unit is used to excite the sample under test to generate a plasma signal. The control unit is used to adjust the system's operating status based on the collected parameters; The acquisition unit is used to acquire and output the initial detection data of the sample to be tested; The prediction unit is used to obtain the association model corresponding to the initial detection data, and predict the acquisition parameters that make the preset spectral quality index reach the optimal based on the association model. The output unit is used to output the optimal acquisition parameters.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the system optimal acquisition parameter prediction method as described in any one of claims 1 to 7.

12. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by the processor, they implement the system optimal acquisition parameter prediction method as described in any one of claims 1 to 7.

Images