CN106290263A - A kind of LIBS calibration and quantitative analysis method based on genetic algorithm - Google Patents
A kind of LIBS calibration and quantitative analysis method based on genetic algorithm Download PDFInfo
- Publication number
- CN106290263A CN106290263A CN201510259910.6A CN201510259910A CN106290263A CN 106290263 A CN106290263 A CN 106290263A CN 201510259910 A CN201510259910 A CN 201510259910A CN 106290263 A CN106290263 A CN 106290263A
- Authority
- CN
- China
- Prior art keywords
- spectral line
- libs
- quantitative analysis
- genetic algorithm
- line
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Investigating, Analyzing Materials By Fluorescence Or Luminescence (AREA)
Abstract
The present invention relates to a kind of LIBS calibration and quantitative analysis method based on genetic algorithm, concretely comprise the following steps: 1) obtain LIBS spectroscopic data;2) elemental characteristic spectrum to be measured is obtained;3) parameter coding, forms genetic algorithm initial population;4) each ideal adaptation degree in population is calculated;5) by selecting, intersect and mutation probability formation new population;6) repeat 4), 5) to meeting termination condition, the optimum spectral line of output (spectral line to) position;7) (internal calibration) quantitative analysis is calibrated according to optimum spectral line (spectral line to).The optimum spectral line that obtains using this method (spectral line to) is as analytical line (analytical line and reference line), it is possible to achieve quantitative analysis accurate to concentration of element to be measured.Have an advantage in that without artificial selection's analytical line (reference line), the element spectral line (spectral line to) of high coefficient of determination (R2), low detection limit (LOD) and low relative standard deviation (RSD) can be found accurately as analytical line (analytical line and reference line).
Description
Technical field
The invention belongs to spectrum analysis and material component analysis field, be specifically a kind of based on
The LIBS calibration and quantitative analysis method of genetic algorithm.
Background technology
LIBS (LIBS) analytical technology, is a kind of to use pulse laser as energy source
Emission spectrographic analysis technology, it is possible to achieve the qualitative and quantitative analysis of material chemical element.It has without system
Sample, directly quickly, the feature such as sample loss amount is little, become in recent years metallurgical analysis, historical relic's protection,
The study hotspot in the fields such as matter chemistry, environmental project.
Calibration (internal calibration) tracing analysis method is by measuring the LIBS of concentration known standard sample, drawing
The relation curve (calibration curve) of the intensity of spectral line concentration of element, after measuring analysis sample spectrum intensity,
The quantitative analysis method of the concentration of element is directly obtained by relation curve.
As most basic quantitative analysis method, either basic calibration analysis method based on single the intensity of spectral line
Being also based on the internal calibration analytic process of analytical line and reference line pair, what it relied primarily on is analytical line (reference line)
Selection.Select key the most accurate, that disturb little spectral line to be scaling method.Traditional analytical line (reference
Line) it is by observing spectral line by analysis personnel, carry out selecting in conjunction with spectra database and experience.With
The increase of LIBS measurement data amount, the method inefficiency of this artificial selection's analytical line (reference line),
Substantially cannot find the spectral line of global optimum, consequent calibration curve carries out quantitative analysis to measuring samples
It is extremely difficult to good and stable effect.
Summary of the invention
For above-mentioned weak point present in prior art, the technical problem to be solved in the present invention is to provide one
Plant and can automatically search for by optimum spectral line, and it is fixed to utilize the optimum spectral line obtained to realize calibration curve method concentration of element
The LIBS calibration and quantitative analysis method based on genetic algorithm of component analysis.
The technical scheme is that a kind of LIBS based on genetic algorithm of the present invention is fixed
Demarcate analysis method, comprise the following steps:
Step 1: obtain the LIBS data of standard sample, determine wave-length coverage;
Step 2: be loaded into characteristic spectral line data base, read element to be measured in the wave-length coverage that step 1 determines
All characteristic spectrum positional informationes, peak-seeking near the characteristic spectral line position of characteristic spectral line data base's corresponding element,
Determine characteristic spectral line particular location corresponding in the LIBS data of measurement;
Step 3: according to LIBS data length, select spectral line quantity to determine coding figure place, form genetic algorithm
Initial population;
Step 4: using the weighted sum of coefficient of determination R2, detection limit LOD and relative standard deviation RSD as suitable
Response function, the individuality that in finding population, optimum spectral line is corresponding;
Step 5: select initial population, intersect and mutation operation, inserts former by the new individual weight obtained
Population forms new population;
Step 6: repeat step 4, step 5, until genetic algorithm meets termination condition, terminate whole algorithm
Process, the optimum position of spectral line that output finally gives;
Step 7: the optimum spectral line of application carries out calibration and quantitative analysis to concentration of element to be measured.
Described characteristic spectral line data base is Given information, high purity substance measure or atomic emission spectrum data
Storehouse obtains.
Also include: to the characteristic spectrum positional information obtained in step 2, in the phase of testing sample LIBS data
Answer and near position, do peak-seeking process.
Described according to LIBS data length, select spectral line quantity to determine coding figure place, particularly as follows: for growing
Degree is for selecting n bar spectral line in the data of length, its coding figure place is by determiningDetermine.
Described fitness function is:
Wherein R2For coefficient of determination,For coefficient of determination penalty threshold, LOD is detection limit, LOD0For detection
Limit penalty threshold, RSD is relative standard deviation, RSD0For relative standard deviation's penalty threshold, a, b, c divide
It it is not the index of penalty factor corresponding to three parameters.
Described step 5 is according to LIBS data length selected population size, crossover probability, mutation probability, selection
Probability, forms new population.
Described crossover probability between 0.7~0.9, mutation probability between 0.05~0.15, select probability
Between 0.7~0.9.
Described termination condition is: after the evolution of some generations, adaptive optimal control degree individuality is not changed in or reaches to arrange
Macroevolution algebraically.
The present invention has the following advantages and beneficial effect:
1. it is not required to by observing LIBS measure spectrum artificial selection's analytical line (reference line), only need to be by arranging
The reasonably i.e. available corresponding optimum spectral line of fitness function, the spectral line so obtained has more global optimum
Property, by its calibration curve set up, the relation between itself and concentration of element can be described more accurately.
2., by changing genetic algorithm chromosome coding structure and length, adjust fitness function accordingly, can
To select different element, a plurality of spectral line of different evaluation index flexibly, i.e. can be used for the quantitative of unitary variant
Analysis method, can be again that multivariable quantitative analysis method provides primary data.
3. the method applied in the present invention, is application genetic algorithm, believes in conjunction with characteristic spectral line in spectra database
Breath, arranges rational fitness function, unit to be measured in reaching automatically to search for spectral region by Evolution of Population
The purpose of the optimum spectral line of element, finally provides the optimum spectral line in the range of LIBS measures, as analytical line (ginseng
Examine line) set up calibration analysis curve concentration of element is carried out quantitative analysis.
Accompanying drawing explanation
Fig. 1 is the inventive method flowchart;
Fig. 2 is that low-alloy steel sample optimum spectral line is to calibration curve quantitative analysis result the inventive method flow process
Figure.
Detailed description of the invention
Below in conjunction with the accompanying drawings and embodiment the present invention is described in further detail.
The present invention is directed to different elemental characteristic position of spectral line in LIBS signal different, the same element of variable concentrations
The feature that characteristic spectral line intensity level is different, finds optimal characteristics position of spectral line, and it is right to determine according to its intensity of spectral line
Answer concentration of element.
As it is shown in figure 1, method reads the experimental spectrum data obtained by LIBS experiment porch as defeated after starting
Enter, find each spectral line that element to be measured is corresponding, by genetic algorithm optimization, select optimum position of spectral line,
And set up calibration curve by it concentration of element to be measured is analyzed.Implement step as follows:
Step 1: obtain the LIBS data of sample, determine wave-length coverage.The sample being directed to is standard
Sample, unit to be measured have actual concentrations;LIBS data are measured by experiment and are obtained.
Step 2: be loaded into characteristic spectral line data base, read element to be measured in the wave-length coverage that step 1 determines
All characteristic spectrum positional informationes.Characteristic spectral line data base is measured by high purity substance or atomic emission spectrum number
Obtain according to storehouse, for standard feature spectral line data storehouse, it is adaptable to all samples.Corresponding characteristic spectral line data base
Peak-seeking near the characteristic spectral line position of element, determines that characteristic spectral line corresponding in the LIBS data of measurement is concrete
Position.
Due to experimental situation and parameter, the difference of operation, measure the LIBS data character pair spectral line position obtained
Can there is skew in the spectral peak put, grasped by the least displacement peak-seeking of characteristic spectrum position environs in java standard library
Make, find the characteristic spectral line position in concrete LIBS data.
Step 3: determine coding figure place according to the final spectral line quantity obtained, form genetic algorithm initial population.
Step 4: with coefficient of determination (R2), detection limit (LOD) and the weighting of relative standard deviation (RSD)
With as fitness function, find the individuality that optimum spectral line is corresponding.
By adding, delete the parameter participating in evaluating and adjusting parameter weights, can regulate what optimal value was paid close attention to
Emphasis, is more adapted to the fitness function of testing sample characteristic and experimental situation.
Step 5: initial population selects (duplication), intersects and mutation operation, heavily inserts original seed group's shape
Become new population.By retaining the high fitness that the high fitness of parent population is individual and inserts in progeny population
Body forms one to be had higher fitness and is different from the new population of parent population.Select according to LIBS data length
Select Population Size (generally 100), crossover probability (0.7~0.9), mutation probability (0.05~0.15), select
Probability (0.7~0.9), forms new population.
Step 6: repeat step 4, step 5, until genetic algorithm meets termination condition, terminate whole algorithm
Process, optimum spectral line (spectral line to) position that output finally gives.
After algorithm termination condition is traditionally arranged to be N generation evolution, adaptive optimal control degree individuality is not changed in or reaches to arrange
Maximum evolutionary generation, to ensure that the optimal value obtained is as global optimum as far as possible.
Step 8: it is quantitative that concentration of element to be measured is calibrated (internal calibration) by the optimum spectral line of application (spectral line to)
Analyze.
Whole method is in addition to route selection realizes calibration curve quantitative analysis, it is also possible to by changing code length
And structure, adjust fitness function, select a plurality of optimum spectral line, carry for other multivariate quantitative analysis methods
For primary data.
Cr, Ni, Mn, Si tetra-kinds in 10 pieces of low-alloy steel standard sample is analyzed respectively by approach described above
The concentration of element, using wherein 8 pieces of samples as training sample, remains 2 pieces of samples as checking sample, survey
The final effect of method for testing.
Setting fitness function is as follows:
Wherein R2For coefficient of determination, LOD is detection limit, and RSD is relative standard deviation, and a, b, c are respectively
It is the index of penalty factor corresponding to three parameters, sets as 0.95 < R2When≤1, a=0;R2When≤0.95,
A=1.Equally, 0 < during LOD≤1000, b=0;During LOD > 1000, b=1.And during 0≤RSD≤0.1, c=0;
During RSD > 0.1, c=1.I.e. when individual coefficient of determination, detection limit or relative standard deviation are beyond ideal range
Time, give corresponding fitness function part one great penalty factor so that it is to evolve the next generation
In be eliminated, it is ensured that new population contains be all coefficient of determination, detection limit and relative standard deviation meet want
The individuality asked.
According to the Analysis of Genetic Algorithms that above fitness function is corresponding, obtain four kinds of analytical lines corresponding to element and
Its corresponding reference line (Fe characteristic spectral line) is as shown in the table.
According to the analytical line obtained, reference line pair, set up internal calibration curve to four kinds of concentration of element in sample
Carrying out quantitative analysis, result is as in figure 2 it is shown, the most respectively with formulaWith
FormulaThe root-mean-square error and the checking collection root-mean-square that calculate every kind of elementary analysis result miss
Difference (C in formulaiWithBeing respectively the concentration of element actual value in sample i and measured value, t and v is respectively training set
With checking collection sample number), finally give the LIBS internal calibration based on genetic algorithm that application this patent relates to
Quantitative analysis results such as following table, it is seen that this method can effectively reach wanting of quantitative analysis sample coherent element concentration
Ask.
Claims (8)
1. a LIBS calibration and quantitative analysis method based on genetic algorithm, it is characterised in that include following
Step:
Step 1: obtain the LIBS data of standard sample, determine wave-length coverage;
Step 2: be loaded into characteristic spectral line data base, read element to be measured in the wave-length coverage that step 1 determines
All characteristic spectrum positional informationes, peak-seeking near the characteristic spectral line position of characteristic spectral line data base's corresponding element,
Determine characteristic spectral line particular location corresponding in the LIBS data of measurement;
Step 3: according to LIBS data length, select spectral line quantity to determine coding figure place, form genetic algorithm
Initial population;
Step 4: using the weighted sum of coefficient of determination R2, detection limit LOD and relative standard deviation RSD as suitable
Response function, the individuality that in finding population, optimum spectral line is corresponding;
Step 5: select initial population, intersect and mutation operation, inserts former by the new individual weight obtained
Population forms new population;
Step 6: repeat step 4, step 5, until genetic algorithm meets termination condition, terminate whole algorithm
Process, the optimum position of spectral line that output finally gives;
Step 7: the optimum spectral line of application carries out calibration and quantitative analysis to concentration of element to be measured.
A kind of LIBS calibration and quantitative analysis method based on genetic algorithm the most according to claim 1,
It is characterized in that, described characteristic spectral line data base is Given information, high purity substance measure or atomic emissions
Spectra database obtains.
A kind of LIBS calibration and quantitative analysis method based on genetic algorithm the most according to claim 1,
It is characterized in that, also include: to the characteristic spectrum positional information obtained in step 2, at testing sample LIBS
Do peak-seeking near the relevant position of data to process.
A kind of LIBS calibration and quantitative analysis method based on genetic algorithm the most according to claim 1,
It is characterized in that, described according to LIBS data length, select spectral line quantity determine coding figure place, particularly as follows:
For selecting n bar spectral line in the data of a length of length, its coding figure place is by determining
Determine.
A kind of LIBS calibration and quantitative analysis method based on genetic algorithm the most according to claim 1,
It is characterized in that, described fitness function is:
Wherein R2For coefficient of determination,For coefficient of determination penalty threshold, LOD is detection limit, LOD0For detection
Limit penalty threshold, RSD is relative standard deviation, RSD0For relative standard deviation's penalty threshold, a, b, c divide
It it is not the index of penalty factor corresponding to three parameters.
A kind of LIBS calibration and quantitative analysis method based on genetic algorithm the most according to claim 1,
It is characterized in that, described step 5 is general according to LIBS data length selected population size, crossover probability, variation
Rate, select probability, form new population.
A kind of LIBS calibration and quantitative analysis method based on genetic algorithm the most according to claim 6,
It is characterized in that, described crossover probability between 0.7~0.9, mutation probability between 0.05~0.15,
Select probability is between 0.7~0.9.
A kind of LIBS calibration and quantitative analysis method based on genetic algorithm the most according to claim 1,
It is characterized in that, described termination condition is: after the evolution of some generations, adaptive optimal control degree individuality is not changed in or reaches
The maximum evolutionary generation arranged.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510259910.6A CN106290263B (en) | 2015-05-19 | 2015-05-19 | A kind of LIBS calibration and quantitative analysis methods based on genetic algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510259910.6A CN106290263B (en) | 2015-05-19 | 2015-05-19 | A kind of LIBS calibration and quantitative analysis methods based on genetic algorithm |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106290263A true CN106290263A (en) | 2017-01-04 |
CN106290263B CN106290263B (en) | 2018-08-24 |
Family
ID=57633843
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510259910.6A Active CN106290263B (en) | 2015-05-19 | 2015-05-19 | A kind of LIBS calibration and quantitative analysis methods based on genetic algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106290263B (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107887289A (en) * | 2017-11-13 | 2018-04-06 | 北京半导体专用设备研究所(中国电子科技集团公司第四十五研究所) | A kind of method and device for obtaining thin film parameter value to be measured |
CN108414475A (en) * | 2018-01-30 | 2018-08-17 | 中国科学院上海技术物理研究所 | The LIBS analysis methods rebuild based on optical chromatography Simultaneous Iteration |
CN109668862A (en) * | 2017-10-17 | 2019-04-23 | 中国科学院沈阳自动化研究所 | A kind of aluminium electrolyte molecular proportion detection method based on laser induced breakdown spectroscopy |
CN110780021A (en) * | 2019-10-30 | 2020-02-11 | 广船国际有限公司 | Method and device for determining standard substance, terminal and storage medium |
CN112414996A (en) * | 2020-07-24 | 2021-02-26 | 北京工商大学 | Finite difference and difference evolution algorithm-based ICP-AES spectral line overlapping interference correction method |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101915753A (en) * | 2010-07-30 | 2010-12-15 | 浙江师范大学 | Genetic Neural NetworkQuantitative analysis method for laser induced breakdown spectroscopy based on gGenetic Neural Networkgenetic neural network |
CN103778469A (en) * | 2013-01-23 | 2014-05-07 | 辽宁工程技术大学 | Blasting scheme selection method based on neural network optimization genetic algorithm |
-
2015
- 2015-05-19 CN CN201510259910.6A patent/CN106290263B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101915753A (en) * | 2010-07-30 | 2010-12-15 | 浙江师范大学 | Genetic Neural NetworkQuantitative analysis method for laser induced breakdown spectroscopy based on gGenetic Neural Networkgenetic neural network |
CN103778469A (en) * | 2013-01-23 | 2014-05-07 | 辽宁工程技术大学 | Blasting scheme selection method based on neural network optimization genetic algorithm |
Non-Patent Citations (3)
Title |
---|
刘金桐: ""多光谱峰值分离技术在LIBS中的应用研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
沈沁梅: ""基于遗传神经网络的激光诱导击穿光谱元素定量分析技术"", 《中国优秀硕士学位论文全文数据库 基础科学辑》 * |
邹孝恒 等: ""基于遗传算法和偏最小二乘法的土壤激光诱导击穿光谱定量分析研究"", 《分析化学》 * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109668862A (en) * | 2017-10-17 | 2019-04-23 | 中国科学院沈阳自动化研究所 | A kind of aluminium electrolyte molecular proportion detection method based on laser induced breakdown spectroscopy |
CN109668862B (en) * | 2017-10-17 | 2021-02-05 | 中国科学院沈阳自动化研究所 | Aluminum electrolyte molecular ratio detection method based on laser-induced breakdown spectroscopy |
CN107887289A (en) * | 2017-11-13 | 2018-04-06 | 北京半导体专用设备研究所(中国电子科技集团公司第四十五研究所) | A kind of method and device for obtaining thin film parameter value to be measured |
CN107887289B (en) * | 2017-11-13 | 2021-03-09 | 北京半导体专用设备研究所(中国电子科技集团公司第四十五研究所) | Method and device for obtaining parameter value of film to be measured |
CN108414475A (en) * | 2018-01-30 | 2018-08-17 | 中国科学院上海技术物理研究所 | The LIBS analysis methods rebuild based on optical chromatography Simultaneous Iteration |
CN108414475B (en) * | 2018-01-30 | 2020-06-26 | 中国科学院上海技术物理研究所 | LIBS analysis method based on optical chromatography simultaneous iterative reconstruction |
CN110780021A (en) * | 2019-10-30 | 2020-02-11 | 广船国际有限公司 | Method and device for determining standard substance, terminal and storage medium |
CN112414996A (en) * | 2020-07-24 | 2021-02-26 | 北京工商大学 | Finite difference and difference evolution algorithm-based ICP-AES spectral line overlapping interference correction method |
CN112414996B (en) * | 2020-07-24 | 2022-06-17 | 北京工商大学 | Finite difference and difference evolution algorithm-based ICP-AES spectral line overlapping interference correction method |
Also Published As
Publication number | Publication date |
---|---|
CN106290263B (en) | 2018-08-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106290263A (en) | A kind of LIBS calibration and quantitative analysis method based on genetic algorithm | |
CN103534578B (en) | Determine spectral device and the method for composition present in sample | |
Shi et al. | Development of a national VNIR soil-spectral library for soil classification and prediction of organic matter concentrations | |
CN105300923A (en) | Correction method of measuring point free temperature compensation model during process of online application of near infrared spectrum analyzer | |
CN105181678B (en) | Rice varieties discrimination method based on LIBS | |
CN104778349B (en) | One kind is used for rice table soil nitrogen application Classified Protection | |
CN107356569A (en) | The construction method of method and its model based on chlorophyll fluorescence prediction wheat grain yield | |
CN104483292B (en) | A kind of method that use multiline ratio method improves laser microprobe analysis accuracy | |
CN102830096A (en) | Method for measuring element concentration and correcting error based on artificial neural network | |
CN113435707B (en) | Soil testing formula fertilization method based on deep learning and weighting multi-factor evaluation | |
CN106980750B (en) | A kind of Soil Nitrogen estimation method of reserve based on typical correspondence analysis | |
CN104198512A (en) | Support vector machine-based X-ray fluorescence spectrum analysis method and support vector machine-based X-ray fluorescence spectrum analysis device | |
CN112231621B (en) | Method for reducing element detection limit based on BP-adaboost | |
Yang et al. | Application of fluorescence spectrum to precisely inverse paddy rice nitrogen content. | |
CN106990056A (en) | A kind of total soil nitrogen spectrum appraising model calibration samples collection construction method | |
CN104730042A (en) | Method for improving free calibration analysis precision by combining genetic algorithm with laser induced breakdown spectroscopy | |
CN110503156A (en) | A kind of multivariate calibration characteristic wavelength selection method based on least correlativing coefficient | |
CN107132190A (en) | A kind of soil organism spectra inversion model calibration samples collection construction method | |
Brennan et al. | Returns to investment in new breeding technologies | |
CN109142251A (en) | The LIBS quantitative analysis method of random forest indirect labor's neural network | |
Mazzei et al. | HRMAS-NMR metabolomics of Aglianicone grapes pulp to evaluate terroir and vintage effects, and, as assessed by the electromagnetic induction (EMI) technique, spatial variability of vineyard soils | |
Röll et al. | Implementation of an automatic time‐series calibration method for the DSSAT wheat models to enhance multi‐model approaches | |
CN110567941B (en) | Rice seed moisture content grading detection method based on main element spectral intensity | |
CN115598162A (en) | Soil heavy metal content detection method based on stacked model | |
CN109211814B (en) | It is a kind of to be set a song to music the soil profile kind identification methods of face partition characteristics based on three-dimensional light |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |