CN104568811B - The new quick determination method of caloric value in coal sample - Google Patents
The new quick determination method of caloric value in coal sample Download PDFInfo
- Publication number
- CN104568811B CN104568811B CN201410579110.8A CN201410579110A CN104568811B CN 104568811 B CN104568811 B CN 104568811B CN 201410579110 A CN201410579110 A CN 201410579110A CN 104568811 B CN104568811 B CN 104568811B
- Authority
- CN
- China
- Prior art keywords
- sample
- coal
- detection model
- caloric value
- spectrum
- 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.)
- Expired - Fee Related
Links
Landscapes
- Investigating Or Analysing Materials By Optical Means (AREA)
- Investigating Or Analyzing Materials Using Thermal Means (AREA)
Abstract
The invention discloses a kind of new quick determination method of caloric value in coal sample.It the described method comprises the following steps：S1. several coal samples are collected and prepare, conventional method determines the caloric value content of each sample respectively；S2. the spectroscopic data and curve of the coal sample are collected with the scanning of NIRS analyzers；S3. the spectroscopic data of sample obtained by S2 is handled, the calibration equation for obtaining caloric value is calculated through returning, amendment and checking caloric value calibration equation set up detection model；S4., coal sample to be measured is directly filled to the injector of NIR instrument successively, starts scanning key, the automatic record storage sample spectra of NIR instrument determines that sample belongs to spectrogram type, selects corresponding detection model, obtains testing result.The degree of accuracy of quantity measuring method of being generated heat in the coal that the present invention is set up conforms to current standards the requirement of method repeatability limit difference, and operation is very easy to be quick, can just complete within 50 seconds or so the Scanning Detction of a sample, include the output of data.During whole detection operation, without claiming sample, without adding any chemical reagent, with quick, convenient, free of contamination feature.
Description
Technical field
The present invention relates to coal analysis detection technique field, more particularly, to a kind of the new of coal sample caloric value
Quick determination method.
Background technology
Coal resources are the first big energy of China, are the basic energy resource materials needed for industrial development.Coal resources in China
Enrich very much, account for 70% or so of China's total energy.
Coal resources are the first big energy of China, are the basic energy resource materials needed for industrial development.Coal resources in China
Enrich very much, account for 70% or so of China's total energy.
According to international coal examination criteria, coal usually requires the quality projects such as detection fixed carbon.By state of China domestic discipline and family rules
Fixed and trade needs, and coal need to generally detect the indexs such as Nei Shui, full sulphur, volatile matter, ash content, caloric value, fixed carbon, flammable body,
Current standard methods are mainly：Interior water has GB/T 212, ISO 11722, ASTMD3173, and ash content has GB/T 212, ISO
1171st, ASTMD 3174, volatile matter have GB/T 212,
ISO 562, ASTMD 3175, full sulphur have GB/T 214, ISO 351, ASTMD 4239, and caloric value has GB/T
213rd, the requirement in the Industry Analysis Method of ISO 1928, ASTMD5865, fixed carbon and flammable body reference coal is carried out.
The classical or modern chemical method and physical-chemical process that above-mentioned standard method is generally used：Caloric value, volatile matter, ash
Divide, fixed carbon, flammable body are all classical gravimetric methods, be related to the equipment such as balance, baking oven, high temperature furnace, carbonization, ashing, perseverance
Weight, weigh, calculate be frequently necessary to carry out the step of, it is quite cumbersome time-consuming；Even if using more advanced instrument, one at present
Technical staff completes these projects and also obtained 5~6 days, while the operation of the physical chemistry instrument of modernization, maintenance, demarcation, verification
It is also very heavy Deng work.The traditional detection method of coal is numerous and diverse due to working, and spends cost higher, spends the time long.
Quick demand of the classification sale of coal, the export trade and power plant etc., traditional coal round of visits is oversize, is unfavorable for coal
Sale, the Check-Out Time for shortening coal is needed for this, it is necessary to find the new method of inspection.
NIRS is the english abbreviation of near infrared spectrum.NIRS technologies are high-new analysis skills with the fastest developing speed nearly ten years
One of art.NIRS analytical technology application spectrum section wave-length coverages are about 3~0.70mm, the dry infrared range of spectrum of category, and can
See that light is the same, be all a part of electromagnetic wave, the general characteristic shown when being acted on electromagnetic wave and object, such as thoroughly
Penetrate, diffusing reflection, absorption etc..In addition, the characteristics of it is most prominent be this SPECTRAL REGION for hydric group (OH, SH, CH, NH) times
Frequency and sum of fundamental frequencies uptake zone.The near infrared spectrum of material is the frequency multiplication of wherein each group vibration and the synthesis absorption table of combination frequency
It is existing.
NIRS is quickly grown nearly ten years, and China is mainly used in the quality point of agricultural product since the eighties in last century
Analysis, through being applied to every field, expands to petrochemical industry and basic organic chemical industry, height from traditional agricultural byproducts analysis
The field such as molecule chemical industry, pharmacy and clinical medicine, biochemical industry, environmental science, textile industry and food industry.But, at present
Research report of the NIRS technologies in terms of coal is seldom.
Data shows that the presence that foreign countries' report is detected using NIRS technologies to coal is much difficult, and there is use the country
Fourier transform near infrared spectrum method sets up coal volatile matter, determination of moisture model, but does not propose to determine using NIRS technologies
Amount analysis Coal ' moisture, the concrete technical scheme of volatile matter.The applicant is summed up using near red by long-term substantial amounts of research
The correlation merit project quick determination method that outer analysis technology sets up coal be it is feasible, still, the following emphasis skill to be solved
Art problem is not solved effectively always：(1) the suitable mathematics transition form of coal NIRS spectroscopic datas is found；(2) find and close
Reliable functional relation between the NIRS detection data of suitable mathematics conversion and the detection data of coal quality composition conventional method；(3)
To further amendment and the checking of gained functional relation.Have no the quick inspection for the caloric value that coal is set up using NIR technology
The technology report of survey method.
The content of the invention
The technical problem to be solved in the present invention is that the detection for being directed to heating value of coal is closed there is provided specific NIRS spectroscopic datas
Suitable mathematics transition form, sum up the NIRS detection data of suitable mathematical conversion and the detection number of coal quality composition conventional method
The reliable functional relation between, and gained functional relation is further corrected and verified there is provided a kind of coal sample caloric value
New quick determination method.
The purpose of the present invention is achieved by the following technical programs：
A kind of new quick determination method of coal sample caloric value is provided, comprised the following steps：
S1. several coal samples are collected and prepare, conventional method determines the caloric value content of each sample respectively；
S2. the spectroscopic data and curve of the coal sample are collected with the scanning of NIRS analyzers；
S3. the spectroscopic data of sample obtained by S2 is handled, the calibration equation for obtaining caloric value, amendment is calculated through returning
With checking caloric value calibration equation, detection model is set up；
S4., coal sample to be measured is directly filled to the injector (being not required to claim sample) of NIR instrument successively, starts scanning key, NIR instrument
Automatic record storage sample spectra.Determine that sample belongs to spectrogram type, select corresponding detection model, obtain testing result.
Scan with NIRS analyzers that to collect the spectroscopic data of the sample be by sample (being not required to claim sample), successively directly described in S2
The injector of the full NIR instrument of tipping, is scanned using digital raster system, is recorded automatically by NIRS instrument and storage sample light
Spectrum.The sample original spectrum being collected into is classified by its different variation tendency, it is identical or close with variation tendency according to peak type
Collection of illustrative plates is combined, and is sorted out, and respectively obtains Y types original spectrum, W types original spectrum, p-type original spectrum and X-type primary light
Spectrum.
It is that spectrum analysis is carried out using WinISI softwares to the spectroscopic data progress Mathematical treatment of sample obtained by S2 described in S3
With set up detection model, the spectroscopic data of sample obtained by S2 is imported into NIR instrument, detection model is determined, testing result is printed.Spectrum
The sides such as trend converter technique, standard normal variable transformation approach, Multivariate Discrete correction, the correction of anti-phase Multivariate Discrete are respectively adopted in pretreatment
The one or more of method, it is final to determine optimization process method with reference to coal sample specificity analysis；Regression correction method is using progressively
Regression analysis (SMLR), PCA (PCA) and minimal error analytical method (PLS) are numerous to eliminate by Data Dimensionality Reduction
Overlapped message part and quantization to spectrum is finally accomplished in information co-exist.
Caloric value content in the sample of one group of unknown component content to be measured is detected using detection model, then by obtained by NIR methods
Detected value be compared and evaluate with Typical physical chemical method detected value.The comparative result of two methods prediction standard deviation
(SEP, Standard Error of Prediction) and the corresponding coefficient of determination (RSQp) or coefficient R p are weighed.
The method for building up of described detection model comprises the following steps：
S31.GH values are analyzed, and the sample that the GH values are more than 3.0 are rejected, the sample sets for being less than 3.0 with GH values are built respectively
Calibration (detection) model of vertical respective type spectrogram；
S32. the SEC values and RSQ values of (detection) model are calibrated described in S31 by calculating；
S33. evaluation experimental is carried out to detection model and determines optimum detection model.The present invention uses validation-cross error
(Standard Error of cross validation, SECV) and the validation-cross coefficient of determination (1minus the
Variance ratio, 1-VR) or coefficient R v weigh detection model.Detection can be effectively assessed by the two indexs
The prediction accuracy of model.Good detection model is evaluated with owning using with low SECV values and high (1-VR) or Rv values
Detection model interacts checking test, selects the detection model of minimum SECV values and highest (1-VR) or Rv values, is defined as most
Good detection model.
Beneficial effects of the present invention are as follows：
The principle of the existing examination criteria of coal is all classical or modern chemical method and physical-chemical process：Interior water, volatile matter,
Ash content, fixed carbon, flammable body are all classical gravimetric methods, are related to the equipment such as balance, baking oven, high temperature furnace, be carbonized, be ashed,
Constant weight, weighing, calculate be frequently necessary to carry out the step of, it is quite cumbersome time-consuming；Caloric value, full sulphur, state-of-the-art at present is to adopt
Use HTHP combustion method, but the operation of the physical chemistry instrument of modernization, maintenances, demarcation are also very heavy.
For a long time, near-infrared spectrum technique is only used for analyzing pure organic matter.Because the wave number of near infrared spectrum exists
4000cm-1Above (i.e. below 2500nm), therefore, only vibration frequency is in 2000cm-1Vibration above, is only possible near red
One-level frequency multiplication is produced in outskirt, and can be in 2000cm-1Produce fundamental vibration above mainly contains hydrogen functional group, such as C-H, N-
H, S-H and O-H stretching vibration.Almost in organic matter all hydric groups information, can be able near infrared spectrum anti-
Reflect.
Coal is a kind of flammable rock.It is this by major part that near-infrared spectrum technique is applied to coal by the present invention first
The compounding substances system of organic substance and part mineral matter and moisture composition, and it is successfully established the assay method of caloric value content,
So as to prove that near-infrared spectrum technique can apply to analyze inorganic substances to a certain extent, prior art prejudice is overcome.
The present invention establishes the near infrared detection method of heating value of coal, has filled up prior art blank, and solve with
Lower focus technology problem：Sum up what the suitable mathematics transition form of coal NIRS spectroscopic datas was changed there is provided suitable mathematical
NIRS detects reliable functional relation between data and the detection data of coal quality composition conventional method, and gained functional relation is carried
Practical, reliable and stable amendment and verification method are supplied.
Based on the inventive method, 4 detection models of caloric value content in coal are obtained, and sets up and can be applied to
New detection method in daily practice examining work.Caloric value content in present invention detection coal sample needs a few minutes, Er Qiewu
Sample need to be claimed, without using the test condition such as chemical reagent or high temperature, high pressure, high current, chemistry, biological or electromagnetism will not be produced dirty
Operating personnel and environment will not be had undesirable effect by dye, continuous detection can be achieved, to mass detection task, its superiority
More protrude.
It is of the invention compared with existing oxygen bomb combustion：The original spectrogram calibration standard deviation of Y types is 0.12MJ/Kg, calibrates phase
Relation number is 0.9960；Validation-cross standard deviation 0.14MJ/Kg, validation-cross coefficient correlation 0.9951；The standard of Preliminary Applications
Deviation is 0.26MJ/Kg, and coefficient correlation is 0.983.The original spectrogram calibration standard deviation of W types is 0.22MJ/Kg, calibrates phase relation
Number is 0.9815；Validation-cross standard deviation 0.31MJ/Kg, validation-cross coefficient correlation 0.9610；The standard deviation of Preliminary Applications
For 0.27MJ/Kg, coefficient correlation is 0.972.The original spectrogram calibration standard deviation of p-type is 0.48MJ/Kg, and calibration coefficient correlation is
0.9734；Validation-cross standard deviation 0.60MJ/Kg, validation-cross coefficient correlation 0.9594；The standard deviation of Preliminary Applications is
0.88MJ/Kg, coefficient correlation is 0.971.The original spectrogram calibration standard deviation of X-type is 0.45%, and calibration coefficient correlation is
0.9830；Validation-cross standard deviation 0.72MJ/Kg, validation-cross coefficient correlation 0.9582；The standard deviation of Preliminary Applications is
1.20MJ/Kg, coefficient correlation is 0.901.
Present invention only requires using near-infrared spectrometers, it is possible to instead of prior art is a variety of, many analyses
Instrument, it is only necessary to grind away equipment, it is not necessary to assay balance, generally only needs people's operation, and within a few minutes, pass through collection one
The spectrum of secondary sample, it is possible to while completing the measure of caloric value content.During spectra collection in addition to electric energy is consumed,
Consumption any reagent and standard substance is not required to, can so save a large amount of instrument and equipments the expense such as purchases, operates, repair, saves
Substantial amounts of time and manpower are saved, analysis cost is greatly reduced, the efficiency of detection work is significantly improved.
Brief description of the drawings
The near-infrared diffusing reflection DDS original spectrum total figures of 509 anthracite samples of Fig. 1.
The near-infrared diffusing reflection DDS original spectrum total figures of 134 bituminous coal samples of Fig. 2.
The near-infrared diffusing reflection DDS original spectrum total figures of 155 steam coal samples of Fig. 3.
The near-infrared diffusing reflection DDS original spectrum total figures of 60 meager lean coal samples of Fig. 4.
The near-infrared diffusing reflection DDS original spectrum total figures of 134 power coal samples of Fig. 5.
The near-infrared diffusing reflection DDS original spectrum total figures for the sample that packet is not known in Fig. 6 179.
The near-infrared diffusing reflection DDS original spectrums total figure (sample in 2008) of 202 anthracite samples of Fig. 7.
The near-infrared diffusing reflection DDS original spectrums total figure (sample in 2008) for the sample that packet is not known in Fig. 8 235.
The original spectrogram of Fig. 9 standard specimens (being above 103f anthracites, lower is 101L bituminous coal).
The original spectrogram of Figure 10 certified reference coals.
Figure 11 anthracites in 2008 and the anthracitic original spectrogram of standard specimen.
The original spectrogram of Figure 12 2005-2008 bituminous coal and standard specimen bituminous coal.
Figure 13 anthracites in 2008 and standard specimen anthracite original spectrum handle figure through (NONE+0011).
Figure 14 anthracites in 2008 and standard specimen anthracite original spectrum handle figure through (NONE+1441).
Figure 15 anthracites in 2008 and standard specimen anthracite original spectrum handle figure through (D+1441).
Figure 16 anthracites in 2008 and standard specimen anthracite original spectrum handle figure through (D+1441).
The original spectrum of Figure 17 2005-2008 bituminous coal and standard specimen bituminous coal handles figure through (NONE+0011).
The original spectrum of Figure 18 2005-2008 bituminous coal and standard specimen bituminous coal handles figure through (D+1441).
The original spectrum of Figure 19 2005-2008 bituminous coal and standard specimen bituminous coal handles figure through (D+0011).
509 anthracite sample original spectrums of Figure 20 handle figure through (NONE+1441).
509 anthracite sample original spectrums of Figure 21 handle figure through (NONE+0011).
509 anthracite sample original spectrums of Figure 22 handle figure through (D+1441).
Figure 23 Y types (374) primary light spectrogram.
Figure 24 W types (1367) primary light spectrogram.
Figure 25 p-types (436) original spectrogram.
Figure 26 X-types (52) original spectrum.
The GH Distribution value figures of the original spectrogram of Figure 27 Y types.
The GH Distribution value figures of the original spectrogram of Figure 28 W types.
The GH Distribution value figures of the original spectrogram of Figure 29 p-types.
The GH Distribution value figures of the original spectrogram of Figure 30 X-types.
Y type original spectrum caloric value detection models (PLS+NONE+2441) of the Figure 31 through validation-cross.
W type original spectrum caloric value detection models (PLS+NONE+1441) of the Figure 32 through validation-cross.
P-type original spectrum caloric value detection models (PLS+D+1441) of the Figure 33 through validation-cross.
X-type original spectrum caloric value detection models (MPLS+SNV+1441) of the Figure 34 through validation-cross.
Embodiment
Below in conjunction with the accompanying drawings the present invention is further illustrated with specific embodiment.The equipment that the embodiment of the present invention is used can refer to
《High temperature process furnances combustion method analyzes the standard specimen method of sulfur content in coal and coke analysis sample》(ASTMD-4239-
2010e1)、《Proximate analysis of coal》(GB/T212-2008)、《The heat output determining method of coal》(GB/T213-2008) in
Listed equipment and reagent.Other unless stated otherwise, the reagent and equipment that the embodiment of the present invention is used is commonly used in the art
Reagent and equipment.
Embodiment 1
The present embodiment provides a kind of new quick determination method of caloric value in coal sample, comprises the following steps：
S1. several coal samples are collected and prepare, conventional method determines the caloric value content of each sample respectively；
S2. the spectroscopic data and curve of the coal sample are collected with the scanning of NIRS analyzers；
S3. the spectroscopic data of sample obtained by S2 is handled, the calibration equation for producing caloric value, amendment is calculated through returning
With checking caloric value calibration equation, detection model is set up；
S4., coal sample to be measured is directly filled to the injector (being not required to claim sample) of NIR instrument successively, starts scanning key, NIR instrument
Automatic record storage sample spectra.After judgement sample ownership spectrogram type, corresponding detection model is clicked on, detection knot is obtained
Really.
Wherein, the method for coal sample being collected and preparing described in S1 is reference《Sample for commercial coal takes method》(GB475-
2008) and《The preparation method of coal sample》(GB474-2008) requirement in is carried out.The conventional method determination sample caloric value
Method be by《The heat output determining method of coal》(GB/T213-2008) requirement in is carried out.
Scan with NIRS analyzers that to collect the spectroscopic data of the sample be by sample (being not required to claim sample), successively directly described in S2
The injector of the full NIR instrument of tipping, is scanned using digital raster system, is recorded automatically by NIRS instrument and storage sample light
Spectrum.
The present embodiment uses the SY-3650-2 type near-infrared analysis that FOSS ANALYTICAL A/S companies of Denmark produce
Instrument, monochromator：Single beam holography digital raster, 1100~2500nm of wave-length coverage；Light source：Tengsten lamp；Detector：PBS, automatically
Mobile conveying multipoint positioning detection, test point is more than 32；Operation temperature：15~32 DEG C；Analysis time：It is per minute to use continuous light
Spectrum scans more than 32 subsamples and completes its spectrum analysis；System noise：Noise signal value is less than 2 × 10-5AU；As a result report
Accuse：The sample of result and display " having surmounted correcting range " is shown on a terminal screen, while reporting result by printer, also may be used
It is connected by RS-232C interfaces with external electrical computer.
Operating method：Carefully stirring coal sample, fills and takes appropriate sample in the sample cell of quartz system, must during filling
Make thickness of sample uniform, cover small cardboard, press lightly on, make sample tight distribution.After filling is finished, observation sample rooved face
Sample situation, if it find that having gap or sample to have the phenomenon of loosening, should recharge sample.The sample cell filled is placed in
On sample introduction track, click on " scanning ", by the input mode pumped, completion Scanning Detction and according to output.Entirely
During detection operation, without claiming sample, without adding any chemical reagent.
First, collection sample is scanned using digital raster system (Digital Dispersive System, DDS)
Spectrum.The present invention is scientifically classified, by coal sample respectively according to anthracite, bituminous coal, steam coal, meager lean coal, thermal coal
It is grouped, one group is then individually classified as to the sample do not classified specifically.Coal sample need not claim sample, and NIR instrument is directly filled successively
Injector, then the sample for the injector that overfills is scalped, the DDS spectrum of the automatic record storage sample of near-infrared analyzer are obtained
Original spectrogram, as shown in accompanying drawing 1 to accompanying drawing 8.The display of the primary light spectrogram shown in accompanying drawing 1 to accompanying drawing 8, only accompanying drawing 4
The variation tendency of the near-infrared diffusing reflection DDS original spectrums of meager lean coal sample relatively or unanimously, other coals such as anthracite,
All there is the variation tendency of obvious difference in bituminous coal, steam coal, the original spectrogram of thermal coal.Because accompanying drawing 1 to Fig. 6 is nearly three
The spectrogram of individual month concentrated collection, considers the influence that analysis Sample Storage time and environment exist to its original spectrum, this hair
Bright collection coal sample in 2008 gathers spectrogram in time, obtains Fig. 7 to Fig. 8, it can be seen that Fig. 7 is basically identical with Fig. 1, the various coals of Fig. 8
The near-infrared diffusing reflection DDS original spectrums of sample cover foregoing anthracite, bituminous coal, steam coal, the original spectrum of thermal coal substantially
The variation tendency of figure.This illustrates that our storage of samples condition or Spectral acquisition times influences unobvious to original spectrum, can be with
Meet present invention research needs.More preferably to investigate coal original spectrum, we are gathered to coal standard sample using the same manner
Original spectrum, obtains Fig. 9 to Figure 10.And the original spectrogram of gained standard sample original spectrum and corresponding coal is combined progress
Observation, obtains Figure 11 to Figure 12.Spectrum change trend can see by Figure 11 and Figure 12 and foregoing corresponding coal is basically identical.
Then, the present invention carries out Pretreated spectra respectively to Figure 11 to Figure 12, obtains Figure 13 to Figure 19.Pretreated spectra mode
Respectively：NONE+0011, NONE+1441, D+1441, D+1441, NONE+0011, D+1441, D+0011 (wherein, NONE tables
Show and do not carry out Pretreated spectra, as original spectrogram, D represents trend converter technique, 1441 represent to make by interval of every 4 spectrum point
The Mathematical treatment of first derivative, 0011 represents not making any derivative processing.For example, D+1441 is to carry out trend to primary light spectrogram
After converter technique pretreatment, then make by interval of every 4 spectrum point first derivative progress Mathematical treatment).Can be with by Figure 13 to Figure 19
See that spectrogram variation tendency after pretreatment is more obvious, clear, but imitated obtained by the various Different treatments of same coal
The fruit visibly different peak type of generally existing and change.Totally apparently, Pretreated spectra fails to obtain than more consistent spectrogram.
The foregoing various coal original spectrums collected are pre-processed, gained situation is essentially identical, original with 509 anthracite samples
Exemplified by spectrum is through (NONE+1441) processing figure, result is see shown in accompanying drawing 20 to accompanying drawing 22, and other sample pretreatment figures exist
This is not repeated (figure is omited) one by one.
S2 is that the sample original spectrum that will be collected into is classified by its different variation tendency, identical with variation tendency according to peak type
Or close collection of illustrative plates is combined, and is sorted out, Y types original spectrum, W types original spectrum, p-type original spectrum and X are respectively obtained
Type original spectrum.The foregoing original spectrum collected is grouped by the present invention by its different variation tendency, then peak type and change are become
Gesture is identical or collection of illustrative plates of relatively is combined, and is sorted out, obtains spectrogram 23 to Figure 26.Figure 23 spectrogram is former labeled as Y types
Beginning spectrum, represents the coal such as most steam coal and part bituminous coal, partial power coal, has 374 samples.Figure 24 is labeled
For W type original spectrums, the coals such as most anthracite, part bituminous coal, partial power coal are represented, 1367 samples are had.Figure
25 are marked as p-type original spectrum, represent the coals such as most meager lean coal, part bituminous coal, partial power coal, have 436
Sample.Figure 26 is marked as X-type original spectrum, represents non-common coal, has 195 samples.
It is that spectrum analysis is carried out using WinISI softwares to the spectroscopic data progress Mathematical treatment of sample obtained by S2 described in S3
With set up detection model, the spectroscopic data of sample obtained by S2 is imported into NIR instrument, detection model is determined, testing result is printed.Spectrum
The sides such as trend converter technique, standard normal variable transformation approach, Multivariate Discrete correction, the correction of anti-phase Multivariate Discrete are respectively adopted in pretreatment
The one or more of method, it is final to determine optimization process method with reference to coal sample specificity analysis；Regression correction method is using progressively
Regression analysis (SMLR), PCA (PCA) and minimal error analytical method (PLS) are numerous to eliminate by Data Dimensionality Reduction
Overlapped message part and quantization to spectrum is finally accomplished in information co-exist.
Caloric value content in the sample of one group of unknown component content to be measured is detected using detection model, then by obtained by NIR methods
Detected value be compared and evaluate with Typical physical chemical method detected value.The comparative result of two methods prediction standard deviation
(SEP, Standard Error of Prediction) and the corresponding coefficient of determination (RSQp) or coefficient R p are weighed.
The method for building up of described detection model comprises the following steps：
S31.GH values are analyzed：GH values are mahalanobis distance, be spectroscopic data regression correction score graphics in, each sample
The distance of distance center sample spot.GH values are generally set to 3.0 near infrared spectrum data analysis, are meant that standard makes a variation
3 times of unit, that is, be exactly be similar to standard deviation (SD) 2.84 times, it means that can be big by the GH values for the sample for having about 10%
In 3.0.If the GH values of which sample are more than 3.0, the sample needs to reject, and separately performs an analysis.The present invention uses PCA
(PCR) clustering is carried out respectively to Y types original spectrum, W types original spectrum, p-type original spectrum and X-type original spectrum；As a result
As shown in accompanying drawing 27 to accompanying drawing 30, analyzed from accompanying drawing 27 to 30, the GH value overwhelming majority of the original spectrogram of four types is less than
The sample that 3.0, Y types are more than 3.0 has 8, and the sample that W types are more than 3.0 has 10, and the sample that p-type is more than 3.0 has 13, and X is more than
3.0 sample has 9, and the GH values are more than after the rejecting of 3.0 sample by the present invention, and 3.0 sample sets point are less than with GH values
Calibration (detection) model of respective type spectrogram, wherein 366, Y types, 1357, W types, p-type 423, X-type 186 are not set up.
S32. the SEC values and RSQ values of (detection) model are calibrated described in S31 by calculating；
Spectral manipulation is carried out to above-mentioned four classes original spectrum using spectral manipulation in WinISI softwares and regression correction method
And data analysis.Pretreated spectra be respectively adopted trend converter technique, standard normal variable transformation approach, Multivariate Discrete correction, it is anti-phase
The one or more of the methods such as Multivariate Discrete correction, it is final to determine optimization process method with reference to coal sample specificity analysis；Mathematics
Processing is used makees first derivative (1441) or second dervative (2441) method, (0011) table with the interval (Gap) of every 4 spectrum point
Show and do not do any derivative processing.Regression correction method uses Stepwise Regression Method (Stepwise Mutiple Linear
Regression, SMLR), PCA (Principal Component Analysis, PCR), minimum deflection analysis
Method (Partial Least Squares Regression, PLS) and inclined minimal error analytical method (Modified Partial
Least Squares Regression, MPLS) calculate the SEC values and RSQ values for calibrating (detection) model.
SEC is calibration standard deviation (Standard Error of Calibration, SEC), refers to the calibration by foundation
Model calibration sample collection is predicted obtained by near-infrared analysis value and conventional chemical methods assay value standard deviation, be back
Return the mark of reading and actual read number degree of agreement.SEC is lower, illustrates that near-infrared analysis result is more kissed with traditional analysis result
Close, confidence level is higher.RSQ (R squared) is the calibration coefficient of determination, is coefficient correlation (Rc) square, refers to calibration model pair
The percentage that the variation of calibration sample collection can be depicted, represents that near-infrared analysis value and conventional method assay value linear relationship are close
Degree.One good detection model requirement has low SEC and high RSQ (Rc).It is theoretical according to the existing common knowledge in this area
On, the model for possessing minimum SEC values and highest RSQ values should be just optimal detection model.But the present invention is by a large amount of long
The experimental summary of phase is found, and the model of not least SEC values and highest RSQ values is exactly optimal detection model, but finally can quilt
Be set to optimum detection model necessarily has relatively low SEC values and higher RSQ values.The whole of foundation is detected for this present invention
Model carries out evaluation test one by one.By taking the original spectrogram of Y types as an example, other can refer to Y types, not repeat one by one.The hair of coal sample
Heat spectral manipulation and regression correction result are shown in Table 1 that (by taking sample segment as an example, the experimental data of other samples does not exist respectively
This is repeated one by one)：
The original spectrogram spectral manipulation of the Y types of table 1 and regression correction effect (caloric value)
The validation-cross result (caloric value) of the Y type detection models of table 2
S33. evaluation experimental is carried out to detection model and determines optimum detection model.
No matter the which type of coal spectrum it can be seen from Tables 1 and 2, the detection model of each coal index, its
SEC values, the RSQ values of the characteristic value of validation-cross result -- SECV values, (1-VR) value and corresponding model are all different, SECV values
All it is slightly larger than SEC values, (1-VR) value is all slightly less than RSQ values.SECV values, (1-VR) value can more reflect detection than SEC values, RSQ values
Model is used to predict the degree of accuracy of unknown sample in the future, because what SEC values, RSQ values reflected is detection model to calibrating sample
Product collection be predicted obtained by near-infrared analysis value and the standard deviation of conventional chemical methods assay value, near-infrared analysis value with often
The close degree of rule method assay value linear relationship；And the reflection of SECV values, (1-VR) value is the detection when validation-cross is calculated
Model the sample for not participating in calibration modeling is predicted obtained by near-infrared analysis value and conventional chemical methods assay values
Standard deviation, the sample near-infrared analysis value and the close journey of conventional method assay value linear relationship for not participating in calibration modeling
Degree.As can be seen here, a good detection model not only requires low SEC and high RSQ (Rc), more requires there are low SECV values
With high (1-VR) or Rv values.
No matter which type of coal spectrum, the detection model of each coal index, one possess minimum SECV values and
The detection model of highest (1-VR) or Rv values, necessarily with relatively low SEC values and higher RSQ values.The present invention is based on more kinds of
The knot summarized in sample detection research in addition to coal sample is consistent with result of study of the present invention, four kinds of spectrum types of the invention
Each coal index there is the detection model of minimum SECV values and highest (1-VR) value to be determined as optimum detection model, and will be main
Information summary is wanted to analyze in table 3.
The main characteristic parameters and its validation-cross result of the optimum detection model of table 3
From table 3, multiple spectrum processing is carried out to the near-infrared diffusing reflection original spectrum of coal using WinISI softwares
And regression correction, as a result with using standard normal variable conversion (SNV), trend conversion (D), without scattering (NONE) processing and once
Smoothing (0011), made with the interval (Gap) of 4 spectrum points first derivative (1441), second dervative (2441) processing and
The combined effect of MPLS, PLS regression correction is optimal.The principal character value and corresponding homing method of gained optimum detection model,
Spectral manipulation mode is referred to listed by table 3.The optimum N IR detection models of each coal index have 692 parameters, wherein including one
Constant term, refers to table 4 to table 7, equation curve (detection model) is as shown in Figure 31 to Figure 34, the parameter of other detection models and side
Journey curve map is omited.
The Y type original spectrum detection model parameters (caloric value MJ/Kg) of table 4
The W type original spectrum detection model parameters (caloric value MJ/Kg) of table 5
The p-type original spectrum detection model parameter (caloric value MJ/Kg) of table 6
The X-type original spectrum detection model parameter (caloric value MJ/Kg) of table 7
The applicating evaluating of embodiment 2 is tested
The detection model that the present embodiment is established using embodiment is detected in the coal sample of one group of unknown component content to be measured
Caloric value content, then the detected value obtained by NIR methods is compared and evaluated with Typical physical chemical method detected value.Two methods
Comparative result prediction standard deviation (SEP, Standard Error of Prediction) and the corresponding coefficient of determination
Or coefficient R p is weighed (RSQp).
Some coal samples are selected at random, is not required to claim sample, the injector of NIR instrument is directly filled successively, start scanning key,
The automatic record storage sample spectra of NIR instrument.After judgement sample ownership spectrogram type, corresponding detection model is clicked on, it is possible to
Testing result is printed, the detection work of the caloric value index of quality of coal sample is rapidly completed.This application tests tetra- kinds of Y, W, P, X
Scan values (the average that experiment value (average and standard deviation), the Near-Infrared Absorption Method that sample number, the conventional method of type use are obtained are obtained
And standard deviation), SEP and corresponding RSQp, Rp value are shown in Table 8 respectively.
The various coals of table 8 present invention analysis testing result
Prediction standard deviation, the coefficient of determination and the correlation of caloric value detection method of content in table 8, coal of the present invention
The indexs such as coefficient are close with corresponding calibration and the result.Specific conventional method Physical Chemistry Experiment value, NIR methods scan values and
Both difference, the repeatability of current standard methods limit differences are shown in Table 9 to table 12, only list the knot of 10 samples per project per type
Really, other abundant experimental results can not be repeated one by one herein.
The Y type detection model Preliminary Applications result of table 9-caloric value (MJ/Kg)
The W type detection model Preliminary Applications result of table 10-caloric value (MJ/Kg)
The p-type detection model Preliminary Applications result of table 11-caloric value (MJ/Kg)
The X-type detection model Preliminary Applications result of table 12-caloric value (MJ/Kg)
The degree of accuracy of heating quantity measuring method meets existing in the coal set up from above-mentioned table 9 to table 12, the present invention
The requirement of standard method repeatability limit difference.Operation is very easy to be quick, can just complete within 50 seconds or so the scanning inspection of a sample
Survey, include the output of data.During whole detection operation, without claiming sample, without adding any chemical reagent, with fast
Fast, convenient, free of contamination feature.
Claims (3)
1. a kind of new quick determination method of coal sample caloric value, it is characterised in that comprise the following steps：
S1. several coal samples are collected and prepare, conventional method determines the caloric value content of each sample respectively；
S2. the spectrum for collecting the coal sample is scanned using digital raster system with digital raster near-infrared diffusing reflection instrument
Data and curve；
S3. the spectroscopic data of sample obtained by step S2 is handled, the calibration equation for obtaining caloric value, amendment is calculated through returning
With checking caloric value calibration equation, detection model is set up；
S4., coal sample to be measured is directly filled to the injector of digital raster near-infrared diffusing reflection instrument successively, starts scanning key, number
The automatic record storage sample spectra of word grating near-infrared diffusing reflection instrument, determines that sample belongs to spectrogram type, the corresponding detection of selection
Model, obtains testing result；
Wherein, the spectroscopic data for scanning the collection coal sample described in step S2 with NIRS analyzers is to distinguish coal sample
It is grouped according to anthracite, bituminous coal, steam coal, meager lean coal, thermal coal, one is then individually classified as to the sample do not classified specifically
Group, original spectrum is gathered to coal standard sample using the same manner；By the sample original spectrum being collected into according to peak type and change
The identical or close collection of illustrative plates of change trend is combined, and is sorted out；
It is that spectrum analysis is carried out using WinISI softwares and built to the spectroscopic data progress processing of sample obtained by S2 described in step S3
Vertical detection model, imports NIR instrument by the spectroscopic data of sample obtained by S2, sets up detection model；
The method for building up of described detection model comprises the following steps：
S31.GH values are analyzed, and the sample that the GH values are more than 3.0 are rejected, the sample sets for being less than 3.0 with GH values set up phase respectively
Answer the detection model of type spectrogram；
S32. the SEC values and RSQ values of detection model described in calculation procedure S31 are passed through；
S33. evaluation experimental is carried out to detection model and determines optimum detection model；
It is to use validation-cross error and the validation-cross coefficient of determination or coefficient correlation that optimum detection model is determined described in step S33
To weigh detection model.
2. the new quick determination method of coal sample caloric value according to claim 1, it is characterised in that be using having
Low validation-cross error amount and the high validation-cross coefficient of determination or correlation coefficient value evaluate good detection model with owning
Detection model interacts checking test, selects minimum validation-cross error amount and the highest validation-cross coefficient of determination or related
The detection model of coefficient value, is defined as optimum detection model.
3. the new quick determination method of coal sample caloric value according to claim 1, it is characterised in that step S2 is also wrapped
Include the sample original spectrum that will be collected into be combined according to the peak type collection of illustrative plates identical or close with variation tendency, sorted out,
Respectively obtain Y types original spectrum, W types original spectrum, p-type original spectrum and X-type original spectrum.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410579110.8A CN104568811B (en) | 2014-10-24 | 2014-10-24 | The new quick determination method of caloric value in coal sample |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410579110.8A CN104568811B (en) | 2014-10-24 | 2014-10-24 | The new quick determination method of caloric value in coal sample |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104568811A CN104568811A (en) | 2015-04-29 |
| CN104568811B true CN104568811B (en) | 2017-08-25 |
Family
ID=53085388
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410579110.8A Expired - Fee Related CN104568811B (en) | 2014-10-24 | 2014-10-24 | The new quick determination method of caloric value in coal sample |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104568811B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107451408B (en) * | 2017-07-31 | 2020-11-06 | 华能平凉发电有限责任公司 | Coal-fired heating value data checking method |
| WO2021114109A1 (en) * | 2019-12-10 | 2021-06-17 | 深圳市能源环保有限公司 | Method for rapidly measuring and calculating calorific value of sludge on the basis of cielab color space |
| CN112710638B (en) * | 2020-12-19 | 2022-05-13 | 华中科技大学 | Method for rapidly detecting coal quality based on photoluminescence spectrum comprehensive analysis |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102042968A (en) * | 2010-11-24 | 2011-05-04 | 南京财经大学 | Grain quality near infrared rapid detection wireless system |
| CN104111234A (en) * | 2014-07-29 | 2014-10-22 | 中国农业大学 | Method and device for online detection of biomass basic characteristics based on near infrared spectroscopy |
-
2014
- 2014-10-24 CN CN201410579110.8A patent/CN104568811B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102042968A (en) * | 2010-11-24 | 2011-05-04 | 南京财经大学 | Grain quality near infrared rapid detection wireless system |
| CN104111234A (en) * | 2014-07-29 | 2014-10-22 | 中国农业大学 | Method and device for online detection of biomass basic characteristics based on near infrared spectroscopy |
Non-Patent Citations (5)
| Title |
|---|
| ASTM clustering for improving coal analysis by near-infrared spectroscopy;J.M.Andrés and M.T.Bona;《Talanta》;20060703;第70卷;第711-719页 * |
| Coal analysis by diffuse reflectance near-infrared spectroscopy: Hierarchical cluster and linear discriminant analysis;M.T.Bona and J.M.Andrés;《Talanta》;20070130;第72卷;第1423-1424页 * |
| 小麦蛋白质近红外模型的建立及特异种质资源筛选与分析;温亮;《中国优秀硕士学位论文全文数据库 农业科技辑》;20111215(第S1期);第15-17页 * |
| 应用NIRS分析技术快速检测煤炭质量;苏彩珠等;《检验检疫科学》;20071231;第17卷(第6期);第34-35页 * |
| 应用近红外光谱分析技术测量煤质发热量;李凤瑞等;《电站系统工程》;20040531;第20卷(第3期);第19-20页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104568811A (en) | 2015-04-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104390927B (en) | The quick determination method of ash content in coal sample | |
| Huang et al. | Improved generalization of spectral models associated with Vis-NIR spectroscopy for determining the moisture content of different tea leaves | |
| CN1982872B (en) | Near-infrared diffuse reflection spectral method for fastly inspecting drop water content | |
| CN106706553A (en) | Method for quick and non-destructive determination of content of amylase in corn single grains | |
| CN101413885A (en) | Near-infrared spectrum method for rapidly quantifying honey quality | |
| CN109540837B (en) | Method for rapidly detecting lignocellulose content of ramie leaves by near infrared | |
| CN105319179B (en) | A kind of method using middle infrared spectrum prediction hydrogen sulfide content in desulfurized amine | |
| CN104568811B (en) | The new quick determination method of caloric value in coal sample | |
| Feng et al. | Nondestructive and rapid determination of lignocellulose components of biofuel pellet using online hyperspectral imaging system | |
| CN204255854U (en) | A kind of system simultaneously detecting 7 indexs in coal sample | |
| Lu et al. | Fast and nondestructive determination of protein content in rapeseeds (Brassica napus L.) using Fourier transform infrared photoacoustic spectroscopy (FTIR‐PAS) | |
| Ge et al. | Visible and near‐infrared reflectance spectroscopy analysis of soils | |
| CN103115889A (en) | Method for predicating sulphur content of crude oil by infrared transmittance spectroscopy | |
| CN103134763B (en) | The method of oil density is predicted by infrared spectrum | |
| CN112179871A (en) | Method for nondestructive detection of caprolactam content in sauce food | |
| CN104390928B (en) | A kind of near infrared spectrum recognition methods for adulterating adulterated coal | |
| CN106018321A (en) | Establishment method for single corn seed protein detection model and application of establishment method | |
| CN104849234A (en) | Assay method for analyzing contents of principal components of imidacloprid based on near-infrared spectrum | |
| CN104535525B (en) | The quick determination method of fixed carbon in coal sample | |
| CN102954946A (en) | Method for determining sulfur content of crude oil through infrared spectrometry | |
| CN104316492A (en) | Method for near-infrared spectrum measurement of protein content in potato tuber | |
| CN110376154A (en) | Fruit online test method and system based on spectrum correction | |
| CN103134764A (en) | Method for predicting crude oil true boiling point distillation curve by transmission infrared spectroscopy | |
| CN104390936B (en) | The quick determination method of volatile matter in coal sample | |
| CN104515749B (en) | The quick determination method of water in coal sample |
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 | ||
| CB03 | Change of inventor or designer information |
Inventor after: Su Caizhu Inventor after: Yan Biheng Inventor after: He Xiaodong Inventor after: Li Peishen Inventor after: Zheng Jianguo Inventor after: Li Guowei Inventor after: Li Chun Inventor after: He Hongxiang Inventor after: Zheng Shuyun Inventor after: Qiu Minmin Inventor after: Zhang Weimin Inventor after: Ruan Xiongjie Inventor after: Xie Wenxi Inventor before: Su Caizhu Inventor before: Zheng Jianguo Inventor before: Liu Erlong Inventor before: Qiu Minmin Inventor before: Zheng Shuyun Inventor before: Cai Yingjun Inventor before: Yao Baihui |
|
| CB03 | Change of inventor or designer information | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170825 Termination date: 20181024 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |