CN107748146A - A kind of crude oil attribute method for quick predicting based near infrared spectrum detection - Google Patents
A kind of crude oil attribute method for quick predicting based near infrared spectrum detection Download PDFInfo
- Publication number
- CN107748146A CN107748146A CN201710984193.2A CN201710984193A CN107748146A CN 107748146 A CN107748146 A CN 107748146A CN 201710984193 A CN201710984193 A CN 201710984193A CN 107748146 A CN107748146 A CN 107748146A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- sample
- mover
- msubsup
- 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.)
- Pending
Links
- 239000010779 crude oil Substances 0.000 title claims abstract description 100
- 238000000034 method Methods 0.000 title claims abstract description 78
- 238000002329 infrared spectrum Methods 0.000 title claims abstract description 49
- 238000001514 detection method Methods 0.000 title claims description 8
- 238000001228 spectrum Methods 0.000 claims abstract description 58
- 238000012549 training Methods 0.000 claims abstract description 47
- 238000012937 correction Methods 0.000 claims abstract description 16
- 238000002835 absorbance Methods 0.000 claims abstract description 11
- 238000000513 principal component analysis Methods 0.000 claims abstract description 10
- 238000012545 processing Methods 0.000 claims abstract description 10
- 239000000203 mixture Substances 0.000 claims description 51
- 239000011159 matrix material Substances 0.000 claims description 19
- 230000008569 process Effects 0.000 claims description 16
- IJGRMHOSHXDMSA-UHFFFAOYSA-N Atomic nitrogen Chemical compound N#N IJGRMHOSHXDMSA-UHFFFAOYSA-N 0.000 claims description 14
- 238000002790 cross-validation Methods 0.000 claims description 11
- 239000002253 acid Substances 0.000 claims description 9
- 238000000605 extraction Methods 0.000 claims description 9
- 230000003595 spectral effect Effects 0.000 claims description 9
- 238000012360 testing method Methods 0.000 claims description 9
- NINIDFKCEFEMDL-UHFFFAOYSA-N Sulfur Chemical compound [S] NINIDFKCEFEMDL-UHFFFAOYSA-N 0.000 claims description 8
- 238000009835 boiling Methods 0.000 claims description 7
- 230000001419 dependent effect Effects 0.000 claims description 7
- 238000004821 distillation Methods 0.000 claims description 7
- 230000000694 effects Effects 0.000 claims description 7
- 229910052757 nitrogen Inorganic materials 0.000 claims description 7
- 229910052717 sulfur Inorganic materials 0.000 claims description 7
- 239000011593 sulfur Substances 0.000 claims description 7
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical group [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 claims description 6
- 238000010238 partial least squares regression Methods 0.000 claims description 5
- 230000008859 change Effects 0.000 claims description 4
- 238000009826 distribution Methods 0.000 claims description 4
- 238000012417 linear regression Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000003993 interaction Effects 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 238000007689 inspection Methods 0.000 claims description 2
- 238000004458 analytical method Methods 0.000 abstract description 11
- 239000000523 sample Substances 0.000 description 117
- 239000003921 oil Substances 0.000 description 12
- 238000005516 engineering process Methods 0.000 description 8
- 238000011156 evaluation Methods 0.000 description 5
- 241000894007 species Species 0.000 description 4
- 238000005259 measurement Methods 0.000 description 3
- 241000772415 Neovison vison Species 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 230000035945 sensitivity Effects 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- UFWIBTONFRDIAS-UHFFFAOYSA-N Naphthalene Chemical compound C1=CC=CC2=CC=CC=C21 UFWIBTONFRDIAS-UHFFFAOYSA-N 0.000 description 1
- 239000005864 Sulphur Substances 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 239000010426 asphalt Substances 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000003610 charcoal Substances 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 239000000039 congener Substances 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000009415 formwork Methods 0.000 description 1
- 239000000295 fuel oil Substances 0.000 description 1
- 239000012535 impurity Substances 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000011005 laboratory method Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000010986 on-line near-infrared spectroscopy Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 239000003208 petroleum Substances 0.000 description 1
- 238000012628 principal component regression Methods 0.000 description 1
- 238000004064 recycling Methods 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 150000003839 salts Chemical class 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000004611 spectroscopical analysis Methods 0.000 description 1
- 238000010561 standard procedure Methods 0.000 description 1
- 239000004575 stone Substances 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/359—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using near infrared light
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/3577—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light for analysing liquids, e.g. polluted water
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N2021/3595—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using FTIR
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N2201/00—Features of devices classified in G01N21/00
- G01N2201/12—Circuits of general importance; Signal processing
- G01N2201/129—Using chemometrical methods
- G01N2201/1293—Using chemometrical methods resolving multicomponent spectra
Landscapes
- Physics & Mathematics (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Other Investigation Or Analysis Of Materials By Electrical Means (AREA)
Abstract
The present invention relates to structure for the method for the near-infrared model of crude oil attribute forecast and the crude oil attribute method for quick predicting detected based near infrared spectrum.The inventive method includes:Detect the attribute of different types of crude oil sample;Determine the atlas of near infrared spectra of the crude oil sample;To 12500~4000cm in the atlas of near infrared spectra that is obtained‑1The absorbance for composing area carries out first differential processing and multiplicative scatter correction;Principal component analysis is carried out to pretreated spectrum, calculates the T2 values of each crude oil sample, the sample more than threshold value is rejected, establishes initial training collection;The principal component scores concentrated using the initial training of acquisition select sample, it is determined that final training set as characteristic variable using the Euclidean distance between sample;One or more wave number sections are selected using moving window PLS;And the near-infrared model established using homing method between crude oil attribute and near infrared spectrum.The present invention is especially suitable for the online fast prediction analysis of crude oil.
Description
Technical field
The present invention relates to a kind of crude oil attribute real-time predicting method based near infrared spectrum detection.
Background technology
With the rapid development of modern industry, oil played as national strategy goods and materials in national economy it is most important
Effect.The petroleum resources in China are deficienter, according to《BP world energy sourceses statistical yearbook 2016》Report, 2015, Chinese stone
Oily net importation amount increases by 770,000 barrel per days, and China turns into the maximum oil importers in the whole world again, and the species of imported crude oil is very
More, wherein it is much so-called " chance oil " to have, either proportion is very big or very high containing acid for they, or impurity is a lot.Have
When, oil tanker, which is arrived at after harbour, just knows it is what oil.This just needs related personnel quickly to be made an appraisal to these crude oil, Er Qieyuan
Depending on oil price wants the result that based on crude is evaluated, if evaluation is time-consuming long, it cannot fix price according to quality in time, it is possible to cause
Economic loss, therefore crude oil Fast Evaluation increasingly shows its importance.
On the other hand, the crude oil change of the refinery of input at present is frequent, will usually become once within 3 two days.When extreme, daily
The crude oil for entering factory is all different.Frequently change can cause many difficulties, especially destilling tower to following process.If crude oil can be made
The speed of evaluation is accelerated, such as is once evaluated for every 10 minutes, it is possible to using the method for automatic governing, as much as possible not
Congener crude oil is suitably concocted, and great fluctuation process is crossed to reduce into the oil property of destilling tower.
In addition, with the continuous heaviness of world wide crude oil, the mink cell focus that refinery of China processes even super mink cell focus
Ratio more and more higher, the requirement to processing of heavy oil are also increasingly finer and efficient.Therefore can not using traditional evaluation method
Meets the needs of practical application.The computer technology continued to develop at present, enable to analyze mass data reality in very short time
It is existing, the rapid development of Modern Analytical Instrument technology is promoted, these have all established solid foundation for crude oil Fast Evaluation.NIR
Analytical technology is that most have one of prospect and most widely used rapid analysis method at present, and On-line near infrared analyzer technology is also to develop most
One of rapid process analysis technique.Because near-infrared analyzer is secondary meter, i.e., near-infrared analyzer can not be straight
Measurement of species attribute is connect, and the mathematical modeling that must first establish between determinand qualitative attribution and near infrared spectrum is then according to mould
Type carrys out measurement of species attribute.Therefore, it is that can near infrared technology effectively be applied to establish the near-infrared model that precision is high, robustness is good
Key.
The content of the invention
In view of the above problems, the present invention proposes a kind of crude oil attribute fast prediction side based near infrared spectrum detection
Method.On the basis of this method is using offline/on-line nir system collection crude oil atlas of near infrared spectra, using first differential and
The method of multiplicative scatter correction pre-processes to the crude oil sample atlas of near infrared spectra collected, eliminates interference;By it is main into
Divide analytical technology and T2 screening sample methods, select the suitable wave-number range of crude oil spectra figure;Finally, according to offset minimum binary
Method, the mathematical modeling established between crude oil property value and its near infrared spectrum data, can be according to real-time by using the model
Spectrum samples (simultaneously) prediction crude oil multiple property values.By offline/on-line nir system, meet to be adapted to crude oil
The quick analysis of matter, particularly online real-time forecast analysis.
First aspect present invention provides a kind of method for building the near-infrared model for crude oil attribute forecast, methods described
Including:
(1) attribute of different types of crude oil sample is detected;
(2) atlas of near infrared spectra of the crude oil sample is determined;
(3) 12500~4000cm in the atlas of near infrared spectra obtained to step (2)-1The absorbance for composing area carries out single order
Differential process and multiplicative scatter correction;
(4) principal component analysis is carried out to the spectrum that step (3) obtains, preserves the principal component scores of each sample, and count
The T2 values of each crude oil sample are calculated, and are contrasted with threshold value, the sample more than threshold value is rejected, establishes initial training collection;
(5) principal component scores of the initial training concentration obtained using step (4) utilize the Europe between sample as characteristic variable
Formula distance selects sample, it is determined that final training set;
(6) one or more wave number sections are selected using moving window PLS, for establishing model;With
(7) near-infrared model established using homing method between crude oil attribute and near infrared spectrum.
In one or more embodiments, the attribute is selected from:Density, sulfur content, nitrogen content, acid number, carbon residue and reality
One or more of boiling point distillation data.
In one or more embodiments, the quantity of step (1) Crude Oil sample is 100~600 parts.
In one or more embodiments, the near-infrared of crude oil sample is gathered using offline or on-line nir system
Spectrum.
In one or more embodiments, in the measure described in step (2), spectral scanning range 4000-
12500cm-1, resolution ratio 2-32cm-1, multiple scanning 10-400 times, take average near infrared light spectrum.
In one or more embodiments, the first differential processing described in step (3) includes with multiplicative scatter correction:
(a) averaged spectrum of sample is calculatedWith
(b) to a spectrum x withCarry out linear regression,B is asked for least square method0And b;
Wherein, the spectrum after processing is xMSC=(x-b0)/b。
In one or more embodiments, in step (4), T2 values are calculated as below:
In above formula, t is variables of the original spectrum matrix X after principal component analysis dimensionality reduction, and σ is t standard deviation, and Iter is
The principal component number of extraction, iter refer to i-th ter principal component, and j refers to j-th of sample.
In one or more embodiments, the threshold value is calculated as below in step (4):Calculate all spectrum in Sample Storehouse
The T2 values of sample, and using 99% confidential interval as upper threshold, the threshold value is calculated by following formula according to F distribution tables, its
In, n is sample number, and F0.001 refers to F inspections, confidence level 0.01:
In one or more embodiments, step (5) includes:Become the principal component scores of each sample as feature
Amount, the Euclidean distance of all samples between any two is calculated, two farthest samples of chosen distance enter training set, then calculate residue
The distance between sample and the two selected samples, choose its minimum value and establish a data set, then from the data set
Choose the sample corresponding to maximum to be added in training set, repeat said process, until choosing z training sample, establish most
Whole training set.
In one or more embodiments, step (5) includes:
E) the Euclidean distance d of all samples between any two is calculated as followsij, chosen distance farthest two samples x1 and x2
Into training set:
In formula, dij:Euclidean distance between i-th of sample and j-th of sample;
m:For number of principal components;
xi:For i-th of sample;With
xj:For j-th of sample;
F) above formula is pressed, calculates the distance between remaining n-2 sample and selected two samples x1 and x2, and respectively take most
Small value min (dI, x1, dI, x2), then choose wherein maximum max (min (dI, x1, dI, x2)) corresponding to sample x3 enter instruction
Practice collection;
G) remaining n-3 sample and selected before three samples x1, the distance between x2 and x3 are calculated, and are respectively taken most
Small value min (dI, x1, dI, x2, dI, x3), then choose wherein maximum max (min (dI, x1, dI, x2, dI, x3)) corresponding to a sample
This x4 enters training set;With
H) said process is repeated, until choosing z training sample.
In one or more embodiments, step (6) includes, first wave number point by spectral window from whole spectrum
Start to be moved rearwards until last, the wave number point fallen in window arrives (i+w-1) for i, and wherein i is the starting wavenumber point of window, w
For window width, (l-w+1) individual wavelet number area for including w wave number point is thus selected successively from whole spectrum;For every
One Ge Zi wave numbers area, sub- PLS regression models are established respectively, and the son for including different PLS number of principal components is calculated using interaction effect
The cross validation root-mean-square error RMSECV of PLS regression models;Then the starting of resulting RMSECV values with corresponding window
Figure is done in position, one or several wavelet number areas containing useful information is selected from figure, to establish final model.
In one or more embodiments, for density, the wave-number range of selection is 4586-5025cm-1;For residual
Charcoal, the wave-number range of selection is 4247-6107cm-1;For acid number, the light wave-number range of selection is 4597-5461cm-1With
6094-7513cm-1;For sulfur content, the wave-number range of selection is 4484-6493cm-1;For nitrogen content, the wave number model of selection
Enclose for 4500-6600cm-1;For wax content, the wave-number range of selection is 4500-6600cm-1;For asphalt content, selection
Wave-number range be 4500-6600cm-1;It is 4599-7500cm with the wave-number range for true boiling point distillation, selection-1。
In one or more embodiments, prediction error RMSECV calculation formula is as follows:
In formula:Z is training set sample number;yiIt is the actual value that i-th of sample measures;It is the predicted value of i-th of sample.
In one or more embodiments, step (7) using PLS establish crude oil attribute with it is near red
Near-infrared model between external spectrum.
In one or more embodiments, the near-infrared model that step (7) is established is:
Y=a0+a1x1+a2x2+…+anxn
Wherein:Y is to predict attribute, aiFor model parameter, xiFor the absorbance of i-th of wave number point of spectrum.
In one or more embodiments, step (7) includes:
(a) first pair of composition of two set of variables is extracted respectively, and is allowed to correlation up to maximum;
It is proposed that first pair of composition is t respectively from two groups of variables1And u1, t1It is independent variable collection X=(x1..., xm)TLinear group
Close:u1It is dependent variable collection Y=(y1..., yp)TLinear combination:Wherein, it is desirable to t1And u1Each change of set of variables where extraction as much as possible
Different information;And t1And u1Degree of correlation reach maximum;
By the standardization observation data matrix E of two groups of variables sets0And F0The score vector of first pair of composition is calculated, is designated asWith
Wherein, E0Remember for spectrum matrix, F0Remember for concentration matrix, it is as follows:
Use the score vector of first pair of compositionInner product calculate first couple of composition t1And u1Covariance Cov (t1,
u1), so two above requirement can turn to constrained extremal problem mathematically, it is shown below:
Using Lagrange Multiplier Methods, problem, which turns to, seeks unit vector w1And v1, makeIt is maximum;
By calculating m m matrixCharacteristic value and characteristic vector, and M eigenvalue of maximum isIt is corresponding single
Position characteristic vector is exactly required solution w1, and v1By w1It is calculated
(b) y is established1,…,ypTo t1Recurrence and x1,…,xmTo t1Recurrence;
If regression model is:
Wherein α1=(α11... α1m)T, β1=(β11... β1m)TIt is the parameter vector in many-to-one regression model respectively,
E1And F1It is residual error battle array, regression coefficient vector α1, β1Least-squares estimation be
Claim α1, β1For model effect load;
(c) residual error battle array E is used1And F1Instead of E0And F0Repeat above step;
NoteThen residual error battle arrayIf residual error battle array F1
The absolute value of middle element is approximately 0, then it is assumed that the regression equation precision established with first composition has met needs, can stop taking out
Composition is taken, otherwise with residual error battle array E1And F1Instead of E0And F0Above step is repeated to produce:
w2=(W21,…w2m)T;v2=(v21... v2p)T
AndFor the score vector of second pair of composition;
The load of respectively Y, X second pair of composition, at this moment
(d) z × m data matrixes E is set0Order be r≤min (z-1, m), then r composition t be present1, t2..., trSo that
Substitute into Y=t1β1+…+trβr, p are produced because becoming
The PLS equation of amount:
yj=aj1x1+…+ajmxm, (j=1,2 ..., p)
HereMeet
In one or more embodiments, step (7) also includes cross validation test:I.e. for extraction needed for modeling
Composition number l, determined by cross validation test:Cast out i-th of observation (i=1,2 ..., n) every time, to remaining
N-1 observation is modeled with partial least-square regression method, and considers to extract the regression equation being fitted after h composition, then house
I-th of the observation station gone substitutes into be fitted regression equation, obtains yj(j=1,2 ..., p) it is pre- in i-th of observation station
Measured value y(i)j(h);The checking of the above is repeated to i=1,2 ..., n, produces j-th of dependent variable y when extracting h compositionj(j=1,
2 ..., p) Prediction sum squares, be
Y=(y1..., yp)TPrediction sum squares be
In addition, using all sample points again, the regression equation containing h composition is fitted;At this moment, i-th sample point is remembered
Predicted value isBy yjError sum of squares be defined as
Define Y error sum of squares be
When PRESS (h) reaches minimum value, corresponding h is required composition number;
Defining Cross gain modulation isCalculate and terminate in each step of modeling
Before, cross validation test is carried out, if walked in hThen model reaches precision and wanted
Ask, stop extract component;IfRepresent the t of h step extractionshThe contributrion margin of composition is notable, should continue h+1
Step calculates.
Second aspect of the present invention provides a kind of crude oil attribute forecast method based near infrared spectrum detection, methods described bag
Include:
(A) atlas of near infrared spectra of crude oil to be detected is determined;
(B) 12500~4000cm in the atlas of near infrared spectra obtained to step (A)-1The absorbance for composing area carries out single order
Differential process and multiplicative scatter correction;
(C) near-infrared model for crude oil attribute forecast that structure obtains herein is utilized to carry out the crude oil attribute pre-
Survey.
In one or more embodiments, the attribute is selected from:Density, sulfur content, nitrogen content, acid number, carbon residue and reality
One or more of boiling point distillation data.
It is to be detected using the collection of offline or on-line nir system in step (A) in one or more embodiments
The near infrared spectrum of crude oil.
In one or more embodiments, in the measure described in step (A), spectral scanning range 4000-
12500cm-1, resolution ratio 2-32cm-1, multiple scanning 10-400 times, take average near infrared light spectrum.
In one or more embodiments, the first differential processing described in step (B) includes with multiplicative scatter correction:
(a) averaged spectrum of each sample is calculated
(b) to a spectrum x withCarry out linear regression,B is asked for least square method0And b;With
(c) spectrum after handling is xMSC=(x-b0)/b。
In one or more embodiments, in step (C), in 12500~4000cm-1Selection wave number area in wavenumber range
Between, the crude oil attribute is predicted using the near-infrared model for crude oil attribute forecast that structure obtains herein.
Beneficial effects of the present invention are as follows:Comprehensive modeling method proposed by the present invention, i.e., gathered based on near-infrared analyzer
Crude oil spectra figure, the crude oil sample atlas of near infrared spectra collected is carried out using the method for first differential and multiplicative scatter correction
Pretreatment, eliminate interference;By principal component analysis technology and T2 screening sample methods, the suitable wave number model of crude oil spectra figure is selected
Enclose;Finally, according to PLS, the mathematical modeling established between crude oil property value and its near infrared spectrum data, realize
The fast prediction analysis of unknown crude oil property value.This method is adapted to offline, on-line nir system.Using entering infrared point online
Analyzer, the near infrared spectrum of crude oil sample is obtained in real time, can be realized based on the present invention pre- real-time to the progress of crude oil attribute
Survey.For the quick analysis demand of crude oil, crude oil near infrared spectrum comprehensive modeling method of the present invention, have pretreated near red
The advantages that external spectrum signal to noise ratio is high, analyze speed is fast, modeling accuracy is high, the present invention can detect the close of the oil product of crude oil simultaneously
Degree, carbon residue, nitrogen content, sulfur content, acid number, true boiling point distillation yield etc.;Solve the time-consuming, laborious etc. of conventional method of analysis to ask
Topic, improves analysis efficiency, is a kind of effective method.
Brief description of the drawings
Fig. 1:The flow chart of crude oil attribute method for quick predicting based near infrared spectrum detection.
Fig. 2:Crude oil near-infrared primary light spectrogram.
Fig. 3:Pretreated crude oil near-infrared spectrogram.
Fig. 4:The error line that moving window offset minimum binary obtains.
Fig. 5:Near-infrared oil density value regression model.
Embodiment
The flow chart of crude oil attribute method for quick predicting of Fig. 1 displays present invention based near infrared spectrum detection, while
The near-infrared model building process of crude oil attribute fast prediction of the present invention is shown, specifically includes following steps:
(1) attribute of different types of crude oil sample is detected;
(2) atlas of near infrared spectra of the crude oil sample is determined;
(3) 12500~4000cm in the atlas of near infrared spectra obtained to step (2)-1The absorbance for composing area carries out single order
Differential process and multiplicative scatter correction;
(4) principal component analysis is carried out to the spectrum that step (3) obtains, preserves the principal component scores of each sample, and count
The T2 values of each crude oil sample are calculated, and are contrasted with threshold value, the sample more than threshold value is rejected, establishes initial training collection;
(5) principal component scores of the initial training concentration obtained using step (4) utilize the Europe between sample as characteristic variable
Formula distance selects sample, it is determined that final training set;
(6) one or more wave number sections are selected using moving window PLS, for establishing model;With
(7) near-infrared model established using homing method between crude oil attribute and near infrared spectrum.
(1) attribute of different types of crude oil sample is detected
Different types of crude oil sample is collected, generally covers paraffinic base crude oil, intermediate base crude and naphthene base crude etc..
Generally, collected crude oil sample quantity is no less than 50.It is preferred that the density (20 DEG C) of collected crude oil sample, sulphur
Content and acid number index control the model in 0.7~1.1g/cm3,0.03%~5.50% and 0.01~12.00mgKOH/g respectively
Within enclosing.Then multiple attributes of collected crude oil are measured using traditional standard method, as density, carbon residue, nitrogen content,
Sulfur content, acid number, salt content, wax content, asphalitine and true boiling point distillation yield etc., and record data.Crude oil species can be 50
More than individual, preferably at least more than 100.
(2) atlas of near infrared spectra of the crude oil sample is determined
The offline or On-line NIR instrument that suitable types can be chosen is equipped with the Transflective near-infrared fibre-optical of specific model
Probe is scanned, and fibre-optical probe light path can select in 0.5mm~20mm scopes, straight using Transflective near-infrared fibre-optical is popped one's head in
The metering system patched into crude oil sample (crude oil sample temperature maintains 30 DEG C) obtains the near infrared spectrum of every part of sample.Every
The spectral scan time is 10-400 times, is averaged.Spectral scanning range is 4000-12500cm-1, resolution ratio 2-32cm-1.It is former
Oily pre-processed spectrum is shown in Fig. 2.
(3) 12500~4000cm in the atlas of near infrared spectra obtained to step (2)-1The absorbance for composing area carries out single order
Differential process and multiplicative scatter correction
The atlas of near infrared spectra of the crude oil sample of collection is pre-processed, preprocess method is to 12500-4000cm-1
Spectrum area carry out first differential processing, eliminate the interference of baseline and other backgrounds, resolution ratio and sensitivity improved, afterwards to spectrum
Carry out multiplicative scatter correction.Specific method is as follows:
A) averaged spectrum of each sample is calculated(i.e. " preferable spectrum ");With
B) to a spectrum x (1 × m) withCarry out linear regression,B is asked for least square method0And b;
Wherein, pretreated spectrum is xMSC=(x-b0)/b。
The interference of baseline and other backgrounds can be effectively eliminated using first differential method, weight peak is differentiated, improves and differentiate
Rate and sensitivity;And multiplicative scatter correction can then eliminate that distribution of particles is uneven and granular size caused by scattering influence, enter
And establish initial training collection.
(4) principal component analysis is carried out to the spectrum that step (3) obtains, preserves the principal component scores of each sample, and count
The T2 values of each crude oil sample are calculated, and are contrasted with threshold value, the sample more than threshold value is rejected, establishes initial training collection
Step (4) is that the selection of sample is corrected to the spectrum after pretreated in step (3), is selected from Sample Storehouse
Representative strong Sample Establishing model, the storage area established speed, reduce model library of model so can be not only improved, more
For it is important that it is easy to the renewal and maintenance of model.
Before training sample set is selected, rejecting abnormalities sample, these exceptional samples it should may contain abnormal chemical first
Component or concentration of component are more extreme, significant difference be present with other samples., can shadow if these sample points participate in the foundation of model
Ring the accuracy and robustness of calibration model.Rejecting abnormalities sample can be counted to detect exceptional value using T2, and T2 is united
Larger sample is measured therefrom to reject.
The description formula of T2 statistics is as follows:
In above formula, t is variables of the original spectrum matrix X after PCA dimensionality reductions, and σ is t standard deviation, and Iter is extraction
Principal component number, iter refer to i-th ter principal component, and j refers to j-th of sample.Because the T2 values of exceptional sample can be far longer than normally
Sample, so the T2 values of the spectrum samples in all Sample Storehouses are calculated, and using 99% confidential interval as upper threshold, under
Formula, and look into F distribution tables threshold value can be calculated,
By the T2 values of all samples in Sample Storehouse compared with threshold value, the sample more than threshold value is rejected, establishes initial instruction
Practice collection.After obtaining initial training collection, using the Euclidean distance between variable, sample is chosen in feature space.
(5) principal component scores of the initial training concentration obtained using step (4) utilize the Europe between sample as characteristic variable
Formula distance selects sample, it is determined that final training set
After carrying out principal component (PCA) to sample spectrum, characteristic variable is used as by the use of principal component scores, between recycling sample
Euclidean distance select sample, it is determined that final training set.Comprise the following steps that:
If sharing n sample, z sample is therefrom selected;
A) the Euclidean distance d of all samples between any two is calculated as followsij, chosen distance farthest two samples x1 and x2
Into training set:
In formula, dij:Euclidean distance between i-th of sample and j-th of sample;
m:For number of principal components;
xi:For i-th of sample;
xj:For j-th of sample;
B) above formula is pressed, calculates the distance between remaining n-2 sample and selected two samples x1 and x2, and respectively take most
Small value min (dI, x1, dI, x2), then choose wherein maximum max (min (dI, x1, dI, x2)) corresponding to sample x3 enter instruction
Practice collection;
C) remaining n-3 sample and selected before three samples x1, the distance between x2 and x3 are calculated, and are respectively taken most
Small value min (dI, x1, dI, x2, dI, x3), then choose wherein maximum max (min (dI, x1, dI, x2, dI, x3)) corresponding to a sample
This x4 enters training set;
D) said process is repeated, until choosing z training sample.
(6) one or more wave number sections are selected using moving window PLS, for establishing model
The step (6) of the present invention is to carry out wave number selection to the spectrum samples in training set, and traditional viewpoint thinks polynary
Bearing calibration has stronger antijamming capability, can all-wave number participate in model foundation, but with to offset minimum binary etc. side
The further investigation of method, characteristic waves are screened by ad hoc approach or section is possible to obtain preferably quantitative model.Pass through wave number
Selection can be with simplified model, and can reject incoherent variable by wave number selection, and it is stronger to obtain predictive ability, robustness
More preferable model.
The present invention uses moving window PLS, and the basic thought of this method is the spectrum by a width for w
Window is moved rearwards since first wave number point of whole spectrum up to last, and it is i to (i+w- to fall the wave number point in window
1), wherein i is the starting wavenumber point of window, and w is window width.So, we select can successively from whole spectrum
(l-w+1) the individual wavelet number area comprising w wave number point establishes sub- PLS regression models for each wavelet number area respectively, and
The intersection that the sub- PLS models comprising different PLS number of principal components (having a maximum PLS principal components limitation) are calculated using interaction effect is tested
Demonstrate,prove root-mean-square error RMSECV.Then resulting RMSECV values can be made Fig. 4 by we with the original position of corresponding window
It is shown.We can easily select one or several wavelet number areas containing useful information from figure.Recycle what is selected
Wavelet number area establishes final model.
Predict that error RMSECV calculation formula is as follows:
In formula:Z is training set sample number;yiIt is the actual value that i-th of sample measures;It is the predicted value of i-th of sample.
(7) near-infrared model established using homing method between crude oil attribute and near infrared spectrum
The step (7) of the present invention is to utilize the near-infrared by the crude oil sample in the training set of pretreatment and wave number selection
Spectrogram establishes model with property value, and the property value for unknown sample is predicted.Built in the present invention using PLS
Formwork erection type, the main thought of this method are as follows:
PLS not only only accounts for spectrum matrix, while have also contemplated that concentration compared with principal component regression
The influence of matrix, for convenience's sake, spectrum matrix are designated as E0, concentration matrix is designated as F0, it is as follows:
Comprise the following steps that:
(1) first pair of composition of two set of variables is extracted respectively, and is allowed to correlation up to maximum.
Assuming that propose that first pair of composition is t respectively from two groups of variables1And u1, t1It is independent variable collection X=(x1..., xm)TLine
Property combination:u1It is dependent variable collection Y=(y1..., yp)TLinear combination:For the needs of regression analysis, it is desirable to:
1)t1And u1The variation information of set of variables where each extracting as much as possible;
2)t1And u1Degree of correlation reach maximum.
By the standardization observation data matrix E of two groups of variables sets0And F0, the score vector of first pair of composition can be calculated, is designated asWith
First couple of composition t1And u1Covariance Cov (t1,u1) score vector of first pair of composition can be usedInner product come
Calculate.So two above requirement can turn to constrained extremal problem mathematically:
Using Lagrange Multiplier Methods, problem, which turns to, seeks unit vector w1And v1, makeIt is maximum.
The solution of problem need only be by calculating m m matrixCharacteristic value and characteristic vector, and M maximum feature
It is worth and isCorresponding unit character vector is exactly required solution w1, and v1Can be by w1It is calculated
(2) y is established1,…,ypTo t1Recurrence and x1,…,xmTo t1Recurrence.
It is assumed that regression model is:
Wherein α1=(α11... α1m)T, β1=(β11... β1m)TIt is the parameter vector in many-to-one regression model respectively,
E1And F1It is residual error battle array.Regression coefficient vector α1, β1Least-squares estimation be
Claim α1, β1For model effect load.
(3) residual error battle array E is used1And F1Instead of E0And F0Repeat above step.
NoteThen residual error battle arrayIf residual error battle array F1
The absolute value of middle element is approximately 0, then it is assumed that the regression equation precision established with first composition has met needs, can stop
Extract composition.Otherwise residual error battle array E is used1And F1Instead of E0And F0Above step is repeated to produce:
w2=(w21... w2m)T;v2=(v21... v2p)T
AndFor the score vector of second pair of composition.
The load of respectively Y, X second pair of composition.At this moment have
(4) z × m data matrixes E is set0Order be r≤min (z-1, m), then r composition t be present1, t2..., trSo that
Substitute into Y=t1β1+…+trβr, produce p because
The PLS equation of variable:
yj=aj1x1+…+ajmxm, (j=1,2 ..., p)
HereMeet
(5) cross validation test.
Generally, PLS and existing r composition t need not be selected1, t2..., trReturned to establish
Formula, and as principal component analysis, only select preceding l composition (r >=l), you can obtain the preferable regression model of predictive ability.It is right
In the composition number l extracted needed for modeling, can be determined by cross validation test.
Cast out i-th of observation (i=1,2 ..., n) every time, to remaining n-1 observation PLS side
Method models, and considers to extract the regression equation being fitted after h composition, then i-th of the observation station cast out is substituted into time being fitted
Return equation, obtain yjThe predicted value y of (j=1,2 ..., p) in i-th of observation station(i)j(h).To i=1,2 ..., n repeat with
On checking, produce extract h composition when j-th of dependent variable yjThe Prediction sum squares of (j=1,2 ..., p) are
Y=(y1..., yp)TPrediction sum squares be
In addition, using all sample points again, the regression equation containing h composition is fitted.At this moment, i-th sample point is remembered
Predicted value isY can then be definedjError sum of squares be
Define Y error sum of squares be
When PRESS (h) reaches minimum value, corresponding h is required composition number.Generally, always there is PRESS (h) big
In SS (h), and SS (h) is then less than SS (h-1).Therefore, in extract component, always wish that ratio PRESS (h)/SS (h) is smaller more
It is good;Limits value can typically be set as 0.05, that is, worked as
When, increase composition thBe advantageous to the raising of model accuracy.Or conversely speaking, when
When, it is considered as increasing new composition th, to reducing the prediction error of equation without obvious improvement result.
Therefore, definition Cross gain modulation isSo, in each step of modeling
Before calculating terminates, cross validation test is carried out, if had in h steps Then model reaches
To required precision, extract component can be stopped;IfRepresent the t of h step extractionshThe contributrion margin of composition is notable,
H+1 steps should be continued to calculate.
The present invention determines the atlas of near infrared spectra of crude oil sample to be measured first when predicting the property value of sample to be tested, and
And using the method described in step (3), the atlas of near infrared spectra of crude oil sample to be measured is pre-processed, afterwards according to step
(6) the wave-number range selection variable of selection, utilizes the multiple attributes for the model prediction crude oil sample to be measured established in step (7)
Value.
The inventive method can predict original by establishing the model between the atlas of near infrared spectra of crude oil sample and property value
A variety of physical datas of oil, effective, reliable foundation is provided for industrial process, (quick to divide especially suitable for on-line prediction simultaneously
Analysis) crude oil multiple property values such as, density, carbon residue, nitrogen content, sulfur content, acid number, true boiling point distillation yield etc..
The present invention is specifically described below by embodiment, this example is not limited to this by taking oil density as an example.
It is necessarily pointed out that following examples are served only for, the invention will be further described, it is impossible to is interpreted as protecting the present invention
Protect the limitation of scope, some nonessential improvement and tune that professional and technical personnel in the field makes according to present disclosure
It is whole, still fall within protection scope of the present invention.
Embodiment 1
The instrument that crude oil near infrared spectrum is gathered in embodiment is Brooker Matrix-F Fourier Transform Near Infrareds
Instrument, spectral region scanning range are 4000-12500cm-1, resolution ratio 32cm-1, add up scanning times 32 times, transflector measurement side
Formula.
Step 1:Different types of crude oil sample 150 is gathered, the near infrared spectrum of crude oil sample is determined, such as Fig. 2 institutes
Show.And the density of crude oil sample is measured according to laboratory method.
Step 2:Choose 8000-4000cm-1The absorbance of Spectral range, first differential+multiplicative scatter correction is carried out to it
Pretreatment, establishes crude oil sample near infrared light spectrum matrix.Fig. 2 is the original atlas of near infrared spectra of crude oil sample of measure, spectrum
Baseline drift is serious, and peak overlap is serious;Fig. 3 is the spectrogram after pretreatment.
Step 3:To pretreated crude oil sample spectrum matrix carry out principal component analysis, preserve each sample it is main into
Get point, and calculate the T2 values of each crude oil sample, and contrasted with threshold value 7.06, reject the sample more than threshold value
This, establishes initial training collection.
Step 4:Using the principal component scores of each sample recorded before as characteristic variable, all samples are calculated first
Euclidean distance between any two, two farthest samples of chosen distance enter training set, then calculate remaining sample and selected
The distance between the two samples, and choose its minimum value and establish a data set, then choose maximum from the data set
Corresponding sample is added in training set, said process is repeated, until choosing 80 training samples.
Step 5:Wave number selection is carried out to it using moving window PLS to the sample in training set, wherein
Window size is 21, and the selection of the wavelet number of gained indulges seat as shown in figure 4, abscissa is the initial point position of moving window in figure
The cross validation prediction error RMSECV for the PLS models that corresponding moving window includes the foundation of wavelet number area is designated as, is therefrom chosen
Wavenumber range is 4586-5025cm-1。
Step 6:Establish the regression model of density value and near infrared spectrum with PLS, prediction attribute with it is near
Relation such as following formula between infrared spectrum
Y=aO+a1x1+a2x2++anxn
Wherein:Y is to predict attribute, aiFor model parameter, xiFor the absorbance of i-th of wave number point of spectrum.
The regression model built is as shown in Figure 5.
In addition, also verified using checking the set pair analysis model.Specifically, to crude oil sample to be measured first according to step
Method in two, to 8000-4000cm-1First differential+multiplicative scatter correction pretreatment is carried out in spectrum area to it, is selected afterwards
4586-5025cm-1Wavenumber range, finally it is predicted using the model in step 6.The coefficient of determination up to 0.9593,
Validation-cross mean square error is 0.000571.
Based on the above method, the comparative result of oil density predicted value and actual value based near infrared spectroscopic method is such as
Shown in table 1, the model prediction is accurate, efficient.
Table 1:Predicted value and actual value result
Claims (10)
- A kind of 1. method for building the near-infrared model for crude oil attribute forecast, it is characterised in that methods described includes:(1) attribute of different types of crude oil sample is detected;(2) atlas of near infrared spectra of the crude oil sample is determined;(3) 12500~4000cm in the atlas of near infrared spectra obtained to step (2)-1The absorbance for composing area carries out first differential Processing and multiplicative scatter correction;(4) principal component analysis is carried out to the spectrum that step (3) obtains, preserves the principal component scores of each sample, and calculated every The T2 values of one crude oil sample, and contrasted with threshold value, the sample more than threshold value is rejected, establishes initial training collection;(5) principal component scores concentrated using the initial training that step (4) obtains as characteristic variable, using between sample it is European away from From sample is selected, it is determined that final training set;(6) one or more wave number sections are selected using moving window PLS, for establishing model;With(7) near-infrared model established using homing method between crude oil attribute and near infrared spectrum.
- 2. the method as described in claim 1, it is characterised in thatThe attribute is selected from:One or more of density, sulfur content, nitrogen content, acid number, carbon residue and true boiling point distillation data;The quantity of step (1) Crude Oil sample is 100~600 parts;Using offline or on-line nir system collection crude oil sample The near infrared spectrum data of product.
- 3. method as claimed in claim 1 or 2, it is characterised in that in the measure described in step (2), spectral scanning range is 4000-12500cm-1, resolution ratio 2-32cm-1, multiple scanning 10-400 times, take average near infrared light spectrum.
- 4. such as the method any one of claim 1-3, it is characterised in that the processing of first differential described in step (3) with Multiplicative scatter correction includes:(a) averaged spectrum of each sample is calculatedWith(b) to a spectrum x withCarry out linear regression,B is asked for least square method0And b;Wherein, the spectrum after processing is xMSC=(x-b0)/b。
- 5. such as the method any one of claim 1-3, it is characterised in thatIn step (4), T2 values are calculated as below:<mrow> <msubsup> <mi>T</mi> <mi>j</mi> <mn>2</mn> </msubsup> <mo>=</mo> <munderover> <mo>&Sigma;</mo> <mrow> <mi>i</mi> <mi>t</mi> <mi>e</mi> <mi>r</mi> <mo>=</mo> <mn>1</mn> </mrow> <mrow> <mi>I</mi> <mi>t</mi> <mi>e</mi> <mi>r</mi> </mrow> </munderover> <mfrac> <msubsup> <mi>t</mi> <mrow> <mi>i</mi> <mi>t</mi> <mi>e</mi> <mi>r</mi> </mrow> <mn>2</mn> </msubsup> <msubsup> <mi>&sigma;</mi> <msub> <mi>t</mi> <mrow> <mi>i</mi> <mi>t</mi> <mi>e</mi> <mi>r</mi> </mrow> </msub> <mn>2</mn> </msubsup> </mfrac> </mrow>In above formula, t is variables of the original spectrum matrix X after principal component analysis dimensionality reduction, and σ is t standard deviation, and Iter is extraction Principal component number, iter refers to i-th ter principal component, and j refers to j-th of sample;In step (4), the threshold value is calculated as below:The T2 values of all spectrum samples in Sample Storehouse are calculated, and are put with 99% Letter section is upper threshold, and the threshold value is calculated by following formula according to F distribution tables, wherein, n is sample number, and F0.001 refers to F inspections Test, confidence level 0.01:<mrow> <msup> <mi>T</mi> <mn>2</mn> </msup> <mi>c</mi> <mi>r</mi> <mi>i</mi> <mi>t</mi> <mo>=</mo> <mfrac> <mrow> <mi>I</mi> <mi>t</mi> <mi>e</mi> <mi>r</mi> <mrow> <mo>(</mo> <mi>n</mi> <mo>-</mo> <mn>1</mn> <mo>)</mo> </mrow> <mrow> <mo>(</mo> <mi>n</mi> <mo>+</mo> <mn>1</mn> <mo>)</mo> </mrow> </mrow> <mrow> <mi>n</mi> <mrow> <mo>(</mo> <mi>n</mi> <mo>-</mo> <mi>I</mi> <mi>t</mi> <mi>e</mi> <mi>r</mi> <mo>)</mo> </mrow> </mrow> </mfrac> <msub> <mi>F</mi> <mn>0.01</mn> </msub> <mrow> <mo>(</mo> <mi>I</mi> <mi>t</mi> <mi>e</mi> <mi>r</mi> <mo>,</mo> <mi>n</mi> <mo>-</mo> <mi>I</mi> <mi>t</mi> <mi>e</mi> <mi>r</mi> <mo>)</mo> </mrow> <mo>.</mo> </mrow>
- 6. such as the method any one of claim 1-5, it is characterised in that step (5) includes:By the master of each sample Component score calculates the Euclidean distance of all samples between any two, two farthest samples of chosen distance enter as characteristic variable Enter training set, then calculate the distance between remaining sample and the two selected samples, choose its minimum value and establish a number According to collection, then from the data set choose maximum corresponding to sample be added in training set, repeat said process, until choosing Z training sample, establish final training set;Preferably, step (5) includes:A) the Euclidean distance d of all samples between any two is calculated as followsij, chosen distance farthest two samples x1 and x2 enter Training set:<mrow> <msub> <mi>d</mi> <mrow> <mi>i</mi> <mi>j</mi> </mrow> </msub> <mo>=</mo> <msqrt> <mrow> <munderover> <mo>&Sigma;</mo> <mrow> <mi>k</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>m</mi> </munderover> <msup> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mrow> <mi>i</mi> <mi>k</mi> </mrow> </msub> <mo>-</mo> <msub> <mi>x</mi> <mrow> <mi>j</mi> <mi>k</mi> </mrow> </msub> <mo>)</mo> </mrow> <mn>2</mn> </msup> </mrow> </msqrt> </mrow>In formula, dij:Euclidean distance between i-th of sample and j-th of sample;m:For number of principal components;xi:For i-th of sample;Withxj:For j-th of sample;B) above formula is pressed, calculates the distance between remaining n-2 sample and selected two samples x1 and x2, and respectively take minimum value min(dI, x1, dI, x2), then choose wherein maximum max (min (dI, x1, dI, x2)) corresponding to sample x3 enter training Collection;C) remaining n-3 sample and selected before three samples x1, the distance between x2 and x3 are calculated, and respectively take minimum value min(dI, x1, dI, x2, dI, x3), then choose wherein maximum max (min (dI, x1, dI, x2, dI, x3)) corresponding to a sample x4 Into training set;WithD) said process is repeated, until choosing z training sample.
- 7. such as the method any one of claim 1-6, it is characterised in that step (6) includes, by spectral window from whole First wave number point of individual spectrum starts to be moved rearwards until last, and it is i to (i+w-1), wherein i to fall the wave number point in window For the starting wavenumber point of window, w is window width, thus selects that (l-w+1) is individual to include w wave number successively from whole spectrum The wavelet number area of point;For each wavelet number area, sub- PLS regression models are established respectively, and calculate and include using interaction effect The cross validation root-mean-square error RMSECV of the sub- PLS regression models of different PLS number of principal components;Then resulting RMSECV Value does figure with the original position of corresponding window, one or several wavelet number areas containing useful information is selected from figure, to build Vertical final model;Preferably, step (6) is included in 4000-12500cm-1Wave-number range in the one or more wave number sections of selection, preferably exist 4000-8000cm-1Wave-number range in;Preferably, predict that error RMSECV calculation formula is as follows:<mrow> <mi>R</mi> <mi>M</mi> <mi>S</mi> <mi>E</mi> <mi>C</mi> <mi>V</mi> <mo>=</mo> <msqrt> <mrow> <mfrac> <mn>1</mn> <mrow> <mi>z</mi> <mo>-</mo> <mn>1</mn> </mrow> </mfrac> <munderover> <mo>&Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>z</mi> </munderover> <msup> <mrow> <mo>(</mo> <msub> <mi>y</mi> <mi>i</mi> </msub> <mo>-</mo> <mover> <msub> <mi>y</mi> <mi>i</mi> </msub> <mo>^</mo> </mover> <mo>)</mo> </mrow> <mn>2</mn> </msup> </mrow> </msqrt> </mrow>In formula:Z is training set sample number;yiIt is the actual value that i-th of sample measures;It is the predicted value of i-th of sample.
- 8. such as the method any one of claim 1-7, it is characterised in that step (7) is using PLS The near-infrared model established between crude oil attribute and near infrared spectrum;Preferably, the near-infrared model that step (7) is established is:Y=a0+a1x1+a2x2+…+anxnWherein:Y is to predict attribute, aiFor model parameter, xiFor the absorbance of i-th of wave number point of spectrum.
- 9. method as claimed in claim 8, it is characterised in that step (7) includes:(a) first pair of composition of two set of variables is extracted respectively, and is allowed to correlation up to maximum;It is proposed that first pair of composition is t respectively from two groups of variables1And u1, t1It is independent variable collection X=(x1..., xm)TLinear combination:u1It is dependent variable collection Y=(y1..., yp)TLinear combination:Wherein, it is desirable to t1And u1Each change of set of variables where extraction as much as possible Different information;And t1And u1Degree of correlation reach maximum;By the standardization observation data matrix E of two groups of variables sets0And F0The score vector of first pair of composition is calculated, is designated asWithWherein, E0Remember for spectrum matrix, F0Remember for concentration matrix, it is as follows:Use the score vector of first pair of compositionInner product calculate first couple of composition t1And u1Covariance Cov (t1,u1), So two above requirement can turn to constrained extremal problem mathematically, it is shown below:<mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mrow> <mo><</mo> <mover> <msub> <mi>t</mi> <mn>1</mn> </msub> <mo>^</mo> </mover> <mo>,</mo> <mover> <msub> <mi>u</mi> <mn>1</mn> </msub> <mo>^</mo> </mover> <mo>></mo> <mo>=</mo> <mo><</mo> <msub> <mi>E</mi> <mn>0</mn> </msub> <msub> <mi>w</mi> <mn>1</mn> </msub> <mo>,</mo> <msub> <mi>F</mi> <mn>0</mn> </msub> <msub> <mi>v</mi> <mn>1</mn> </msub> <mo>></mo> <mo>=</mo> <msubsup> <mi>w</mi> <mn>1</mn> <mi>T</mi> </msubsup> <msubsup> <mi>E</mi> <mn>0</mn> <mi>T</mi> </msubsup> <msub> <mi>F</mi> <mn>0</mn> </msub> <msub> <mi>v</mi> <mn>1</mn> </msub> <mo>&DoubleRightArrow;</mo> <mi>m</mi> <mi>a</mi> <mi>x</mi> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <msubsup> <mi>w</mi> <mn>1</mn> <mi>T</mi> </msubsup> <mi>w</mi> <mo>=</mo> <mo>|</mo> <mo>|</mo> <msub> <mi>w</mi> <mn>1</mn> </msub> <mo>|</mo> <msup> <mo>|</mo> <mn>2</mn> </msup> <mo>=</mo> <mn>1</mn> <mo>,</mo> <msubsup> <mi>v</mi> <mn>1</mn> <mi>T</mi> </msubsup> <mi>v</mi> <mo>=</mo> <mo>|</mo> <mo>|</mo> <msub> <mi>v</mi> <mn>1</mn> </msub> <mo>|</mo> <msup> <mo>|</mo> <mn>2</mn> </msup> <mo>=</mo> <mn>1</mn> </mrow> </mtd> </mtr> </mtable> </mfenced>Using Lagrange Multiplier Methods, problem, which turns to, seeks unit vector w1And v1, makeIt is maximum;Pass through Calculate m m matrixCharacteristic value and characteristic vector, and M eigenvalue of maximum isCorresponding unit is special Sign vector is exactly required solution w1, and v1By w1It is calculated(b) y is established1,…,ypTo t1Recurrence and x1,…,xmTo t1Recurrence;If regression model is:<mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mrow> <msub> <mi>E</mi> <mn>0</mn> </msub> <mo>=</mo> <mover> <msub> <mi>t</mi> <mn>1</mn> </msub> <mo>^</mo> </mover> <msubsup> <mi>&alpha;</mi> <mn>1</mn> <mi>T</mi> </msubsup> <mo>+</mo> <msub> <mi>E</mi> <mn>1</mn> </msub> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <msub> <mi>F</mi> <mn>0</mn> </msub> <mo>=</mo> <mover> <msub> <mi>u</mi> <mn>1</mn> </msub> <mo>^</mo> </mover> <msubsup> <mi>&beta;</mi> <mn>1</mn> <mi>T</mi> </msubsup> <mo>+</mo> <msub> <mi>F</mi> <mn>1</mn> </msub> </mrow> </mtd> </mtr> </mtable> </mfenced>Wherein α1=(α11... α1m)T, β1=(β11... β1m)TIt is the parameter vector in many-to-one regression model respectively, E1And F1 It is residual error battle array, regression coefficient vector α1, β1Least-squares estimation be<mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mrow> <msub> <mi>&alpha;</mi> <mn>1</mn> </msub> <mo>=</mo> <msubsup> <mi>E</mi> <mn>0</mn> <mi>T</mi> </msubsup> <mover> <msub> <mi>t</mi> <mn>1</mn> </msub> <mo>^</mo> </mover> <mo>/</mo> <mo>|</mo> <mo>|</mo> <mover> <msub> <mi>t</mi> <mn>1</mn> </msub> <mo>^</mo> </mover> <mo>|</mo> <msup> <mo>|</mo> <mn>2</mn> </msup> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <msub> <mi>&beta;</mi> <mn>1</mn> </msub> <mo>=</mo> <msubsup> <mi>F</mi> <mn>0</mn> <mi>T</mi> </msubsup> <mover> <msub> <mi>t</mi> <mn>1</mn> </msub> <mo>^</mo> </mover> <mo>/</mo> <mo>|</mo> <mo>|</mo> <mover> <msub> <mi>t</mi> <mn>1</mn> </msub> <mo>^</mo> </mover> <mo>|</mo> <msup> <mo>|</mo> <mn>2</mn> </msup> </mrow> </mtd> </mtr> </mtable> </mfenced>Claim α1, β1For model effect load;(c) residual error battle array E is used1And F1Instead of E0And F0Repeat above step;NoteThen residual error battle arrayIf residual error battle array F1Middle member The absolute value of element is approximately 0, then it is assumed that the regression equation precision established with first composition has met needs, can stop being taken into Point, otherwise with residual error battle array E1And F1Instead of E0And F0Above step is repeated to produce:w2=(w21... w2m)T;v2=(v21... v2p)TAndFor the score vector of second pair of composition;<mrow> <msub> <mi>&alpha;</mi> <mn>2</mn> </msub> <mo>=</mo> <mfrac> <mrow> <msubsup> <mi>E</mi> <mn>1</mn> <mi>T</mi> </msubsup> <mover> <msub> <mi>t</mi> <mn>2</mn> </msub> <mo>^</mo> </mover> </mrow> <mrow> <mo>|</mo> <mo>|</mo> <mover> <msub> <mi>t</mi> <mn>2</mn> </msub> <mo>^</mo> </mover> <mo>|</mo> <msup> <mo>|</mo> <mn>2</mn> </msup> </mrow> </mfrac> <mo>,</mo> <msub> <mi>&beta;</mi> <mn>2</mn> </msub> <mo>=</mo> <mfrac> <mrow> <msubsup> <mi>F</mi> <mn>1</mn> <mi>T</mi> </msubsup> <mover> <msub> <mi>t</mi> <mn>2</mn> </msub> <mo>^</mo> </mover> </mrow> <mrow> <mo>|</mo> <mo>|</mo> <mover> <msub> <mi>t</mi> <mn>2</mn> </msub> <mo>^</mo> </mover> <mo>|</mo> <msup> <mo>|</mo> <mn>2</mn> </msup> </mrow> </mfrac> </mrow>The load of respectively Y, X second pair of composition, at this moment<mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mrow> <msub> <mi>E</mi> <mn>0</mn> </msub> <mo>=</mo> <mover> <msub> <mi>t</mi> <mn>1</mn> </msub> <mo>^</mo> </mover> <msubsup> <mi>&alpha;</mi> <mn>1</mn> <mi>T</mi> </msubsup> <mo>+</mo> <mover> <msub> <mi>t</mi> <mn>2</mn> </msub> <mo>^</mo> </mover> <msubsup> <mi>&alpha;</mi> <mn>2</mn> <mi>T</mi> </msubsup> <mo>+</mo> <msub> <mi>E</mi> <mn>2</mn> </msub> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <msub> <mi>F</mi> <mn>0</mn> </msub> <mo>=</mo> <mover> <msub> <mi>t</mi> <mn>1</mn> </msub> <mo>^</mo> </mover> <msubsup> <mi>&beta;</mi> <mn>1</mn> <mi>T</mi> </msubsup> <mo>+</mo> <mover> <msub> <mi>t</mi> <mn>2</mn> </msub> <mo>^</mo> </mover> <msubsup> <mi>&beta;</mi> <mn>2</mn> <mi>T</mi> </msubsup> <mo>+</mo> <msub> <mi>F</mi> <mn>2</mn> </msub> </mrow> </mtd> </mtr> </mtable> </mfenced>(d) z × m data matrixes E is set0Order be r≤min (z-1, m), then r composition t be present1, t2..., trSo that<mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mrow> <msub> <mi>E</mi> <mn>0</mn> </msub> <mo>=</mo> <mover> <msub> <mi>t</mi> <mn>1</mn> </msub> <mo>^</mo> </mover> <msubsup> <mi>&alpha;</mi> <mn>1</mn> <mi>T</mi> </msubsup> <mo>+</mo> <mo>...</mo> <mo>+</mo> <mover> <msub> <mi>t</mi> <mi>r</mi> </msub> <mo>^</mo> </mover> <msubsup> <mi>&alpha;</mi> <mi>r</mi> <mi>T</mi> </msubsup> <mo>+</mo> <msub> <mi>E</mi> <mi>r</mi> </msub> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <msub> <mi>F</mi> <mn>0</mn> </msub> <mo>=</mo> <mover> <msub> <mi>t</mi> <mn>1</mn> </msub> <mo>^</mo> </mover> <msubsup> <mi>&beta;</mi> <mn>1</mn> <mi>T</mi> </msubsup> <mo>+</mo> <mo>...</mo> <mo>+</mo> <mover> <msub> <mi>t</mi> <mi>r</mi> </msub> <mo>^</mo> </mover> <msubsup> <mi>&beta;</mi> <mi>r</mi> <mi>T</mi> </msubsup> <mo>+</mo> <msub> <mi>F</mi> <mi>r</mi> </msub> </mrow> </mtd> </mtr> </mtable> </mfenced>Substitute into Y=t1β1+…+trβ x, produce p dependent variable PLS equation:yj=aj1x1+…+ajmxm, (j=1,2 ..., p)HereMeetOptionally, step (7) also includes cross validation test:I.e. for the composition number l extracted needed for modeling, pass through intersection Validity check determines:Cast out i-th of observation (i=1,2 ..., n) every time, use remaining n-1 observation an inclined most young waiter in a wineshop or an inn Multiply homing method modeling, and consider to extract the regression equation being fitted after h composition, i-th of the observation station cast out then is substituted into institute The regression equation of fitting, obtains yjThe predicted value y of (j=1,2 ..., p) in i-th of observation station(i)j(h);To i=1, 2 ..., n repeat the checking of the above, produce j-th of dependent variable y when extracting h compositionjThe prediction error of (j=1,2 ..., p) is put down Fang He, it is<mrow> <msub> <mi>PRESS</mi> <mi>j</mi> </msub> <mrow> <mo>(</mo> <mi>h</mi> <mo>)</mo> </mrow> <mo>=</mo> <munderover> <mo>&Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msup> <mrow> <mo>(</mo> <msub> <mi>y</mi> <mrow> <mi>i</mi> <mi>j</mi> </mrow> </msub> <mo>-</mo> <msub> <mover> <mi>y</mi> <mo>^</mo> </mover> <mrow> <mo>(</mo> <mi>i</mi> <mo>)</mo> <mi>j</mi> </mrow> </msub> <mo>(</mo> <mi>h</mi> <mo>)</mo> <mo>)</mo> </mrow> <mn>2</mn> </msup> <mo>,</mo> <mrow> <mo>(</mo> <mi>j</mi> <mo>=</mo> <mn>1</mn> <mo>,</mo> <mn>2</mn> <mo>,</mo> <mo>...</mo> <mo>,</mo> <mi>p</mi> <mo>)</mo> </mrow> </mrow>Y=(y1..., yp)TPrediction sum squares be<mrow> <mi>P</mi> <mi>R</mi> <mi>E</mi> <mi>S</mi> <mi>S</mi> <mrow> <mo>(</mo> <mi>h</mi> <mo>)</mo> </mrow> <mo>=</mo> <munderover> <mo>&Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>p</mi> </munderover> <msub> <mi>PRESS</mi> <mi>j</mi> </msub> <mrow> <mo>(</mo> <mi>h</mi> <mo>)</mo> </mrow> </mrow>In addition, using all sample points again, the regression equation containing h composition is fitted;At this moment, the prediction of i-th of sample point is remembered It is worth and isBy yjError sum of squares be defined as<mrow> <msub> <mi>SS</mi> <mi>j</mi> </msub> <mrow> <mo>(</mo> <mi>h</mi> <mo>)</mo> </mrow> <mo>=</mo> <msubsup> <mi>&Sigma;</mi> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </msubsup> <msup> <mrow> <mo>(</mo> <msub> <mi>y</mi> <mrow> <mi>i</mi> <mi>j</mi> </mrow> </msub> <mo>-</mo> <msub> <mover> <mi>y</mi> <mo>^</mo> </mover> <mrow> <mo>(</mo> <mi>i</mi> <mo>)</mo> <mi>j</mi> </mrow> </msub> <mo>(</mo> <mi>h</mi> <mo>)</mo> <mo>)</mo> </mrow> <mn>2</mn> </msup> </mrow>Define Y error sum of squares be<mrow> <mi>S</mi> <mi>S</mi> <mrow> <mo>(</mo> <mi>h</mi> <mo>)</mo> </mrow> <mo>=</mo> <munderover> <mo>&Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>p</mi> </munderover> <msub> <mi>SS</mi> <mi>j</mi> </msub> <mrow> <mo>(</mo> <mi>h</mi> <mo>)</mo> </mrow> </mrow>When PRESS (h) reaches minimum value, corresponding h is required composition number;Defining Cross gain modulation isBefore each step calculating of modeling terminates, Cross validation test is carried out, if walked in hThen model reaches required precision, stops Extract component;IfRepresent the t of h step extractionshThe contributrion margin of composition is notable, should continue h+1 steps and calculate.
- A kind of 10. crude oil attribute forecast method based near infrared spectrum detection, it is characterised in that methods described includes:(A) atlas of near infrared spectra of crude oil to be detected is determined;(B) 12500~4000cm in the atlas of near infrared spectra obtained to step (A)-1The absorbance for composing area carries out first differential Processing and multiplicative scatter correction;(C) the obtained near-infrared mould for crude oil attribute forecast is built using the method any one of claim 1-9 Type is predicted to the crude oil attribute.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710984193.2A CN107748146A (en) | 2017-10-20 | 2017-10-20 | A kind of crude oil attribute method for quick predicting based near infrared spectrum detection |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710984193.2A CN107748146A (en) | 2017-10-20 | 2017-10-20 | A kind of crude oil attribute method for quick predicting based near infrared spectrum detection |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107748146A true CN107748146A (en) | 2018-03-02 |
Family
ID=61252529
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710984193.2A Pending CN107748146A (en) | 2017-10-20 | 2017-10-20 | A kind of crude oil attribute method for quick predicting based near infrared spectrum detection |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107748146A (en) |
Cited By (22)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108387548A (en) * | 2018-05-24 | 2018-08-10 | 东北农业大学 | A method of sweetener is quickly detected based on infrared spectrum technology |
CN109668858A (en) * | 2019-02-14 | 2019-04-23 | 大连理工大学 | Method based near infrared spectrum detection fermentation process biomass and concentration of component |
CN110514619A (en) * | 2019-08-29 | 2019-11-29 | 山东科技大学 | Near-Infrared Quantitative Analysis model building method based on Biased estimator |
CN110672552A (en) * | 2019-09-25 | 2020-01-10 | 南京林业大学 | Confidence coefficient estimation method for vehicle fuel oil near infrared spectrum detection result |
CN110987862A (en) * | 2019-11-06 | 2020-04-10 | 汉谷云智(武汉)科技有限公司 | Diesel oil on-line blending method |
CN110987866A (en) * | 2019-12-19 | 2020-04-10 | 汉谷云智(武汉)科技有限公司 | Gasoline property evaluation method and device |
CN111413324A (en) * | 2020-05-18 | 2020-07-14 | 南京富岛信息工程有限公司 | Raman spectrum detection method for trace crude oil in naphtha by using fluorescence background |
CN111445964A (en) * | 2020-03-27 | 2020-07-24 | 合肥金星机电科技发展有限公司 | Method for visualizing component analysis results |
CN111829978A (en) * | 2019-04-18 | 2020-10-27 | 中国石油化工股份有限公司 | Method for blending target crude oil from existing crude oil by utilizing near infrared spectrum |
CN112444500A (en) * | 2020-11-11 | 2021-03-05 | 东北大学秦皇岛分校 | Alzheimer's disease intelligent detection device based on spectrum |
CN113094892A (en) * | 2021-04-02 | 2021-07-09 | 辽宁石油化工大学 | Oil concentration prediction method based on data elimination and local partial least squares |
CN113125377A (en) * | 2021-03-30 | 2021-07-16 | 武汉理工大学 | Method and device for detecting diesel oil property based on near infrared spectrum |
CN113158575A (en) * | 2021-04-29 | 2021-07-23 | 晶格码(青岛)智能科技有限公司 | Method for transferring online near-infrared spectrum model of assumed standard sample |
CN113340815A (en) * | 2021-06-01 | 2021-09-03 | 安徽大学 | Crown scale wheat scab disease index evaluation method based on two-waveband combination index |
CN113702328A (en) * | 2021-08-20 | 2021-11-26 | 广东省惠州市石油产品质量监督检验中心 | Method, device, equipment and storage medium for analyzing properties of product oil |
CN113740295A (en) * | 2021-07-29 | 2021-12-03 | 北京易兴元石化科技有限公司 | Near-infrared modeling-based product oil additive detection and analysis method and device |
CN113792082A (en) * | 2021-09-02 | 2021-12-14 | 深圳创景数科信息技术有限公司 | Fabric component retrieval method based on database |
CN113933263A (en) * | 2021-10-11 | 2022-01-14 | 四川启睿克科技有限公司 | Near infrared spectrum model establishing method based on index fitting |
CN114354534A (en) * | 2021-12-30 | 2022-04-15 | 中国航空油料有限责任公司 | Method for establishing aviation kerosene property prediction model by utilizing binary linear classifier |
CN115015126A (en) * | 2022-04-26 | 2022-09-06 | 中国人民解放军国防科技大学 | Method and system for judging activity of powdery biological particle material |
CN116242799A (en) * | 2023-03-14 | 2023-06-09 | 合肥工业大学 | Base oil detection device and detection method based on deep learning infrared multidimensional fusion algorithm |
CN117109906A (en) * | 2023-10-24 | 2023-11-24 | 卡松科技股份有限公司 | Oil online equipment fault analysis method and system based on visualization |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102374975A (en) * | 2010-08-19 | 2012-03-14 | 中国石油化工股份有限公司 | Method for predicting physical property data of oil product by using near infrared spectrum |
CN103729650A (en) * | 2014-01-17 | 2014-04-16 | 华东理工大学 | Selection method for near infrared spectrum modeling samples |
CN105424641A (en) * | 2014-09-04 | 2016-03-23 | 中国石油化工股份有限公司 | Crude oil type near infrared spectrum identification method |
CN105466884A (en) * | 2014-09-04 | 2016-04-06 | 中国石油化工股份有限公司 | Method for identifying type and characteristic of crude oil through near-infrared spectrum |
-
2017
- 2017-10-20 CN CN201710984193.2A patent/CN107748146A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102374975A (en) * | 2010-08-19 | 2012-03-14 | 中国石油化工股份有限公司 | Method for predicting physical property data of oil product by using near infrared spectrum |
CN103729650A (en) * | 2014-01-17 | 2014-04-16 | 华东理工大学 | Selection method for near infrared spectrum modeling samples |
CN105424641A (en) * | 2014-09-04 | 2016-03-23 | 中国石油化工股份有限公司 | Crude oil type near infrared spectrum identification method |
CN105466884A (en) * | 2014-09-04 | 2016-04-06 | 中国石油化工股份有限公司 | Method for identifying type and characteristic of crude oil through near-infrared spectrum |
Non-Patent Citations (3)
Title |
---|
尚廷义 等: "《基于光谱技术的寒地水稻稻瘟病害分析及机理研究》", 30 June 2016, 哈尔滨工程大学出版社 * |
李志刚: "《光谱数据处理与定量分析技术》", 31 January 2017, 北京邮电大学出版社 * |
邱藤 等: "近红外光谱快速原油评价研究", 《光谱学与光谱分析》 * |
Cited By (30)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108387548A (en) * | 2018-05-24 | 2018-08-10 | 东北农业大学 | A method of sweetener is quickly detected based on infrared spectrum technology |
CN109668858A (en) * | 2019-02-14 | 2019-04-23 | 大连理工大学 | Method based near infrared spectrum detection fermentation process biomass and concentration of component |
CN111829978B (en) * | 2019-04-18 | 2023-08-11 | 中国石油化工股份有限公司 | Method for blending target crude oil from existing crude oil by utilizing near infrared spectrum |
CN111829978A (en) * | 2019-04-18 | 2020-10-27 | 中国石油化工股份有限公司 | Method for blending target crude oil from existing crude oil by utilizing near infrared spectrum |
CN110514619A (en) * | 2019-08-29 | 2019-11-29 | 山东科技大学 | Near-Infrared Quantitative Analysis model building method based on Biased estimator |
CN110672552B (en) * | 2019-09-25 | 2022-06-07 | 南京林业大学 | Confidence coefficient estimation method for vehicle fuel oil near infrared spectrum detection result |
CN110672552A (en) * | 2019-09-25 | 2020-01-10 | 南京林业大学 | Confidence coefficient estimation method for vehicle fuel oil near infrared spectrum detection result |
CN110987862A (en) * | 2019-11-06 | 2020-04-10 | 汉谷云智(武汉)科技有限公司 | Diesel oil on-line blending method |
CN110987866A (en) * | 2019-12-19 | 2020-04-10 | 汉谷云智(武汉)科技有限公司 | Gasoline property evaluation method and device |
CN111445964B (en) * | 2020-03-27 | 2023-05-12 | 合肥金星智控科技股份有限公司 | Visualization method of component analysis result |
CN111445964A (en) * | 2020-03-27 | 2020-07-24 | 合肥金星机电科技发展有限公司 | Method for visualizing component analysis results |
CN111413324A (en) * | 2020-05-18 | 2020-07-14 | 南京富岛信息工程有限公司 | Raman spectrum detection method for trace crude oil in naphtha by using fluorescence background |
CN112444500A (en) * | 2020-11-11 | 2021-03-05 | 东北大学秦皇岛分校 | Alzheimer's disease intelligent detection device based on spectrum |
CN113125377A (en) * | 2021-03-30 | 2021-07-16 | 武汉理工大学 | Method and device for detecting diesel oil property based on near infrared spectrum |
CN113125377B (en) * | 2021-03-30 | 2024-02-23 | 武汉理工大学 | Method and device for detecting property of diesel based on near infrared spectrum |
CN113094892A (en) * | 2021-04-02 | 2021-07-09 | 辽宁石油化工大学 | Oil concentration prediction method based on data elimination and local partial least squares |
CN113158575A (en) * | 2021-04-29 | 2021-07-23 | 晶格码(青岛)智能科技有限公司 | Method for transferring online near-infrared spectrum model of assumed standard sample |
CN113340815A (en) * | 2021-06-01 | 2021-09-03 | 安徽大学 | Crown scale wheat scab disease index evaluation method based on two-waveband combination index |
CN113740295A (en) * | 2021-07-29 | 2021-12-03 | 北京易兴元石化科技有限公司 | Near-infrared modeling-based product oil additive detection and analysis method and device |
CN113740295B (en) * | 2021-07-29 | 2024-03-05 | 北京易兴元石化科技有限公司 | Method and device for detecting and analyzing finished oil additive based on near infrared modeling |
CN113702328B (en) * | 2021-08-20 | 2024-05-17 | 广东省惠州市石油产品质量监督检验中心 | Method, device, equipment and storage medium for analyzing properties of finished oil |
CN113702328A (en) * | 2021-08-20 | 2021-11-26 | 广东省惠州市石油产品质量监督检验中心 | Method, device, equipment and storage medium for analyzing properties of product oil |
CN113792082A (en) * | 2021-09-02 | 2021-12-14 | 深圳创景数科信息技术有限公司 | Fabric component retrieval method based on database |
CN113933263A (en) * | 2021-10-11 | 2022-01-14 | 四川启睿克科技有限公司 | Near infrared spectrum model establishing method based on index fitting |
CN114354534A (en) * | 2021-12-30 | 2022-04-15 | 中国航空油料有限责任公司 | Method for establishing aviation kerosene property prediction model by utilizing binary linear classifier |
CN115015126A (en) * | 2022-04-26 | 2022-09-06 | 中国人民解放军国防科技大学 | Method and system for judging activity of powdery biological particle material |
CN116242799B (en) * | 2023-03-14 | 2023-08-18 | 合肥工业大学 | Base oil detection device and detection method based on deep learning infrared multidimensional fusion algorithm |
CN116242799A (en) * | 2023-03-14 | 2023-06-09 | 合肥工业大学 | Base oil detection device and detection method based on deep learning infrared multidimensional fusion algorithm |
CN117109906B (en) * | 2023-10-24 | 2024-01-30 | 卡松科技股份有限公司 | Oil online equipment fault analysis method and system based on visualization |
CN117109906A (en) * | 2023-10-24 | 2023-11-24 | 卡松科技股份有限公司 | Oil online equipment fault analysis method and system based on visualization |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107748146A (en) | A kind of crude oil attribute method for quick predicting based near infrared spectrum detection | |
CN107817223A (en) | The construction method of quick nondestructive real-time estimate oil property model and its application | |
CN105388123B (en) | A kind of method by near infrared spectrum prediction oil property | |
CN105466884B (en) | It is a kind of by near infrared light spectrum discrimination crude oil species and its method for property | |
CN103134767B (en) | Method for liquor quality identification through infrared spectrum revision | |
CN105424641B (en) | A kind of near infrared spectrum recognition methods of crude oil type | |
CN105486655B (en) | The soil organism rapid detection method of model is intelligently identified based on infrared spectroscopy | |
CN102252972B (en) | Near infrared spectrum based detection method for rapid discrimination of oil-tea camellia seed oil real property | |
CN101995389A (en) | Method for fast recognition of crude oil variety through near infrared spectrum | |
CN109324014B (en) | Self-adaptive near-infrared rapid prediction method for crude oil properties | |
CN104990894A (en) | Detection method of gasoline properties based on weighted absorbance and similar samples | |
CN101216419A (en) | Method for quickly detecting yellow wine quality index | |
CN109324013A (en) | A method of it is quickly analyzed using Gaussian process regression model building oil property near-infrared | |
CN102841069B (en) | Method for rapidly identifying types of crude oil by using mid-infrared spectrum | |
CN112098357B (en) | Strawberry sensory quality grade evaluation method based on near infrared spectrum | |
CN108875118B (en) | Method and device for evaluating accuracy of prediction model of silicon content of blast furnace molten iron | |
CN107703097A (en) | Utilize the method and its application of decay total reflection probe and the model of near infrared spectrometer structure fast prediction oil property | |
CN102841070B (en) | Method for identifying types of crude oil by using two-dimensional correlation infrared synchronization spectrum | |
CN107958267A (en) | A kind of oil property Forecasting Methodology represented based on linear | |
CN106383088A (en) | A seed purity rapid nondestructive testing method based on a multispectral imaging technique | |
CN107632010A (en) | A kind of quantitative approach of combination LIBS to steel samples | |
CN106560699A (en) | Combined detection method used for identification of producing area of Wuyi rock tea | |
CN107860743A (en) | Utilize the method and its application of the model of reflective near infrared fibre-optical probe structure fast prediction oil property | |
CN105548027A (en) | Analytical model and method for determining content of tea oil in blend oil based on near infrared spectroscopy | |
CN102937575A (en) | Watermelon sugar degree rapid modeling method based on secondary spectrum recombination |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180302 |
|
RJ01 | Rejection of invention patent application after publication |