CN101498667B - Method for detecting ethylene or ethylene propylene rubber content in ethylene-propylene copolymerization polypropylene - Google Patents

Abstract

The invention discloses a method for detecting the content of ethylene or ethylene propylene rubber in ethylene-propylene polypropylene, comprising the steps of obtaining the raman spectrum of an ethylene-propylene polypropylene sample containing known ethylene or ethylene propylene rubber, preprocessing the raman spectrum, conducting correlation modeling through presetting the content of the known ethylene or ethylene propylene rubber and the intensity of the raman spectrum which corresponds to the raman spectrum characteristic wave number set (RCS) so as to obtain a predicted model, detecting the raman spectrum of the ethylene-propylene polypropylene sample to be detected and preprocessing the raman spectrum, guiding the intensity of the raman spectrum of the ethylene-propylene polypropylene sample to be detected, which corresponds to the RCS into the predicted model so as to detect the content of the ethylene or ethylene propylene rubber in the ethylene-propylene polypropylene sample to be detected, and finally reconstructing the RCS through selecting the characteristic of the raman spectrum in the light of the actual situation. The invention has the advantages of convenience, rapidness, accuracy, environmental protection, and the like and can be used for detecting the content of the ethylene or ethylene propylene rubber in ethylene-propylene polypropylene on line or off line during the manufacturing and the production processes of the ethylene-propylene polypropylene.

Description

Ethene or EP rubbers content detecting method in second third COPP

Technical field

The present invention relates to the Raman spectrum detection range, relate in particular to a kind of Raman spectrum ethene and EP rubbers content detecting method in second third COPP.

Background technology

Polypropylene (Polypropylene, be called for short PP) is to be the polymkeric substance that monomer polymerization makes with the propylene, is an important kind in the general-purpose plastics.But the impact flexibility of HOPP is lower, especially the fragility under the low temperature has limited its use greatly, therefore people carry out modification by the method for adding toughness material blend or copolymerization to it, improve polyacrylic performance effectively, make polyacrylic range of application more extensive.

In second third COPP, the adding of ethene can be introduced EP rubbers, improves the shock resistance of product.And in the multipolymer content of the content of ethene and EP rubbers to the mechanical property important influence of product, for example along with the increase of ethylene contents, polypropene impact-resistant intensity increases, and pulling strengrth and flexural strength reduce, increase along with EP rubbers content, polypropylene normal temperature erosion-resisting characteristics and low-temperature impact resistance increase, and pulling strengrth and bending modulus reduce.Therefore one of production firm's key index that ethylene contents and EP rubbers content are produced as polypropylene is used for polyacrylic production and production quality control.

The method that is used to measure ethylene contents and EP rubbers content at present mainly comprises chemical analysis, radioisotope method, homopolymer blend method, model compound method, chemical extraction method (ISO 16152-2005), (industrial solid nuclear-magnetism Magmonitor is used for the foundation of polyolefin mass parameter method of testing to the nmr analysis method, petrochemical complex, 2002, the 31st the 1st phase of volume, 48～52) and infrared analysis (CN 1124483C, CN 1152245C) etc.Wherein chemical analysis, radioisotope method and model compound method complex operation, take time and effort, can't be applied to polyacrylic online detection; The homopolymer blend method is because seldom use of measuring error; Chemical extraction method is not only consuming time longer, and the organic solvent that uses in the extraction process also may bring environmental pollution; The required equipment volume of nmr analysis method is big, acquisition expenses is high, simultaneously nuclear-magnetism equipment to operating conditions require high, plant maintenance is complicated, brings inconvenience for the device operating personnel; The sample detection of infrared analysis needs the matrix sample preparation, also can't realize online detection.

Summary of the invention

The invention provides accurate, quick, safe, the harmless detection method of ethylene contents and EP rubbers content in a kind of second third COPP, solved the problems that prior art exists effectively; The present invention is suitable for and online detection and offline inspection simultaneously, can provide effective means for the production of second third COPP and the real-time monitoring of process.

Detection method proposed by the invention comprises the steps:

(1) obtain known ethylene contents or EP rubbers content sample second third COPP Raman spectrum and carry out pre-service;

(2) preset features wave number collection with the pairing raman spectrum strength of characteristic wave manifold of sample second third COPP and known ethylene contents or the related modeling of EP rubbers content, obtains forecast model:

Y＝I _RCS*coeff+K

Wherein Y is ethylene contents or EP rubbers content, I _RCSBe characteristic wave manifold characteristic of correspondence raman spectrum strength collection, coeff is a matrix of coefficients, and K is a residual error;

(3) obtain the Raman spectrum of second third COPP to be measured, after the pre-service, extract and the pairing raman spectrum strength of characteristic wave manifold, the substitution forecast model obtains the ethylene contents or the EP rubbers content of second third COPP to be measured.

Detection method proposed by the invention, when obtaining Raman spectrum by Raman spectrometer, the resolution that Raman spectrometer is set is better than 10cm ^-1

The spectrum that Raman spectrometer is gathered has also comprised other irrelevant informations and noise except that the self information of sample, as electric noise, sample background and parasitic light etc.Therefore, the preprocess method of elimination spectroscopic data irrelevant information and noise becomes very crucial and necessary.Spectrogram preprocess method commonly used comprises smoothly, reduces, differential, normalization, standardization, polynary scatter correction, the polynary scatter correction of segmentation, standard normal variable conversion, remove trend, Fourier transform, orthogonal signal correction, wavelet transformation and analytic signal etc. only, concrete grammar can be referring to " spectrum pre-service and Wavelength selecting method progress and application in the infrared analysis " (chemical progress, 2004, the 16th the 4th phase of volume, 528～542).

The detection method that proposes according to the present invention comprises 805,813,1325,1333,1453,1463,1465,1467cm among the characteristic wave manifold RCS that ethylene contents detects ^-1, comprise 799,805,811,1335,1467,1471 among the characteristic wave manifold RCS of EP rubbers content detection, 1477cm ^-1

Use mathematical model with the pairing raman spectrum strength I of feature Raman wave number collection RCS _RCSCarry out relatedly with ethylene contents or EP rubbers content, described mathematical model comprises multiple linear regression, principal component regression, partial least square method etc., and concrete grammar can be referring to " multivariate data processing " (Chemical Industry Press, 1998).

Because the difference of Raman spectroscopy instrument, the above-mentioned characteristic wave manifold that presets may need to adjust and optimize, such as the feature selecting by Raman spectrum the characteristic wave manifold is optimized, the wave number number that is comprised among the characteristic wave manifold RCS also can increase or reduce according to the actual requirements simultaneously.The detection method that proposes according to the present invention, the wave number number that is comprised among the characteristic wave manifold RCS is not less than 5.Described feature selecting comprises after the system of selection of sequence forward direction, the sequence to system of selection, increase that l goes the r system of selection, searching method etc. floats, concrete grammar can be referring to " based on the research and the application of the feature selection approach of support vector machine " (Zhejiang University's PhD dissertation, 2006).

As further extension, detection method proposed by the invention can also be used for the detection of melt polypropylene flow rate, isotacticity, xylene soluble part content etc.

The present invention is compared with prior art easy and simple to handle, and detection speed is fast; The testing result standard, sensing range is wide; Can be implemented in line detects; The output signal of pick-up unit can directly be sent into control system, to realize the real-time monitoring of polypropylene production and process.

Description of drawings

Fig. 1 is the process flow diagram of the inventive method;

Fig. 2 is fitting coefficient distribution plan when ethylene contents is predicted among the embodiment 1;

Fig. 3 is the embodiment 1 ethylene contents synoptic diagram that predicts the outcome;

Fig. 4 is fitting coefficient distribution plan when ethylene contents is predicted among the embodiment 1;

Fig. 5 is embodiment 1 an EP rubbers content prediction result schematic diagram;

Fig. 6 is the ethylene contents synoptic diagram that predicts the outcome among the embodiment 2;

Fig. 7 (a) is an EP rubbers content prediction result schematic diagram among the embodiment 3;

Fig. 7 (b) is the ethylene contents synoptic diagram that predicts the outcome among the embodiment 3;

Fig. 8 increases l to remove r method flow synoptic diagram;

Fig. 9 is the searching method schematic flow sheet that floats;

Figure 10 is horizontal type agitated bed structural representation;

Figure 11 is the Changing Pattern figure of ethylene contents in the polypropylene of horizontal type agitated bed exit.

Embodiment

Embodiment 1

The MultiRAM type Fourier Raman spectrometer that adopts German Bruker company to produce, it is 1064nm that this spectrometer is selected laser source wavelength for use.Set spectrometer spectral scan scope 0～3600cm ^-1, resolution is 1cm ^-1By non-immersion measurement, detect the Raman spectrum of some sample second third COPPs, prediction ethylene contents and EP rubbers content.The ethylene contents of known sample and EP rubbers content are as shown in table 1:

Table 1 polypropylene specimen character

Sample number into spectrum	MFR?	Ethylene contents	EP rubbers content
				1?	1.5?	7.5?	12.5?
2?	2.5?	10.5?	16.1?
				3?	10.4?	6.8?	13.4?
4?	9.0?	9.9?	20.0?
				5?	8.8?	13.5?	22.5?
6?	2.5?	9.5?	18.3?
				7?	1.6?	7.1?	11.9?
8?	2.7?	6.2?	8.8?
				9?	9.7?	14.4?	27.3?
10?	8.2?	15.3?	29.1?

At first, the Raman spectrum that sample is measured carries out pre-service by polynary scatter correction, and its detailed process is as follows:

(1) averaged spectrum of the required correction spectrum of calculating:

$\stackrel{\_}{A}=\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{A}_i}{n}$

In the formula, A _iBe the Raman spectrum matrix of i sample,

Be the averaged spectrum matrix.

(2) averaged spectrum is returned:

$A_i=a_i\stackrel{\_}{A}+b_i$

In the formula, a _i, b _iPolynary scatter correction fitting coefficient for the Raman spectrum matrix.

(3) each bar spectrum is made polynary scatter correction:

$A_i^{\text{'}}=\frac{(A_i-b_i)}{a_i}$

In the formula, A _i ^'Raman spectrum matrix for i sample behind polynary scatter correction.

Detect preset features wave number collection RCS according to ethylene contents ₁, RCS ₁In comprise 805,813,1325,1333,1453,1463,1465,1467cm ^-1, and with RCS ₁Pairing feature raman spectrum strength collection I _RCS1Ethylene contents Y with sample second third COPP _ECCarry out modeling by multiple linear regression, then have

Y _FC＝I _RCS1×coeff ₁+K ₁

Coeff wherein ₁Be matrix of coefficients (can with reference to figure 2), K ₁Be residual error.

In this example, (promptly choose 1 sample by normalization method as forecast sample at every turn, carry out regression modeling with remaining sample as known sample, and selected forecast sample predicted) ethylene contents in second third COPP is predicted, the result as shown in Figure 3, its average relative error is 3.1%.

According to EP rubbers content detection preset features wave number collection RCS ₂, RCS ₂In comprise 799,805,811,1335,1467,1471,1477cm ^-1And with RCS ₂Pairing feature raman spectrum strength collection I _RCS2EP rubbers content Y with second third COPP _RCarry out modeling by multiple linear regression, then have

Y _R＝I _RCS2×coeff ₂+K ₂

Coeff wherein ₂Be matrix of coefficients (can with reference to figure 4), K ₂Be residual error.

In this example, EP rubbers content in second third COPP is predicted by normalization method, the result as shown in Figure 5, its average relative error is 3.2%.

If adopt the polynary scatter correction method of segmentation to carry out pre-service, wherein window number is set to 10, and the measuring error of gained ethylene contents and EP rubbers content decreases, and its average relative error is respectively 2.9% and 3.0%.

Use other preprocess method gained result as shown in table 2.

The different preprocess methods of table 2 compare the influence of final detection result

?

Smoothly

Differential

Normalization

Standardization

Go trend

Fourier transform

Wavelet transformation

Ethylene contents

3.9％?

4.2％?

4.3％?

3.6％?

3.3％?

3.1％?

2.9％?

EP rubbers content

4.1％?

4.3％?

3.7％?

3.2％?

3.4％?

3.2％?

3.1％?

Embodiment 2

The MultiRAM type Fourier Raman spectrometer that adopts German Bruker company to produce, it is 1064nm that this spectrometer is selected laser source wavelength for use.Set spectrometer spectral scan scope 0～3600cm ^-1, resolution is 10cm ^-1By non-immersion measurement, detect the Raman spectrum of the second third COPP sample, the prediction ethylene contents.

At first, the pairing raman spectrum strength collection of the preset features wave number collection RCS I that detects according to ethylene contents _RCSWith polyacrylic ethylene contents Y _ECCarry out modeling by multiple linear regression, detailed process such as embodiment 1 obtain predicting the outcome of ethylene contents in second third COPP, as shown in Figure 6.

Then, RCS is optimized to preset features wave number collection, by the system of selection of sequence forward direction the relevant pairing feature Raman of the composition wave number collection of ethene in second third COPP is extracted, and then realize the reconstruct of characteristic wave manifold RCS, its matlab detailed process is as follows:

If (n * m) is pretreated Raman spectrum matrix to X, and wherein n is the sample number, and m is a Raman wave number number, and Y (n * 1) is the ethylene contents of each sample, then new feature Raman wave number collection RCS _NewLeaching process as follows:

New=X; % inserts pretreated X among the new variables new

RCSN=RCS; % with preset features wave number collection RCS as initial value

for?i＝1:length(RCS)

k＝length(RCS)+1-i；

new(:，k)＝□；

End% rejects the respective items of RCS from new

Below begin to carry out in characteristic wave manifold RCS, to add the feature wave number, obtain RCS _New

For i=1:k%k several numbers of characteristic wave for adding

for?j＝1:length(new)

temp＝[RCSN?new(:，j)]；

B＝inv(temp′*temp)*temp′*Y；

E＝Y-temp*B；

E (j)=norm (E); % with the mould of residual matrix E as criterion function

end

for?m＝1:length(new)

iferror(m)＝min(error)

Add=m; % obtain feature wave number new to be added (:, m)

end

end

RCSN＝[RCSN?new(:，m)]；

new(:，m)＝□；

end

So far, obtain new feature Raman wave number collection RCS _New, it is stored as RCSN in matlab.In this example, getting the k value is 2, promptly adds 2 feature wave numbers in preset features wave number collection RCS.

According to new preset features wave number collection RCS _NewPairing raman spectrum strength collection I _RCSNWith ethylene contents Y _ECCarry out modeling by multiple linear regression, detailed process such as embodiment 1 obtain predicting the outcome of ethylene contents, as shown in Figure 6.

As can be seen from Figure 6, helpful by optimizing characteristic wave manifold RCS to reducing the detection error, but it should be noted that, because the restructuring procedure of RCS depends on the quantity of sample, if increase several numbers of characteristic wave blindly, can cause the appearance of over-fitting, increase on the contrary and detect error.

Embodiment 3

The MultiRAM type Fourier Raman spectrometer that adopts German Bruker company to produce, it is 1064nm that this spectrometer is selected laser source wavelength for use.Set spectrometer spectral scan scope 0～3600cm ^-1, resolution is 4cm ^-1By non-immersion measurement, detect the Raman spectrum of the second third COPP sample, prediction ethylene contents and EP rubbers content.

With the example that is predicted as of EP rubbers content, its process is as follows:

At first, according to the pairing raman spectrum strength collection of the preset features wave number collection RCS I of EP rubbers content detection _RCSWith known EP rubbers content Y _RCarry out modeling by multiple linear regression, detailed process such as embodiment 1 obtain predicting the outcome of EP rubbers content, and its mean relative deviation is shown in Fig. 7 (a).

Then, RCS is optimized to preset features wave number collection, by system of selection before and after the sequence the relevant pairing feature Raman of the composition wave number collection of EP rubbers is extracted, and then realizes the reconstruct of characteristic wave manifold RCS, and its matlab detailed process is as follows:

If (n * m) is pretreated Raman spectrum matrix to X, and wherein n is the sample number, and m is a Raman wave number number, and Y (n * 1) is the ethylene contents of each sample, then new feature Raman wave number collection RCS _NewLeaching process as follows:

RCSN=RCS; % with preset features wave number collection RCS as initial value

Below begin to carry out from characteristic wave manifold RCS, to reject the feature wave number, obtain RCS _New

for?i＝1:k

for?j＝1:length(RCSN)

temp＝RCSN；

temp(:，j)＝□；

B＝inv(temp′*temp)*temp′*Y；

E＝Y-temp*B；

E (j)=norm (E); % with the mould of residual matrix E as criterion function

end

for?m＝1:length(new)

if?error(m)＝min(error)

Add=m; % obtain feature wave number RSCN to be rejected (:, m)

end

end

RSCN(:，m)＝□；

end

So far, obtain new feature Raman wave number collection RCS _New, it is stored as RCSN in matlab.In this example, get the k value and get 1,2,3 successively ..., 6, i.e. characteristic wave manifold RCS _NewRemaining feature wave number is successively decreased successively by 6 to 1.According to new preset features wave number collection RCS _NewPairing raman spectrum strength collection I _RCSNWith EP rubbers content Y _RCarry out modeling by multiple linear regression, detailed process such as embodiment 1 obtain predicting the outcome of EP rubbers content, and its mean relative deviation is shown in Fig. 7 (a).

The testing process of the detection of ethylene contents and EP rubbers content is similar, and after system of selection was rejected the characteristic wave manifold before and after the employing sequence, the testing result mean relative deviation of ethylene contents was shown in Fig. 7 (b).

As can be seen, can exert an influence by the detection of rejecting Partial Feature wave number from preset features wave number collection RCS to polypropylene character from Fig. 7 (a) and Fig. 7 (b), the feature wave number is rejected many more, and it is big more to detect error.

Embodiment 4

The DXR Raman spectrometer that adopts U.S. ThermoFisher company to produce, it is 780nm that this spectrometer is selected laser source wavelength for use, sets spectrometer spectral scan scope 0～3600cm ^-1, resolution is 4cm ^-1By the Raman spectrum of the industrial Raman fiber probe off-line measurement second third COPP sample, prediction ethylene contents and EP rubbers content.

To directly use preset features wave number collection RCS modeling and forecasting (embodiment 1) and increase l and remove r system of selection and unsteady searching method (increasing l goes the program process of r system of selection and unsteady searching method respectively as Fig. 8 and shown in Figure 9; But specific procedure reference example 2,3) compare, the result is as shown in table 3.

The average relative error contrast as a result of table 3 polypropylene property prediction

?	Ethylene contents (%)	EP rubbers content (%)
			Preset features wave number collection	3.1?	3.2?
Increase l and go r system of selection optimization	2.7?	2.7?
			Unsteady searching method optimization	2.6?	2.7?

Embodiment 5

The cold mould in laboratory is tested with horizontal type agitated bed, as shown in figure 10, during operation, speed of agitator 20rpm, and by regulating terminal valve maintenance bed material level basicly stable (polypropylene homopolymer PP-1 feed rate 1.35kg/min).During original state, have PP-1 to amount to 30kg in the still, using ethylene contents is that 12.5% COPP PP-2 amounts to 1.35kg as tracer agent, studies the horizontal type agitated bed time dependent rule of exit ethylene contents.

Detecting instrument is the DXR Raman spectrometer that U.S. ThermoFisher company produces, and selecting laser source wavelength for use is 780nm, sets spectrometer spectral scan scope 0～3600cm ^-1, resolution is 5cm ^-1Check point is located at horizontal type agitated bed exit, detects in real time by industrial Raman fiber probe, and with embodiment 1 gained model prediction ethylene contents.

By using Raman spectrum to detect the horizontal type agitated bed time dependent situation of exit ethylene contents that ethylene concentration obtains as shown in figure 11, its result and sample detecting gained be basically identical as a result.

Claims (5)

1. ethene or EP rubbers content detecting method in second third COPP comprise the steps:

(1) obtain known ethylene contents or EP rubbers content the second third COPP sample Raman spectrum and carry out pre-service;

(2) preset features wave number collection, feature selecting by Raman spectrum is optimized the characteristic wave manifold, with the pairing raman spectrum strength of characteristic wave manifold of the second third COPP sample and known ethylene contents or the related modeling of EP rubbers content, obtain forecast model:

Y＝I _RCS*coeff+K

Wherein Y is ethylene contents or EP rubbers content, I _RCSBe characteristic wave manifold characteristic of correspondence raman spectrum strength collection, coeff is a matrix of coefficients, and K is a residual error;

Described feature selecting adopts after the system of selection of sequence forward direction, the sequence to system of selection, increase l removes r system of selection or unsteady searching method.

(3) obtain the Raman spectrum of second third COPP to be measured, after the pre-service, extract and the pairing raman spectrum strength of characteristic wave manifold, the substitution forecast model obtains the ethylene contents or the EP rubbers content of second third COPP to be measured.

2. detection method as claimed in claim 1 is characterized in that: several numbers of the characteristic wave in the characteristic wave manifold are no less than 5.

3. detection method as claimed in claim 1 is characterized in that: the characteristic wave manifold that ethylene contents detects is 805,813,1325,1333,1453,1463,1465,1467cm ^-1

4. detection method as claimed in claim 1 is characterized in that: the characteristic wave manifold of EP rubbers content detection is 799,805,811,1335,1467,1471,1477cm ^-1

5. detection method as claimed in claim 1 is characterized in that: the resolution that obtains the Raman spectrum employing is better than 10cm ^-1

Images