CN103323423A - Anti-interference method and system for gas concentration analysis based on laser - Google Patents

Abstract

The invention is suitable for the gas analysis field, and provides an anti-interference method and a system for gas concentration analysis based on laser. The anti-interference method comprises the following steps of A. employing a direct absorption method to take points of a ventilation waveform to carry out straight line fitting, B. subtracting the ventilation waveform from a straight line obtained by the straight line fitting to obtain an absorption peak waveform, and C. carrying out curve fitting to the absorption peak waveform to obtain a smooth absorption peak waveform, and calculating the concentration of a sample gas by using the area or the peak value of the smooth absorption peak as a characteristic value. By means of a curve fitting algorithm, the gas concentration can be directly worked out; because the calculation of the curve fitting algorithm calculation is throughout the whole absorption peak, noise interference can be greatly reduced; and by means of the curve fitting algorithm, influence on calculating concentrations by signal intensity change interference is reduced. By means of standard Berr-Lambert's law I=I0e<-kcl>, the method is suitable for detecting high concentration gases.

Description

Based on anti-interference method and the system of laser to the gas concentration analysis

Technical field

The invention belongs to the gas analysis field, relate in particular to a kind of anti-interference method and system that gas concentration is analyzed based on laser.

Background technology

At present environmental protection is the most important thing of country, and it then is a requisite ring in the environmental protection that gas detects.Because the advantages such as monochromaticity, concentration of energy, tunability, temperature and power stability of semiconductor laser with tunable, making the tunable diode laser absorption spectroscopy technology is that present test gas composition, concentration can reach very high precision.

Present modal technology is the method for quadratic harmonics, disturbs but the method for quadratic harmonics easily is subject to outer signals.Another kind of scheme is the Sampling Integral method, and the sensitivity of Sampling Integral method will be lower than the method for quadratic harmonics, but its Measurement sensibility will be higher than the method for quadratic harmonics, and owing to not needing to use lock-in amplifier, is lower than the method for quadratic harmonics in price.At present based on up-to-date high speed signal treatment technology, the sensitivity that can improve the Sampling Integral method makes the sensitivity of Sampling Integral method reach the level of the method for quadratic harmonics.

Aspect the signal processing of Sampling Integral method, the algorithm of gas concentration is particularly important. and be by calculating the absorption peak peak value or calculating the absorption peak area and come calculating concentration at present mostly.Use absorption peak peak value calculating concentration, owing to only by the numerical evaluation of a point, easily being subjected to noise; Although and use absorption peak area calculating concentration can remedy this defective, interference (such as the dust interference) that can not the erasure signal Strength Changes.

Summary of the invention

The invention reside in provides a kind of anti-interference method of gas concentration being analyzed based on laser, and aim is to solve the problem that the erasure signal Strength Changes is disturbed.

The present invention is achieved in that a kind of anti-interference method of gas concentration being analyzed based on laser, and described anti-interference method may further comprise the steps:

A, the direct absorption process of employing are got the ventilation waveform and are a little carried out fitting a straight line;

B, the straight line that utilizes fitting a straight line to obtain deduct the ventilation waveform spike shape that is absorbed;

C, the absorption peak waveform carried out curve fitting obtains level and smooth absorption peak waveform, the concentration of calculating sample gas as eigenwert with area or the peak value of level and smooth absorption peak.

Further technical scheme of the present invention is: also comprise among the described step C:

C1, the absorption peak waveform carried out curve fitting obtains level and smooth absorption peak waveform;

C2, the area that extracts level and smooth absorption peak or peak value are as eigenwert;

C3, calculate the concentration of sample gas with the eigenwert of level and smooth absorption peak.

Further technical scheme of the present invention is: also comprise between described step C2, the C3:

The noise reduction step is used for the eigenwert of extracting is carried out medium filtering and mean filter carries out noise reduction.

Further technical scheme of the present invention is: in the described fitting a straight line in order to prevent that fitting parameter is subject to noise effect and larger error occurs and take to gather in each waveform multiple spot and carry out match, fitting formula:

Wherein k represents the slope of fitting a straight line, and b represents the intercept of fitting a straight line.

Further technical scheme of the present invention is: described curve is to obtain level and smooth absorption peak waveform to adopt multiple spot to carry out match, the curve formula: $\hat{y}=a_0+a_{1}x+a_2{x}^2+\&CenterDot;\&CenterDot;\&CenterDot;+a_n{x}^n,$ A wherein _iBe target component, X represents to participate in the matrix that the horizontal ordinate of the point of match forms through conversion.

Further technical scheme of the present invention is: the curve of described absorption peak becomes the form of normal distribution curve, and the normal distribution curve match belongs to nonlinear fitting, the normal distribution curve equation is carried out the series expansion conversion be linear equation: $\hat{y}=a_0+a_{1}x+a_2{x}^2+\&CenterDot;\&CenterDot;\&CenterDot;+a_n{x}^n,$ A wherein _iBe target component, X represents to participate in the matrix that the horizontal ordinate of the point of match forms through conversion.

Further technical scheme of the present invention is: described fitting a straight line is a fitting a straight line, and described curve is at least the sextic match.

Another object of the present invention is to provide a kind of jamproof system of gas concentration being analyzed based on laser, described jamproof system comprises:

The fitting a straight line module is used for adopting direct absorption process that the ventilation waveform is got and a little carries out fitting a straight line;

Absorption peak is asked for module, and the straight line that is used for utilizing fitting a straight line to obtain deducts the ventilation waveform spike shape that is absorbed;

Sample gas concentration computing module, being used for the absorption peak waveform carried out curve fitting obtains level and smooth absorption peak waveform, the concentration of calculating sample gas as eigenwert with area or the peak value of level and smooth absorption peak.

Further technical scheme of the present invention is:, described sample gas concentration computing module comprises:

The curve unit obtains level and smooth absorption peak waveform for the absorption peak waveform is carried out curve fitting;

The eigenwert extraction unit is used for extracting the area of level and smooth absorption peak or peak value as eigenwert;

The concentration computing unit is for the concentration of calculating sample gas with the eigenwert of level and smooth absorption peak.

Further technical scheme of the present invention is: described sample gas concentration computing module also comprises:

Noise reduction unit is used for the eigenwert of extracting is carried out medium filtering and mean filter noise reduction.

The invention has the beneficial effects as follows: by adopting curve fitting algorithm, can directly obtain gas concentration, spread all over whole absorption peak because curve fitting algorithm calculates, reduce greatly noise, adopt curve fitting algorithm, reduced the impact that the suffered change in signal strength of calculating concentration is disturbed.Bill's Lambert law I=I of employing standard ₀e ^-kcl, so that this method is applicable to the detection of high concentrations of gas.

Description of drawings

Fig. 1 is the process flow diagram of the anti-interference method of gas concentration being analyzed based on laser that provides of the embodiment of the invention;

Fig. 2 is the means for anti-jamming structural drawing of gas concentration being analyzed based on laser that the embodiment of the invention provides;

Fig. 3 is absorption peak oscillogram in the anti-interference method of gas concentration being analyzed based on laser that provides of the embodiment of the invention;

Fig. 4 is least square method and robustness regretional analysis fitted figure in the anti-interference method of gas concentration being analyzed based on laser that provides of the embodiment of the invention;

Fig. 5 be the embodiment of the invention provide based on the oscillogram of ventilating in the anti-interference method of laser to the gas concentration analysis.

Embodiment

Reference numeral: 10-fitting a straight line module 20-absorption peak is asked for module 30-sample gas concentration computing module 301-curve unit 302-eigenwert extraction unit 303-noise reduction unit 304 concentration computing units

Fig. 1 shows the anti-interference method process flow diagram of gas concentration being analyzed based on laser provided by the invention, and details are as follows:

Among the step S1, in the direct absorption process of employing, normal waveform is triangular wave in the time of logical zero gas, a certain section appearance depression in the middle of the triangular wave in the time of logical sample gas.The ventilation waveform is carrying out fitting a straight line to each waveform as shown in Figure 5, and the result of match obtains a straight line, normally the leading portion of each waveform is got several points in the process of match, and the rear end is got some points and carried out multi-point fitting.Since include noise in the waveform, can not be very few so be used for counting of match, if larger error appears in the very few impact that will be subject to noise of counting of match.The formula of fitting a straight line:

K wherein, b is the target component of requirement, and k represents the slope of fitting a straight line, and b represents the intercept of fitting a straight line.Being used for counting of match can not be very few, owing to comprising noise in the waveform, is used for the very few words fitting parameter k of counting of match, b can be subject to noise effect and larger error occur, carries out match if each waveform has 100 points to get at least six points, and these six points are chosen three at leading portion as far as possible, choose three in the rear end, the result that goes out of chosen_point fitting makes noise effect drop to minimum like this, and its best match selected point is the 17th, 18,19,95,96,97 these six somes matches.

Among the step S2, the waveform when straight line that obtains according to step S1 deducts ventilation will obtain absorption peak waveform as shown in Figure 3, the curved structure of described absorption peak waveform.

Among the step S3, can find out that according to Fig. 3 the absorption peak curve waveform has still comprised noise, for the impact of reduce noise, the absorption peak curve waveform be carried out curve fitting, will obtain a comparatively level and smooth absorption peak waveform through curve.Because the absorption peak curve is the form of normal distribution curve, and the normal distribution curve match belongs to nonlinear fitting, realize that in the C-language Programming Design program nonlinear fitting compares money year, considering that the normal distribution curve equation is carried out series expansion, the nonlinear equation that is is varied to the linear equation of following form:

A wherein _iBe target component, X represents to participate in the matrix that the horizontal ordinate of the point of match forms through serial conversion.Being able to the sextic match through test of many times, that noise is dropped to is minimum.The below fits to the detailed step that example provides this algorithm with sextic:

Fit equation:

$\hat{y}=a_0+a_{1}x+a_2{x}^2+a_3{x}^3+a_4{x}^4+a_5{x}^5+a_6{x}^6,$

Being write as matrix form is:

Y=Xα

Wherein: Y=[y ₁y ₂Y _n] ^T,

α=[a ₀?a ₁?…?a ₆] ^T

If

The expression estimated value, e represents error, that is:

$e_i\left(\mathrm{\&alpha;}\right)=Y-\hat{Y},i=\mathrm{1,2},\&CenterDot;\&CenterDot;\&CenterDot;n,$

Then the objective function of robustness regretional analysis is:

$\mathrm{\&Delta;}\left(\mathrm{\&alpha;}\right)=\min \underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\mathrm{\&rho;}\left(\frac{e_i}{\mathrm{\&gamma;}}\right)$

To Δ (α) about a ₀, a ₁, a ₆Ask local derviation and make that its derivative is 0, namely

$\frac{\&PartialD;\mathrm{\&Delta;}\left(\mathrm{\&alpha;}\right)}{{\&PartialD;a}_{\mathrm{<figure-callout\; id="0"\; label="j"\; filenames="130524090030.png"\; state="\{\{state\}\}">j</figure-callout>}}}0,j=\mathrm{0,1},\&CenterDot;\&CenterDot;\&CenterDot;6$

With the result's arrangement that obtains be: $\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}w\left({\stackrel{\&OverBar;}{e}}_i\right)(y_i-{\hat{y}}_i)X_{\mathrm{ij}}=0,i=1\&CenterDot;\&CenterDot;\&CenterDot;n,j=0\&CenterDot;\&CenterDot;\&CenterDot;6,$ Wherein w (x)=ρ ' (x)/x, be to return the weight of estimation variance, γ=med{|e _i-med (e _i) |/0.6745}

Calculate

Wherein:

$W=\mathrm{diag}\left[\begin{array}{ccc}{w}_1& \&CenterDot;\&CenterDot;\&CenterDot;& w_{K^2}\end{array}\right]$ w _i＝e _i/γ,i=1,...,n

W (x) is some specific functions, and the below provides several w (x) as weight, w (x) difference, and the estimated parameter that finally calculates is difference to some extent also, but difference is all little.

$w\left(x\right)=\left\{\begin{array}{ccc}1& \left|x\right|\&le;t& \\ t/\left|x\right|& \left|x\right|>t,& t=1.345\end{array}\right.$

$w\left(x\right)=\frac{1}{1+{(x/t)}^2}t=2.385$

The iterative algorithm that parameter alpha is estimated is:

(i) utilize least square method to calculate initial parameter α ⁽⁰⁾

(ii) to each iteration t, calculate residual error

And weight

Wherein, i=1 ..., n;

(iii) utilize above-mentioned formula to calculate new argument α ^(t)

(iv) repetition (ii) and (iii) step are until sequence { α ^(t)Satisfy the condition of convergence.

Annotate: in the above-mentioned iterative process, get ${\mathrm{\&alpha;}}^{\left(t\right)}={\left[\begin{array}{cccc}{a}_0^{\left(t\right)}& a_1^{\left(t\right)}& \&CenterDot;\&CenterDot;\&CenterDot;& a_6^{\left(t\right)}\end{array}\right]}^T,t=\mathrm{0,1},...,$ This algorithm is highly resistant in the data noise spot to the impact of match.

Among the step S4, extract the area of level and smooth absorption peak waveform or peak value as eigenwert.100 points are arranged is the process of example explanation area and peak value to the absorption peak waveform, area is absorption peak waveform after level and smooth to be carried out integration obtain the area that this waveform and x axle surround, owing to being discrete type, so integration is exactly all additions of y value that are equivalent to the absorption peak waveform, the area at the peak that namely is absorbed.And the value of first point is assigned to a variable during peak value of absorption peak, then remaining all points compare with this variable one by one, if the value of certain point is larger than this variable, just exchange, otherwise continue relatively to go down, when have a few all complete after this variable maximal value of just having stored absorption peak, the i.e. peak value of absorption peak.

Among the step S5, still there is certain noise in the level and smooth absorption peak waveform that obtains through curve, minimum for noise is dropped to, reduce error, the absorption peak eigenwert of extracting is carried out medium filtering and mean filter carries out further noise reduction process.

Among the step S6, bring the concentration of calculating sample gas into, formula through the absorption peak eigenwert of a series of processing:

I wherein ₀Be the primary light strong voltage, the light intensity voltage of I after by gas absorption, Span are the gas concentrations of range of rectifying an instrument.Above-mentioned formula is according to langbobier law: I=I ₀e ^-kclBoth sides ask natural logarithm to obtain: ln (I)=ln (I ₀)-kcl, k wherein, l is constant, further is deformed into: I wherein ₀The primary light strong voltage, the light intensity voltage of I after by gas absorption, k, l are constants, are subjected to the impact of factors according to gas test in the reality, actual all adopt be:

Wherein Span is the gas concentration of range of rectifying an instrument.

Described fitting a straight line get the quantity of getting that a quantity is less than curve.

What present most of fitting algorithm all adopted is least square fitting, and the deadly defect of least square method is to be subjected in the data impact of noise spot larger, fitness bias occurs.The situation of stochastic distribution based on signal noise, fitting algorithm in the algorithm determines to adopt robustness regretional analysis match, it is the least square method of weighting, this algorithm is assigned to a weight to the error of fitting of each data point, the weight that error of fitting is large is little, the weight that error of fitting is little is just large, adopts the parameter of Newton iteration method digital simulation curve, iterations or fitting precision is set as the condition of iteration stopping.Change the impact that algorithm is highly resistant to the match of noise spot in the data.

The below is with the difference of example explanation robustness regretional analysis and least square fitting.

Given straight-line equation y=10-2x for example, x=1,2, y=-10 in the time of 10, x=10, the value of y changes y=0 into as noise spot in the time of now with x=10, then adopt respectively least square method and robustness regretional analysis match x=[1,2,3,4,5,6,7,8,9,10] and y=[8,6,4,2,0 ,-2,-4 ,-6 ,-8,0], the fitted figure picture as shown in Figure 4, clearly the robustness regretional analysis match of can finding out is not affected by noise spot substantially in the image, and least square method is subjected to the impact of a noise spot to depart from actual data trend.

Fig. 2 shows the jamproof system of gas concentration being analyzed based on laser provided by the invention, and described jamproof system comprises:

Fitting a straight line module 10 is used for adopting direct absorption process that the ventilation waveform is got and a little carries out fitting a straight line;

Absorption peak is asked for module 20, and the straight line that is used for utilizing fitting a straight line to obtain deducts the ventilation waveform spike shape that is absorbed;

Sample gas concentration computing module 30, being used for the absorption peak waveform carried out curve fitting obtains level and smooth absorption peak waveform, the concentration of calculating sample gas as eigenwert with area or the peak value of level and smooth absorption peak.

Described sample gas concentration computing module 30 comprises:

Curve unit 301 obtains level and smooth absorption peak waveform for the absorption peak waveform is carried out curve fitting;

Eigenwert extraction unit 302 is used for extracting the area of level and smooth absorption peak or peak value as eigenwert;

Concentration computing unit 304 is for the concentration of calculating sample gas with the eigenwert of level and smooth absorption peak.

Described sample gas concentration computing module 30 also comprises:

Noise reduction unit 303 is used for the eigenwert of extracting is carried out medium filtering and mean filter noise reduction.

The above only is preferred embodiment of the present invention, not in order to limiting the present invention, all any modifications of doing within the spirit and principles in the present invention, is equal to and replaces and improvement etc., all should be included within protection scope of the present invention.

Claims (10)

1. an anti-interference method of gas concentration being analyzed based on laser is characterized in that, described anti-interference method may further comprise the steps:

A, the direct absorption process of employing are got the ventilation waveform and are a little carried out fitting a straight line;

B, the straight line that utilizes fitting a straight line to obtain deduct the ventilation waveform spike shape that is absorbed;

C, the absorption peak waveform carried out curve fitting obtains level and smooth absorption peak waveform, the concentration of calculating sample gas as eigenwert with area or the peak value of level and smooth absorption peak.

2. anti-interference method according to claim 1 is characterized in that, also comprises among the described step C:

C1, the absorption peak waveform carried out curve fitting obtains level and smooth absorption peak waveform;

C2, the area that extracts level and smooth absorption peak or peak value are as eigenwert;

C3, calculate the concentration of sample gas with the eigenwert of level and smooth absorption peak.

3. anti-interference method according to claim 2 is characterized in that, also comprises between described step C2, the C3:

The noise reduction step is used for the eigenwert of extracting is carried out medium filtering and mean filter carries out noise reduction.

4. anti-interference method according to claim 3 is characterized in that: in the described fitting a straight line in order to prevent that fitting parameter is subject to noise effect and larger error occurs and take to gather in each waveform multiple spot and carry out match, fitting formula: Wherein k represents the slope of fitting a straight line, and b represents the intercept of fitting a straight line.

5. anti-interference method according to claim 4 is characterized in that, described curve is to obtain level and smooth absorption peak waveform to adopt multiple spot to carry out match, the curve formula:

A wherein _iBe target component, X represents to participate in the matrix that the horizontal ordinate of the point of match forms through conversion.

6. anti-interference method according to claim 5, it is characterized in that: the curve of described absorption peak becomes the form of normal distribution curve, and the normal distribution curve match belongs to nonlinear fitting, the normal distribution curve equation is carried out the series expansion conversion be linear equation:

A wherein _iBe target component, X represents to participate in the matrix that the horizontal ordinate of the point of match forms through conversion.

7. each described anti-interference method according to claim 1-6, it is characterized in that: described fitting a straight line is a fitting a straight line, and described curve is at least the sextic match.

8. jamproof system of gas concentration being analyzed based on laser, it is characterized in that: described jamproof system comprises:

The fitting a straight line module is used for adopting direct absorption process that the ventilation waveform is got and a little carries out fitting a straight line;

Absorption peak is asked for module, and the straight line that is used for utilizing fitting a straight line to obtain deducts the ventilation waveform spike shape that is absorbed;

Sample gas concentration computing module, being used for the absorption peak waveform carried out curve fitting obtains level and smooth absorption peak waveform, the concentration of calculating sample gas as eigenwert with area or the peak value of level and smooth absorption peak.

9. jamproof system according to claim 8 is characterized in that, described sample gas concentration computing module comprises:

The curve unit obtains level and smooth absorption peak waveform for the absorption peak waveform is carried out curve fitting;

The eigenwert extraction unit is used for extracting the area of level and smooth absorption peak or peak value as eigenwert;

The concentration computing unit is for the concentration of calculating sample gas with the eigenwert of level and smooth absorption peak.

10. jamproof system according to claim 9 is characterized in that, described sample gas concentration computing module also comprises:

Noise reduction unit is used for the eigenwert of extracting is carried out medium filtering and mean filter noise reduction.

Images