A multi-component gas concentration detection method based on spectral analysis algorithm

Abstract

The present invention discloses a multi-component gas concentration detection method based on a spectral analysis algorithm, the steps are as follows: A. changing the wavelength of the incident light and recording the ring-down time τ of the mixed gas t And draw the ring-down curve; B. Normalize the ring-down curve; C. Use the ring-down time τ of the mixed gas t Calculate the normalized ring-down intensity I t ; D. Solve #imgabs0# and plot the mixed spectrum with wave number v as the horizontal coordinate and #imgabs1# as the vertical coordinate; E. Use the spectrum analysis algorithm to perform peak separation on the mixed spectrum to obtain multiple Lorentz distribution function absorption peaks; F. Perform peak area integral calculation on each Lorentz distribution function absorption peak to obtain the peak integral area; G. Substitute the peak integral area and line intensity, as well as the total gas pressure and effective optical path into the gas concentration calculation formula to calculate and obtain the gas concentration of each component in the mixed gas. The present invention uses the spectrum analysis algorithm to perform peak separation and fitting on the mixed spectrum to solve the content of each gas component, and the detection limit is low and the accuracy is high.

Description

Technical Field

[0001] The invention belongs to the technical field of multi-component gas concentration detection, and specifically is a multi-component gas concentration detection method based on a spectrum analysis algorithm. Background Art

[0002] Aiming at different application scenarios, the research on multi-component gas detection technology is constantly improving. Multi-gas detection technology does save a lot of manpower and material resources, and has the advantages of fast speed, convenience and high efficiency.

[0003] In the process of measuring the spectrum, since the absorption peaks of the detected gases are close to each other and the absorption peaks of the gases have tailing phenomena to varying degrees, it brings great obstacles to the detection of gas components.

[0004] Among the existing multi-gas component detection technologies, the tube detection method has low precision and is greatly affected by environmental interference. Although the gas chromatography-mass spectrometry method can detect a wide range of gas components and has high sensitivity, the chromatographic injection time is long, the separation effect is easily affected by the environment, and the price is relatively high. Although the electrochemical sensors and gas sensors have a fast response speed, they have poor selectivity and have disadvantages such as temperature drift and life attenuation. Summary of the invention

[0005] The purpose of the present invention is to provide a multi-component gas concentration detection method based on a spectral analysis algorithm with low detection limit, no secondary calibration and high accuracy in view of the problems existing in the prior art; it is applicable to the field of gas component monitoring using cavity ring-down spectroscopy (CRDS) technology. Since the absorption peaks of multiple gases detected are close, the accuracy of gas measurement will be affected. Therefore, an analysis algorithm based on Lorentz multi-peak fitting is adopted to perform peak separation on the mixed spectrum and then use the peak integrated area of ​​the measured gas to invert the gas content.

[0006] The purpose of the present invention is to be solved by the following technical solutions:

[0007] A multi-component gas concentration detection method based on a spectral analysis algorithm, the steps of the detection method are as follows:

[0008] A. Change the wavelength of the incident light and record the ring-down time τ of the mixed gas t And draw the ring-down curve;

[0009] B. Normalize the ring-down curve;

[0010] C. Using the ring-down time τ of the mixed gas t Calculate the normalized ring-down intensity I t ;

[0011] D. Solve ln The wave number v is the horizontal axis. Plot the mixed spectrum for the ordinate;

[0012] E. Using a spectrum analysis algorithm to perform peak separation on the mixed spectrum to obtain multiple Lorentz distribution function absorption peaks;

[0013] F. Performing peak area integral calculation on each Lorentzian distribution function absorption peak to obtain the peak integrated area;

[0014] G. Substitute the integrated area and line intensity of each peak, as well as the total gas pressure and effective optical path into the gas concentration calculation formula to obtain the gas concentration of each component in the mixed gas.

[0015] The gas concentration calculation formula in step G is:

[0016]

[0017] In formula (6), X represents the concentration of a gas component; It is the peak integral area of ​​a gas component obtained by peak separation using the spectrum analysis algorithm. Represents the incident light intensity I0 and the ring-down light intensity I after transmission t ratio; S(T) represents the line intensity of a gas component; p refers to the total gas pressure; L refers to the effective optical path.

[0018] The ring-down light intensity I in step C t The solution formula is:

[0019]

[0020] In formula (1), I t It represents the optical signal intensity of the mixed gas under the same cavity ring-down time τ0; τ0 is the cavity ring-down time; τ t is the ring-down time of the mixed gas.

[0021] When the ring-down light intensity I t Based on the solution formula, the gas concentration calculation formula in step G is:

[0022]

[0023] The process of the spectrum analysis algorithm in step E is as follows: first determine the absorption fitting function containing the undetermined Lorentz characteristic parameters, use the absorption fitting function to construct the error function, determine the initial value coordinates of the iteration, determine the iteration direction, and then determine the iteration step size according to the Armijo exact linear search method, calculate the gradient and verify whether it meets the requirements, if so, continue, otherwise change the iteration direction and iteration step size at the same time until the minimum fitting error is found and the Lorentz characteristic parameter value in this case is determined, thereby determining the absorption fitting function of the gas to be measured and decomposing the mixed spectrum into multiple absorption peaks.

[0024] The specific steps of the spectrum analysis algorithm in step E are as follows:

[0025] E1. Determine the absorption fitting function containing the unknown Lorentz characteristic parameters:

[0026]

[0027] In formula (2), x is the wave number of the horizontal coordinate in the mixed spectrum, F is the absorbance of the vertical coordinate in the mixed spectrum, m represents the type of gas to be measured, and A i 、w i , is the characteristic parameter of the Lorentz line shape, where A i represents the absorption peak area of ​​the i-th gas component, w i represents the half-height width of the i-th gas component, Indicates the center position of the absorption peak of the i-th gas component;

[0028] E2. Use the absorption fitting function to construct the error function:

[0029]

[0030] In formula (3), Q is the fitting residual, y j represents the absorbance of the mixed spectrum, F(x j ) is the value of the absorption fitting function, and n is the number of spectral sampling points;

[0031] E3, set the iteration precision ε and the initial coordinates of the iteration parameters;

[0032] E4. Determine the search direction k :

[0033]

[0034] In formula (4), d k is the search direction for the kth iteration; d k-1 is the search direction for the k-1th iteration; β k-1 is the conjugate parameter of the k-1th iteration; g k is the kth iteration gradient direction; Transpose the gradient direction for the kth iteration; g is the gradient direction transposition for the k-1th iteration; k-1 is the kth iteration gradient direction;

[0035] E5. Determine the search direction k Then, the Armijo exact linear search method is used to determine the iteration step size α k ;

[0036] E6. Set the initial value of the current iteration (Aik ,w ik )(i is the type of gas, k is the number of iterations), let A k+1 =A k +α k ·d k 、w k+1 =w k +α k ·d k , calculate along the search direction d k and the iteration step α k The next point coordinates (A i(k+1) ,w i(k+1) ) and calculate the coordinates of the next point (A i(k+1) ,w i(k+1) )’s gradient value g k :

[0037]

[0038] In formula (5), is the gradient of the fitted residual matrix;

[0039] E7. Calculate the second norm || g k || and compare with the iteration accuracy ε set in step E3;

[0040] E8, if || g k ||<ε, then stop the operation and output the Lorentz characteristic parameter A k 、w k As the optimal solution, go to step E9; if || g k ||>ε, then return to step E3 and use the coordinates of the iteration parameters of this round as the initial coordinates of the iteration parameters in step E3, and repeat steps E4-E7;

[0041] E9, stop iteration and find the minimum fitting residual Q min , and simultaneously output the corresponding Lorentz characteristic parameter value A ik 、w ik ;

[0042] E10. Determine the absorption fitting function containing the determined Lorentz characteristic parameters of the gas to be measured, and decompose the mixed spectrum into multiple absorption peaks.

[0043] The purpose of the above step E8 is to ensure that the fitting residual Q value drops to the minimum value in the maximum descent direction, that is, to select the fastest descent direction and drop to the lowest point in this direction, thereby quickly finding the minimum fitting residual Q value.

[0044] The search direction d in step E4 k The initial search direction is the negative gradient direction d0 = -g0 of the initial coordinates of the error function.

[0045] The Armijo exact linear search method is used in step E5 to determine the iteration step length α k The detailed steps are:

[0046] E51. Determine the initial parameters δ and σ of the Armijo exact linear search method;

[0047] E52, start cyclic search from m=0;

[0048] E53, if satisfied Then m=m k The iteration step length is

[0049] E54. If not satisfied, set m=m+1 and continue searching until Until establishment.

[0050] The multi-component gas concentration detection method can be applied to the detection of various trace gas concentrations under the background of SF6 / N2 in the CRDS field.

[0051] Compared with the prior art, the present invention has the following advantages:

[0052] The multi-component gas concentration detection method of the present invention adopts a scanning spectrum method to calculate the gas concentration. First, the wavelength is changed to collect the corresponding decay time, and the absorbance is solved using the normalized decay curve and used as the vertical coordinate of the mixed spectrum; when detecting multiple gases, it is necessary to perform peak fitting on the mixed spectrum to obtain the fitted absorption function of each component gas, calculate the absorption peak fitting area of ​​each gas, and then invert the content of each gas component detected based on the concentration calculation formula derived from the Beer-Lambert law.

[0053] The present invention normalizes the collected decay time curve, calculates the mixed spectrum curve for peak fitting, and only selects the absorption peak of the gas to be measured for calculation, which can effectively reduce the influence of other gases on the gas to be measured, and also detects the content of other gas components; the advantage is that under normal temperature and pressure environment, the collision motion between particles is more obvious than the molecular thermal motion, so the selection of the Lorentz linear function describing the collision broadening for peak fitting is closer to reality, making the concentration calculation more accurate.

[0054] The present invention uses a single gas absorption peak obtained by a spectral analysis algorithm, integrates the peak area and substitutes it into a given concentration calculation formula. The concentration value of the trace gas can be calculated by a single-chip microcomputer and displayed on the screen. The advantage is that the algorithm used has a superlinear convergence speed, a simple algorithm structure, and is easy to program. It avoids the calculation of second-order derivatives, greatly reduces the amount of calculation and storage, occupies less resources on the main control chip, and improves the possibility of hardware implementation. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Attached Figure 1 It is a flow chart of the multi-component gas concentration detection method based on spectral analysis algorithm of the present invention;

[0056] Attached Figure 2 is a flow chart of the spectrum analysis algorithm of the present invention;

[0057] Attached Figure 3 It is a flow chart of iterative step length calculation in the spectrum analysis algorithm of the present invention;

[0058] Attached Figure 4 Graph showing the fitted mixed spectrum and the Lorentz distribution function after peak separation in an embodiment of the present invention;

[0059] Attached Figure 5 This is the process of normalizing the e-exponential ring-down curve and calculating the ratio of the emitted and transmitted light intensities in the embodiment of the present invention. DETAILED DESCRIPTION

[0060] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0061] like Figure 1-3 As shown: A multi-component gas concentration detection method based on spectral analysis algorithm, the steps of the detection method are as follows:

[0062] A. Change the wavelength of the incident light and record the ring-down time τ of the mixed gas t And draw the ring-down curve;

[0063] B. Normalize the ring-down curve;

[0064] C. Using the ring-down time τ of the mixed gas t Calculate the normalized ring-down intensity I t , ring-down intensity I t The solution formula is:

[0065]

[0066] In formula (1), I t It represents the optical signal intensity of the mixed gas under the same cavity ring-down time τ0; τ0 is the cavity ring-down time; τ t is the ring-down time of the mixed gas;

[0067] D. Solve The wave number v is the horizontal axis. Plot the mixed spectrum for the ordinate;

[0068] E. Using a spectrum analysis algorithm to perform peak separation on the mixed spectrum to obtain multiple Lorentz distribution function absorption peaks;

[0069] F. Performing peak area integral calculation on each Lorentzian distribution function absorption peak to obtain the peak integrated area;

[0070] G. Substitute the integrated area and line intensity of each peak, as well as the total gas pressure and effective optical path into the gas concentration calculation formula to calculate and obtain the gas concentration of each component in the mixed gas; the gas concentration calculation formula is:

[0071]

[0072] In formula (6), X represents the concentration of a gas component; It is the peak integral area of ​​a gas component obtained by peak separation using the spectrum analysis algorithm. Represents the incident light intensity I0 and the ring-down light intensity I after transmission t ratio; S(T) represents the line intensity of a gas component; p refers to the total pressure of the gas; L refers to the effective optical path;

[0073] When the ring-down light intensity I t Based on the solution formula, the gas concentration calculation formula in step G is:

[0074]

[0075] In the multi-component gas concentration detection method based on the spectral analysis algorithm provided by the present invention, the process of the spectral analysis algorithm in step E is: first determine the absorption fitting function containing the undetermined Lorentz characteristic parameters, use the absorption fitting function to construct the error function, determine the initial value coordinates of the iteration, determine the iteration direction, and then determine the iteration step according to the Armijo exact linear search method, and calculate the gradient and verify whether it meets the requirements. If so, continue; otherwise, change the iteration direction and iteration step at the same time until the minimum fitting error is found and the Lorentz characteristic parameter value in this case is determined, thereby determining the absorption fitting function of the gas to be measured and decomposing the mixed spectrum into multiple absorption peaks.

[0076] The specific steps of the spectrum analysis algorithm in step E are as follows:

[0077] E1. Determine the absorption fitting function containing the unknown Lorentz characteristic parameters:

[0078]

[0079] In formula (2), x is the wave number of the horizontal coordinate in the mixed spectrum, F is the absorbance of the vertical coordinate in the mixed spectrum, m represents the type of gas to be measured, and A i 、w i 、x ci is the characteristic parameter of the Lorentz line shape, where A i represents the absorption peak area of ​​the i-th gas component, wi represents the half-height width of the i-th gas component, Indicates the center position of the absorption peak of the i-th gas component;

[0080] E2. Use the absorption fitting function to construct the error function:

[0081]

[0082] In formula (3), Q is the fitting residual, y j represents the absorbance of the mixed spectrum, F(x j ) is the value of the absorption fitting function, and n is the number of spectral sampling points;

[0083] E3, set the iteration precision ε and the initial coordinates of the iteration parameters;

[0084] E4. Determine the search direction k :

[0085]

[0086] In formula (4), d k is the search direction for the kth iteration; d k-1 is the search direction for the k-1th iteration; β k-1 is the conjugate parameter of the k-1th iteration; g k is the kth iteration gradient direction; Transpose the gradient direction for the kth iteration; g is the gradient direction transposition for the k-1th iteration; k-1 is the kth iteration gradient direction; search direction d k The initial search direction is the negative gradient direction of the initial coordinates of the error function d0 = -g0;

[0087] E5. Determine the search direction k Then, the Armijo exact linear search method is used to determine the iteration step size α k , the specific steps are;

[0088] E51. Determine the initial parameters δ and σ of the Armijo exact linear search method;

[0089] E52, start cyclic search from m=0;

[0090] E53, if satisfied Then m=m k The iteration step length is

[0091] E54. If not satisfied, set m=m+1 and continue searching until Until establishment;

[0092] E6. Set the initial value of the current iteration (A ik ,w ik )(i is the type of gas, k is the number of iterations), let A k+1 =A k +α k ·d k 、w k+1 =w k +α k ·d k , calculate along the search direction d k and the iteration step α k The next point coordinates (A i(k+1) ,w i(k+1) ) and calculate the coordinates of the next point (A i(k+1) ,w i(k+1) )’s gradient value g k :

[0093]

[0094] In formula (5), is the gradient of the fitted residual matrix;

[0095] E7. Calculate the second norm || g k || and compare with the iteration accuracy ε set in step E3;

[0096] E8, if || g k ||<ε, then stop the operation and output the Lorentz characteristic parameter A k 、w k As the optimal solution, go to step E9; if || g k ||>ε, then return to step E3 and use the coordinates of the iteration parameters of this round as the initial coordinates of the iteration parameters in step E3, and repeat steps E4-E7;

[0097] E9, stop iteration and find the minimum fitting residual Q min , and simultaneously output the corresponding Lorentz characteristic parameter value A ik 、w ik ;

[0098] E10. Determine the absorption fitting function containing the determined Lorentz characteristic parameters of the gas to be measured, and decompose the mixed spectrum into multiple absorption peaks.

[0099] The purpose of the above step E8 is to ensure that the fitting residual Q value drops to the minimum value in the maximum descent direction, that is, to select the fastest descent direction and drop to the lowest point in this direction, thereby quickly finding the minimum fitting residual Q value.

[0100] Example

[0101] The present invention is described in detail below in conjunction with specific implementation cases, but the protection scope of the present invention is not limited by the drawings and examples.

[0102] Combination Figure 4 and Figure 5 Now we measure the trace amount of H2S and CO2 in the SF6 / N2 background. From the scanning spectrum, we know that the two peak center wavelengths of H2S and CO2 are 6336.7883cm -1 and 6336.3451cm -1 The present invention is further described. The calculation of multi-gas component concentration based on CRDS mainly includes the following steps:

[0103] 1) Change the wavelength of the incident light and record the ring-down time τ of the mixed gas t And draw the ring-down curve;

[0104] 2) Normalize the ring-down curve and use the ring-down time τ t Calculate the normalized ring-down intensity I t , with wave number v as the horizontal axis, The mixed spectrum is plotted for the ordinate, and the spectrum analysis algorithm is used for peak separation below;

[0105] 3) First determine the fitting peak function

[0106] For H2S and CO2, there should be two absorption peaks that conform to the Lorentz distribution within the spectral scanning range. At this time, the absorption spectrum expression is: The error function is expressed as:

[0107] 4) Fit the four Lorentz characteristic parameters A1, A2, w1, and w2 to solve the gradient of the error function: At the same time, the iteration accuracy ε needs to be set;

[0108] 5) Here, the iteration accuracy is set to ε = 0.001, and the initial value of the iteration is Substitute the initial gradient g0 and calculate the search direction d0 = -g0;

[0109] 6) The iteration step size is determined by the Armijo exact linear search method, and the initial parameters are set to σ = 0.6, δ = 0.4, and m = 0. If the inequality is satisfied Then m=m k The iteration step length is If m=m+1 is not satisfied, continue until the inequality is established;

[0110] 7) Let A k+1 =A k +α k·d k ,w k+1 =w k +α k ·d k , calculate the gradient g1, ||g1|| according to the parameter value of this iteration. If it is less than the iteration accuracy, output the parameter value; if it is greater than the iteration accuracy ε, recalculate the search direction Go back to step 5) and use the Armijo exact linear search method to calculate the iteration step length;

[0111] 8) Repeat steps 5)-7) until the norm of the parameter value gradient of this iteration is within the accuracy range initially set, and then the parameter value after iteration can be output. Here, the corresponding minimum fitting residual Q is found after calculation by the main control chip. min , the corresponding parameter values ​​obtained are A1=0.1081, A2=0.012, w1=0.1686, w2=0.1159;

[0112] 9) According to the obtained Lorentz characteristic parameters and the determined peak center wavelength, the mixed spectrum curve is decomposed into two Lorentz distribution absorption peaks, such as Figure 4 As can be seen, the dashed line is the CO2 absorption peak and the dotted line is the H2S absorption peak, and the area integration calculations are performed on the absorption peaks of the two gases respectively;

[0113] 10) Record the experimental values, the values ​​of the line intensity S(T) corresponding to the central wavelength of the two gases queried in the HITRAN database, and the absorption fitting spectral functions of the two gases respectively. and Perform peak integral area calculation, and substitute the peak integral area, effective optical path L, gas total pressure P, and line intensity S(T) into the given concentration calculation formula for calculation. The H2S gas concentration can be calculated to be 10.29ppm and the CO2 gas concentration is 1.53ppm, which are consistent with the concentration given by the gas distributor.

[0114] Depend on Figure 4 It can be seen that the present invention adopts a new spectrum analysis algorithm to perform peak fitting on the mixed spectrum and solve the concentration, and finally a concentration value with a high similarity to the target concentration can be obtained in the single chip microcomputer, which is a concentration calculation method with high accuracy.

[0115] The spectral analysis algorithm of the present invention performs peak separation processing on multiple data points on the collected absorbance-wavenumber curve, simplifies them into gas absorption peak areas, effectively simplifies data processing; at the same time, it avoids the problem of gas tailing affecting each other, and provides more reliable and accurate data support for concentration calculation and analysis.

[0116] The above embodiments are only for illustrating the technical idea of ​​the present invention, and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the present invention; any technology not involved in the present invention can be realized by existing technologies.

Claims

1. A multi-component gas concentration detection method based on spectral analysis algorithm, characterized in that: The steps of this detection method are as follows: A. Change the wavelength of the incident light and record the ring-down time τ of the mixed gas t And draw the ring-down curve; B. Normalize the ring-down curve; C. Using the ring-down time τ of the mixed gas t Calculate the normalized ring-down intensity I t ; D. Solve The wave number v is the horizontal axis. Plot the mixed spectrum for the ordinate; E. Using a spectrum analysis algorithm to perform peak separation on the mixed spectrum to obtain multiple Lorentz distribution function absorption peaks; F. Performing peak area integral calculation on each Lorentzian distribution function absorption peak to obtain the peak integrated area; G. Substitute the integrated area and line intensity of each peak, as well as the total gas pressure and effective optical path into the gas concentration calculation formula to calculate and obtain the gas concentration of each component in the mixed gas; The process of the spectrum analysis algorithm in step E is as follows: first determine the absorption fitting function containing the undetermined Lorentz characteristic parameters, use the absorption fitting function to construct the error function, determine the initial value coordinates of the iteration, determine the iteration direction, and then determine the iteration step size according to the Armijo exact linear search method, calculate the gradient and verify whether it meets the requirements, if so, continue, otherwise change the iteration direction and iteration step size at the same time until the minimum fitting error is found and the Lorentz characteristic parameter value in this case is determined, thereby determining the absorption fitting function of the gas to be measured and decomposing the mixed spectrum into multiple absorption peaks.

2. The multi-component gas concentration detection method based on spectral analysis algorithm according to claim 1 is characterized in that: The gas concentration calculation formula in step G is: In formula (6), X represents the concentration of a gas component; It is the peak integral area of ​​a gas component obtained by peak separation using the spectrum analysis algorithm. Represents the incident light intensity I0 and the ring-down light intensity I after transmission t ratio; S(T) represents the line intensity of a gas component; p refers to the total gas pressure; L refers to the effective optical path.

3. The multi-component gas concentration detection method based on spectral analysis algorithm according to claim 1 or 2, characterized in that: The ring-down light intensity I in step C t The solution formula is: In formula (1), I t It represents the optical signal intensity of the mixed gas under the same cavity ring-down time τ0; τ0 is the cavity ring-down time; τ t is the ring-down time of the mixed gas.

4. The multi-component gas concentration detection method based on spectral analysis algorithm according to claim 1 is characterized in that: The specific steps of the spectrum analysis algorithm in step E are as follows: E1. Determine the absorption fitting function containing the unknown Lorentz characteristic parameters: In formula (2), x is the wave number of the horizontal coordinate in the mixed spectrum, F is the absorbance of the vertical coordinate in the mixed spectrum, m represents the type of gas to be measured, and A i 、w i , is the characteristic parameter of the Lorentz line shape, where A i represents the absorption peak area of ​​the i-th gas component, w i represents the half-height width of the i-th gas component, Indicates the center position of the absorption peak of the i-th gas component; E2. Use the absorption fitting function to construct the error function: In formula (3), Q is the fitting residual, y j represents the absorbance of the mixed spectrum, F(x j ) is the value of the absorption fitting function, and n is the number of spectral sampling points; E3, set the iteration precision ε and the initial coordinates of the iteration parameters; E4. Determine the search direction d k : In formula (4), d k is the search direction for the kth iteration; d k-1 is the search direction for the k-1th iteration; β k-1 is the conjugate parameter of the k-1th iteration; g k is the kth iteration gradient direction; Transpose the gradient direction for the kth iteration; g is the gradient direction transposition for the k-1th iteration; k-1 is the kth iteration gradient direction; E5. Determine the search direction k Then, the Armijo exact linear search method is used to determine the iteration step size α k ; E6. Set the initial value of the current iteration (A ik ,w ik ), let A k+1 =A k +α k ·d k 、w k+1 =w k +α k ·d k , calculate along the search direction d k and the iteration step α k The next point coordinates (A i(k+1) ,w i(k+1) ) and calculate the coordinates of the next point (A i(k+1) ,w i(k+1) )’s gradient value g k : In formula (5), is the gradient of the fitted residual matrix; E7. Calculate the second norm || g k || and compare with the iteration accuracy ε set in step E3; E8, if || g k ||<ε, then stop the operation and output the Lorentz characteristic parameter A k 、w k As the optimal solution, go to step E9; if || g k ||>ε, then return to step E3 and use the coordinates of the iteration parameters of this round as the initial coordinates of the iteration parameters in step E3, and repeat steps E4-E7; E9, stop iteration and find the minimum fitting residual Q min , and simultaneously output the corresponding Lorentz characteristic parameter value A ik 、w ik ; E10. Determine the absorption fitting function containing the determined Lorentz characteristic parameters of the gas to be measured, and decompose the mixed spectrum into multiple absorption peaks.

5. The multi-component gas concentration detection method based on spectral analysis algorithm according to claim 4 is characterized in that: The search direction d in step E4 k The initial search direction is the negative gradient direction d0 = -g0 of the initial coordinates of the error function.

6. The multi-component gas concentration detection method based on spectral analysis algorithm according to claim 4 is characterized in that: The Armijo exact linear search method is used in step E5 to determine the iteration step length α k The detailed steps are: E51. Determine the initial parameters δ and σ of the Armijo exact linear search method; E52, start cyclic search from m=0; E53, if satisfied Then m=m k The iteration step length is E54. If not satisfied, set m=m+1 and continue searching until Until establishment.

7. The multi-component gas concentration detection method based on spectral analysis algorithm according to claim 1 is characterized in that: The multi-component gas concentration detection method can be applied to the detection of various trace gas concentrations under the background of SF6 / N2 in the CRDS field.

Images