A high-precision spectral noise elimination method and system based on TDLAS technology

Abstract

The application discloses a high-precision spectral noise elimination method and system based on TDLAS technology and relates to the technical field of gas concentration optical detection. The method obtains the center frequency and half-width parameter of a Lorentz model by fitting the standard gas sampling data, determines the adaptive cutoff frequency according to the center frequency and the half-width parameter, and performs digital low-pass filtering on the to-be-detected gas sampling data. The filtered data is divided into an absorption interval and two non-absorption intervals. The local phase information of the interference fringes in the non-absorption intervals is extracted and weighted third-order polynomial fitting is performed to obtain the baseline of the whole region. The application extracts the local phase of the interference fringes in the non-absorption intervals in real time and converts the local phase into the basis for weighted fitting. The processing link is constructed by combining adaptive filtering and constrained Lorentz fitting. The phase extrapolation compensation is used to eliminate the residual error of the interference fringes caused by temperature fluctuation, the spectral absorption area more accurately reflects the real absorption strength, and thus the precision of low-concentration gas detection is improved.

Description

Technical Field

[0001] This invention relates to the field of optical gas concentration detection technology, specifically to a high-precision spectral noise elimination method and system based on TDLAS technology. Background Technology

[0002] Tunable diode laser absorption spectroscopy (TDLAS) is a gas detection technique based on the wavelength tuning characteristics of narrow-linewidth lasers. This technique utilizes the characteristic that the laser output wavelength varies with the injection current or operating temperature to scan the characteristic absorption lines of the target gas. By analyzing the degree of light intensity attenuation after the laser passes through the gas, the gas concentration is deduced using Beer-Lambert's law. Due to the narrow linewidth, high wavelength tuning resolution, and good wavelength stability of semiconductor lasers, TDLAS technology can achieve highly selective measurement of single absorption lines of specific gas molecules, effectively avoiding spectral cross-interference from other gas components. It has been widely used in industrial process control, atmospheric environmental monitoring, combustion diagnostics, and respiratory gas analysis.

[0003] In a TDLAS measurement system, the light signal output from the laser passes sequentially through optical elements such as a collimating lens, a gas absorption cell, and a focusing lens before reaching a photodetector. The detector converts the light signal into an electrical signal, which is then sampled by an analog-to-digital converter to obtain the original spectral signal. Ideally, if there is no target absorbing gas in the optical path, the signal received by the detector should present a smooth spectral baseline. However, in actual optical systems, the laser undergoes multiple reflections and interferences between multiple optical interfaces, superimposing quasi-periodic oscillating interference fringes onto the spectral signal received by the detector. The amplitude and period of the interference fringes are determined by the surface reflectivity and spacing of the optical elements, as well as the laser wavelength, and their oscillation characteristics exhibit a stable sinusoidal distribution in the wavenumber or wavelength domain.

[0004] To obtain accurate gas absorption line shapes for concentration inversion, it is typically necessary to remove interference fringes and other background components from the original spectral signal. The conventional process involves pre-acquiring a background spectral signal under conditions without the target absorbing gas as a reference template. During the actual measurement, the spectral signal of the gas to be measured is subtracted from this reference template to remove the background. Under constant temperature or gradual temperature changes in a laboratory setting, the morphology and phase of the interference fringes are relatively stable, and the aforementioned background subtraction method achieves satisfactory results, thus it is widely used. However, as TDLAS technology expands into applications such as industrial field online monitoring, outdoor atmospheric trace gas observation, and vehicle-mounted or airborne mobile measurements, the measurement system needs to cope with temperature disturbances caused by diurnal temperature variations, seasonal temperature changes, and equipment self-heating, making spectral baseline processing face more complex conditions.

[0005] The limitations of existing technologies include at least the following problems: Existing baseline correction methods are based on the assumption that the interference fringe morphology remains stable across multiple measurements. They use static background subtraction or fitting methods to strip the interference fringes, but do not consider the nonlinear drift of the interference fringe phase under temperature change conditions. When the measurement system is in a temperature fluctuation scenario, the temperature disturbance simultaneously affects the laser wavelength tuning characteristics and the physical length of the optical cavity. The two have a coupled relationship in modulating the phase of the interference fringes, resulting in a phase mismatch between the pre-acquired background template and the interference fringes at the actual measurement time that is difficult to eliminate by translation alignment. After subtraction, the baseline has an oscillating residual. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a high-precision spectral noise elimination method and system based on TDLAS technology, which solves the problem of residual oscillation after baseline subtraction caused by phase mismatch of interference fringes due to temperature fluctuations in existing technologies.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a high-precision spectral noise elimination method based on TDLAS technology, comprising the following steps: acquiring sampling data of known standard gas concentrations and fitting them with a Lorentz model to obtain center frequency parameters and full width at half maximum (FWHM) parameters; acquiring sampling data of the gas concentration to be measured, determining an adaptive cutoff frequency based on the center frequency parameters and FWHM parameters, and performing digital low-pass filtering to obtain filtered sampling data of the gas concentration to be measured; dividing the filtered sampling data of the gas concentration to be measured into an absorption interval and non-absorption intervals located on both sides of the absorption interval; extracting local phase information of interference fringes in the non-absorption interval, calculating the phase change rate based on the local phase information, and generating fitting weight coefficients for each sampling point in the non-absorption interval according to the phase change rate; using the fitting weight coefficients to perform weighted third-order polynomial fitting on the data in the non-absorption interval, and extrapolating the fitted polynomial curve to the entire region to obtain the full-region data. The baseline line shape data of the absorption region is obtained; the center frequency parameter and the full width at half maximum (FWHM) parameter are used as prior constraints for fitting, and the variation tolerance ranges of the center frequency parameter and the FWHM parameter are set. Under the fitting prior constraints, Lorentz fitting is performed on the data in the absorption region to obtain Lorentz fitted signal data; within the absorption region, the Lorentz fitted signal data is subtracted from the baseline line shape data of the entire region to obtain the preliminary absorption signal; the local phase information of the interference fringes in the non-absorption region is extrapolated to the absorption region to obtain the estimated residual oscillation phase in the absorption region; based on the estimated residual oscillation phase and the estimated residual amplitude determined by the amplitude level of the interference fringes in the non-absorption region, the estimated residual of the interference fringes in the absorption region is generated, and the estimated residual of the interference fringes is subtracted from the preliminary absorption signal to obtain the compensated absorption signal; the spectral absorption area is obtained by integrating the compensated absorption signal, and the concentration of the gas to be measured is calculated based on the spectral absorption area.

[0008] Furthermore, the specific steps for determining the adaptive cutoff frequency and performing digital low-pass filtering based on the center frequency parameter and the full width at half maximum (FWHM) parameter are as follows: Perform a Fourier transform on the sampling data of the gas concentration to be measured to obtain the spectral distribution of the sampling data; calculate the theoretical upper limit of the gas absorption signal in the frequency domain based on the center frequency parameter and the FWHM parameter; extend the theoretical upper limit of the gas absorption signal in the frequency domain outward by a preset protection bandwidth ratio to obtain the adaptive cutoff frequency; use the adaptive cutoff frequency to perform digital low-pass filtering on the sampling data of the gas concentration to be measured to obtain the filtered sampling data of the gas concentration to be measured.

[0009] Furthermore, the specific steps for dividing the filtered gas concentration sampling data into absorption intervals and non-absorption intervals located on both sides of the absorption interval are as follows: Locate the start and end positions of the gas absorption characteristics in the filtered gas concentration sampling data; define the interval from the start to the end position of the gas absorption characteristics as the absorption interval; define the continuous interval to the left of the absorption interval without gas absorption characteristics as the non-absorption interval; define the continuous interval to the right of the absorption interval without gas absorption characteristics as the non-absorption interval.

[0010] Furthermore, the specific steps for extracting the local phase information of the interference fringes in the non-absorption interval are as follows: perform autocorrelation operation on the data in the non-absorption interval to obtain the correlation peak position of the interference fringes in the non-absorption interval; calculate the interference fringe phase value of each sampling point in the non-absorption interval based on the correlation peak position of the interference fringes in the non-absorption interval; smooth the interference fringe phase value of each sampling point in the non-absorption interval to remove abnormal phase points; and output the local phase information of the interference fringes in the non-absorption interval.

[0011] Furthermore, the specific steps for generating the fitting weight coefficients for each sampling point within the non-absorption interval based on the phase change rate are as follows: calculate the absolute value of the phase difference between adjacent sampling points as the phase change rate; based on the inverse correspondence between the phase change rate and the fitting weight coefficients, assign higher fitting weight coefficients to sampling points with smaller phase change rates and lower fitting weight coefficients to sampling points with larger phase change rates.

[0012] Furthermore, the specific steps for Lorentz fitting of the data in the absorption interval under the fitting prior constraints are as follows: construct a Lorentz fitting objective function with a center frequency parameter tolerance term and a half-width parameter tolerance term; use an iterative algorithm to fit the data in the absorption interval; stop iterating when the fitting residual converges to a preset convergence threshold to obtain the Lorentz fitted signal data.

[0013] Furthermore, based on the predicted residual oscillation phase and the predicted residual amplitude determined by the amplitude level of the interference fringes in the non-absorption interval, the specific steps for generating the predicted residual of the interference fringes in the absorption interval are as follows: calculate the absolute value of the difference between the data of each sampling point in the non-absorption interval and the corresponding baseline value; average the absolute values ​​of each difference to obtain the mean amplitude of the interference fringes in the non-absorption interval; use the mean amplitude as the predicted residual amplitude in the absorption interval; generate the predicted residual of the interference fringes in the absorption interval based on the predicted residual amplitude and the predicted residual oscillation phase.

[0014] Furthermore, the specific steps for integrating the compensated absorption signal to obtain the spectral absorption area are as follows: determine the lower limit and upper limit of integration of the compensated absorption signal within the absorption interval; perform numerical integration on the compensated absorption signal within the interval from the lower limit to the upper limit; and obtain the spectral absorption area of ​​the gas concentration to be measured.

[0015] Furthermore, the specific steps for calculating the concentration of the gas to be measured based on the spectral absorption area are as follows: retrieve the calibration correspondence between the known standard gas concentration and the spectral absorption area; substitute the spectral absorption area of ​​the gas to be measured into the calibration correspondence between the known standard gas concentration and the spectral absorption area to calculate the initial concentration value of the gas to be measured; perform systematic error correction on the initial concentration value of the gas to be measured, and output the final concentration of the gas to be measured.

[0016] A high-precision spectral noise cancellation system based on TDLAS technology includes: a model parameter acquisition unit for acquiring sampling data of known standard gas concentrations and performing Lorentz model fitting to obtain center frequency parameters and full width at half maximum (FWHM) parameters; an adaptive filtering unit for acquiring sampling data of the gas concentration to be measured, determining an adaptive cutoff frequency based on the center frequency parameters and FWHM parameters, and performing digital low-pass filtering to obtain filtered sampling data of the gas concentration to be measured; an interval division unit for dividing the filtered sampling data of the gas concentration to be measured into absorption intervals and non-absorption intervals located on both sides of the absorption intervals; and a baseline fitting unit for extracting local phase information of interference fringes in the non-absorption intervals, calculating the phase change rate based on the local phase information, generating fitting weight coefficients for each sampling point in the non-absorption intervals based on the phase change rate, performing weighted third-order polynomial fitting on the data in the non-absorption intervals using the fitting weight coefficients, and extrapolating the fitted polynomial curve to the entire region to obtain the full region. The data includes baseline line shape data; an absorption fitting compensation unit, which uses the center frequency parameter and the full width at half maximum (FWHM) parameter as fitting prior constraints, sets the variation tolerance range of the center frequency parameter and the FWHM parameter, and performs Lorentz fitting on the data in the absorption interval under the fitting prior constraints to obtain Lorentz fitted signal data; within the absorption interval, the Lorentz fitted signal data is processed by difference with the baseline line shape data of the entire region to obtain a preliminary absorption signal; the local phase information of the interference fringes in the non-absorption interval is extrapolated to the absorption interval to obtain the estimated residual oscillation phase in the absorption interval; based on the estimated residual oscillation phase and the estimated residual amplitude determined by the amplitude level of the interference fringes in the non-absorption interval, the estimated residual of the interference fringes in the absorption interval is generated, and the preliminary absorption signal is subtracted from the estimated residual of the interference fringes to obtain the compensated absorption signal; the compensated absorption signal is integrated to obtain the spectral absorption area, and the concentration of the gas to be measured is calculated based on the spectral absorption area.

[0017] The present invention has the following beneficial effects:

[0018] (1) The high-precision spectral noise elimination method based on TDLAS technology pre-collects sampling data of known standard gas concentrations and fits the Lorentz model, extracts the center frequency and half width at half maximum (WHM) parameters as core reference benchmarks. When detecting the gas to be tested, it abandons the traditional fixed filter parameter mode, dynamically calculates the theoretical bandwidth upper limit of the gas absorption signal in the frequency domain based on the above two model parameters, and then expands the preset protection bandwidth to generate an adaptive cutoff frequency, thereby driving the digital low-pass filter to work. This processing method makes the filtering operation closely fit the actual characteristics of the current gas absorption profile, which avoids the problem of incorrect removal of useful absorption signal edge components that may be caused by a fixed cutoff frequency, and also prevents unnecessary high-frequency interference introduced by an excessively high cutoff frequency. The filtered sampling data can achieve accurate noise removal while completely preserving the gas absorption profile.

[0019] (2) The high-precision spectral noise elimination method based on TDLAS technology performs autocorrelation calculation and local phase extraction on the interference fringes in the non-absorption region, and converts the obtained phase information into fitting weight coefficients for each sampling point. This achieves an optimization and improvement of the traditional polynomial baseline fitting method. In a temperature fluctuation environment, the phase of the interference fringes will undergo nonlinear drift. The traditional equal-weighted fitting is easily affected by the region of drastic phase change, causing the baseline estimation to deviate from the true background trend. This invention accurately identifies the characteristics of the interference fringes peaks and troughs where the phase change is gentle and the drift interference is small. It assigns higher fitting weights to the sampling points in this region, and reduces the fitting contribution of regions with drastic phase change and steep slope. The third-order polynomial fitting after weighting can more accurately capture the slowly changing background caused by the reflection of optical elements and avoid being constrained by local oscillation fluctuations.

[0020] (3) The high-precision spectral noise elimination method based on TDLAS technology uses the center frequency and half-width parameters obtained from the fitting of standard gas as prior constraints and applies them to the Lorentz fitting process in the absorption range. This effectively solves the technical problem of the fitting algorithm diverging and getting trapped in local optima due to the weak absorption signal under low concentration conditions. Under this constraint framework, the iterative algorithm is limited to searching for the optimal solution within the credible tolerance range of the prior parameters, rather than blindly exploring in the full parameter space. This strategy allows the fitting parameters to be slightly adaptively adjusted according to the actual measurement conditions to match the current line broadening state, and also eliminates the possibility that the center frequency and half-width are dragged to physically unreasonable ranges due to noise interference. Combined with the interference fringe residual extrapolation compensation step driven by the phase information of the non-absorption range, this invention eliminates the phase mismatch residue between the background template and the real-time signal caused by temperature changes, thereby achieving a lower detection limit and more stable repeatable measurement results in trace detection.

[0021] (4) The high-precision spectral noise elimination system based on TDLAS technology provides reliable linear prior knowledge through the model parameter acquisition unit. The adaptive filtering unit performs dynamic noise reduction that matches the absorption characteristics. The interval division unit automatically identifies the boundary between the absorption region and the non-absorption region. The baseline fitting unit restores the accurate background using the interference fringe phase weighting technique. The absorption fitting compensation unit performs phase extrapolation compensation and completes the concentration calculation based on the constrained fitting. No manual intervention is required. It can adapt to measurement drift under different temperature environments and ensure the stability of detection accuracy.

[0022] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0023] Figure 1This is a flowchart of a high-precision spectral noise elimination method based on TDLAS technology according to the present invention.

[0024] Figure 2 This is an example diagram of the original ADC sampling signal of the spectral absorption signal at different gas concentrations in a high-precision spectral noise elimination method based on TDLAS technology according to the present invention.

[0025] Figure 3 This is an example diagram showing the spectral distribution of spectral absorption signals of different gas concentrations in a high-precision spectral noise elimination method based on TDLAS technology according to the present invention.

[0026] Figure 4 This is an example diagram of the original ADC sampling signals of the spectral absorption of three groups of pure nitrogen gases at different times, at low concentrations of 10 ppm and 1000 ppm, in a high-precision spectral noise elimination method based on TDLAS technology according to the present invention.

[0027] Figure 5 This is an example diagram of the signal after low-pass filtering of the original ADC sampling signal of spectral absorption in the high-precision spectral noise elimination method based on TDLAS technology of the present invention.

[0028] Figure 6 This is an example of signal comparison between baseline fitting and Lorentz fitting of the absorption interval in the high-precision spectral noise reduction method based on TDLAS technology of the present invention.

[0029] Figure 7 This is a block diagram of a high-precision spectral noise cancellation system based on TDLAS technology according to the present invention. Detailed Implementation

[0030] Please see Figure 1-6This invention provides a technical solution: a high-precision spectral noise reduction method based on TDLAS technology, comprising the following steps: acquiring sampling data of a known standard gas concentration and fitting it with a Lorentz model to obtain a center frequency parameter and a full width at half maximum (FWHM) parameter; acquiring sampling data of a gas of a desired concentration, determining an adaptive cutoff frequency based on the center frequency parameter and FWHM parameter, and performing digital low-pass filtering to obtain filtered sampling data of the gas of the desired concentration; dividing the filtered sampling data of the gas of the desired concentration into an absorption region and non-absorption regions located on both sides of the absorption region; extracting the local phase information of the interference fringes in the non-absorption region, according to... This paper performs a weighted third-order polynomial fitting on the data in the non-absorption region to obtain the baseline line shape data of the entire region. Using the center frequency parameter and the full width at half maximum (FWHM) parameter as prior constraints, a Lorentz fitting is performed on the data in the absorption region under constraints to obtain the Lorentz fitting signal data. Within the absorption region, the baseline line shape data of the entire region is subtracted from the Lorentz fitting signal data to obtain the preliminary absorption signal. The local phase information of the interference fringes extracted from the non-absorption region is used to perform phase extrapolation compensation on the interference fringe residuals in the absorption region. The compensated absorption signal is integrated to obtain the spectral absorption area. The concentration of the gas to be measured is calculated based on the spectral absorption area.

[0031] Among them, the sampling data of the known standard gas concentration and the sampling data of the gas to be measured are both raw ADC sampling data collected by the detector in the TDLAS system. The sampling frequency is set to 100kHz-500kHz and the number of sampling points is set to 1024-4096.

[0032] The formula for fitting the Lorentz model is:

[0033] ;

[0034] in, For the sampled signal strength, For the maximum intensity of the absorption peak, For the center frequency parameter, The half-width and height parameters, The baseline offset is used to solve for the center frequency and full width at half maximum (FWHM) parameters using the least squares method. The fitting residuals must be less than 10. -6 .

[0035] Specifically, the steps for determining the adaptive cutoff frequency and performing digital low-pass filtering based on the center frequency parameter and the full width at half maximum (FWHM) parameter are as follows:

[0036] Performing a Fourier transform on the sampling data of the gas at the desired concentration yields the spectral distribution of the sampling data, specifically:

[0037] The Discrete Fourier Transform (DFT) is used to transform the data at all sampling points. This transformation converts the time-domain sampling data into a frequency-domain signal, clarifying the distribution of signal energy at different frequencies, thereby distinguishing the frequency range of gas absorption signals from high-frequency noise.

[0038] For example, if the number of sampling points is 2048 and the sampling frequency is 200kHz, the frequency domain resolution is approximately 97.66Hz, which can clearly distinguish the frequency range of gas absorption signals and high-frequency noise.

[0039] The theoretical upper limit of the bandwidth of the gas absorption signal in the frequency domain is calculated based on the center frequency parameter and the full width at half maximum (FWHM) parameter. Specifically:

[0040] Based on the time-frequency correspondence of the Lorentz curve, the center frequency parameter corresponds to the center position in the frequency domain, and the half-width at half-maximum (FWHM) parameter corresponds to the frequency domain broadening range. The formula for calculating the theoretical upper limit of the bandwidth of the gas absorption signal in the frequency domain is as follows:

[0041] ;

[0042] That is, it covers a frequency range of 2.5 times the width and height on both sides of the center frequency, ensuring that all frequency components of the gas absorption signal are completely included;

[0043] For example, if the center frequency parameter is 1570nm, the full width at half maximum (FWHM) parameter is 0.01nm, and the laser emission frequency is 191THz, then the frequency domain broadening range is approximately 12.16MHz, and the theoretical upper limit of the bandwidth is the center position of the frequency domain plus 24.32MHz.

[0044] The adaptive cutoff frequency is obtained by extending the theoretical upper limit of the gas absorption signal in the frequency domain outward by a preset protection bandwidth ratio, specifically as follows:

[0045] The preset protection bandwidth ratio is 10%-20%, and the formula for calculating the adaptive cutoff frequency is:

[0046] ;

[0047] in, To protect the bandwidth ratio ( );

[0048] The purpose of setting a protection bandwidth is to prevent the gas absorption signal from being filtered out due to frequency offset.

[0049] The sampling data of the gas concentration to be measured is digitally low-pass filtered using an adaptive cutoff frequency to obtain the filtered sampling data of the gas concentration to be measured, as follows:

[0050] The filtering is achieved using an FIR low-pass filter with an order of 128-256 and a Hanning window as the window function. The filter retains only the signal components below the adaptive cutoff frequency and filters out signals above that frequency. The filtering process is achieved by weighted summation of the filter coefficients and the delayed original sampled data.

[0051] In this implementation scheme, the spectral distribution is obtained by performing a Fourier transform on the sampling data of the gas concentration to be measured. The theoretical upper limit of the bandwidth of the gas absorption signal in the frequency domain is calculated by using the center frequency parameter and the full width at half maximum (FWHM) parameter obtained by fitting a known standard gas. Then, the preset protection bandwidth is extended outward to form an adaptive cutoff frequency. This allows the cutoff boundary of the low-pass filter to be adjusted in real time according to the current absorption line characteristics, thereby avoiding the hidden danger that a fixed cutoff frequency may eliminate the absorption edge components. The combination of the FIR filter and the Hanning window suppresses high-frequency electrical noise while maintaining the signal flatness and phase linearity in the passband. This ensures that the filtered gas sampling data retains the absorption profile and interference fringe pattern completely while significantly reducing the interference of irrelevant noise.

[0052] Specifically, the steps for dividing the filtered sampling data of the gas concentration to be measured into an absorption interval and non-absorption intervals located on both sides of the absorption interval are as follows:

[0053] The starting and ending positions of the gas absorption characteristics in the sampling data of the gas concentration to be measured after localization filtering are as follows:

[0054] The threshold method is used to locate absorption features. First, the average value of all sampled data after filtering is calculated, and then the standard deviation of the sampled data is calculated. The absorption feature identification threshold is set to the average value plus three times the standard deviation. When the data of a certain sample point is greater than the threshold, the sample point is determined to be an absorption feature point.

[0055] The starting position of the absorption feature is the index of the first sampling point that meets the condition, and the ending position of the absorption feature is the index of the last sampling point that meets the condition.

[0056] For example: if the average value of the filtered sampled data is 2.5V and the standard deviation is 0.1V, then the recognition threshold is 2.8V. The sampling point with the serial number 320 first satisfies the data being greater than 2.8V, and the sampling point with the serial number 580 last satisfies the data being greater than 2.8V. Therefore, the starting position is 320 and the ending position is 580.

[0057] The interval from the start position to the end position of the gas absorption characteristic is defined as the absorption interval, which is as follows:

[0058] The sampling point number range of the absorption interval is from the start position to the end position. The corresponding frequency range is calculated by the sampling point number, sampling frequency and total number of sampling points. This interval completely covers all sampling points of the gas absorption characteristics, ensuring that the subsequent Lorentz fitting is only performed on the absorption signal.

[0059] For example: if the starting position is 320, the ending position is 580, the sampling frequency is 200kHz, and the total number of sampling points is 2048, then the frequency range corresponding to the absorption interval is approximately 31.25kHz to 55.66kHz.

[0060] The continuous interval to the left of the absorption interval, where there is no gas absorption characteristic, is defined as the non-absorption interval, specifically as follows:

[0061] The sampling point number range for the non-absorption region on the left is from 0 to the starting position minus 1, and the corresponding frequency range is from 0 to the starting frequency of the absorption region minus the frequency domain resolution.

[0062] All sampling points within this interval do not meet the absorption feature recognition conditions, have no gas absorption features, and only contain interference fringes and background noise, which are used to extract local phase information of the interference fringes.

[0063] For example: corresponding to the absorption interval mentioned above, the non-absorption interval on the left has a serial number range of 0 to 319 and a frequency range of 0 to 31.15kHz;

[0064] The continuous interval to the right of the absorption interval, where there is no gas absorption characteristic, is defined as the non-absorption interval, specifically as follows:

[0065] The sampling point number range for the non-absorption interval on the right is from the end position plus 1 to the total number of sampling points minus 1, and the corresponding frequency range is from the end frequency of the absorption interval plus the frequency domain resolution to half of the sampling frequency.

[0066] Consistent with the non-absorption region on the left, this region has no gas absorption characteristics, only interference fringes and background noise, and can be matched with the non-absorption region on the left.

[0067] For example, corresponding to the absorption interval mentioned above, the non-absorption interval on the right ranges from 581 to 2047, with a frequency range of 55.76kHz to 100kHz.

[0068] In this implementation scheme, the start and end boundaries of gas absorption features are automatically located by using the mean and standard deviation of the sampled data after statistical filtering and constructing a dynamic threshold. This eliminates the limitations of manually setting fixed intervals, allowing the division of the absorption interval and the non-absorption intervals on both sides to be adaptively completed according to the actual shape of the current signal. The absorption interval completely covers the gas absorption range, ensuring that the subsequent Lorentz fitting operation only acts on the target area and is not affected by non-absorption noise. The non-absorption intervals on both sides do not contain gas absorption features and only carry interference fringes and background noise, providing a pure data source for the extraction of the local phase of the interference fringes. Thus, even when the absorption peak position is slightly shifted due to temperature fluctuations, the boundaries of each region can still be accurately defined, thereby laying a reliable spatial division foundation for weighted baseline fitting and phase extrapolation compensation.

[0069] Specifically, the steps for extracting the local phase information of interference fringes in the non-absorption region are as follows:

[0070] Autocorrelation calculations were performed on the data in the non-absorption region to obtain the positions of the correlation peaks of the interference fringes within the non-absorption region, specifically:

[0071] Autocorrelation is performed on all data in the non-absorption region on the left or right. By calculating the correlation of the data itself under different delay times, the delay time with the strongest correlation is found. This delay time is the position of the correlation peak, which corresponds to the period of the interference fringes.

[0072] For example: if there are 320 sampling points in the non-absorption region on the left, after autocorrelation operation, the correlation is strongest when the delay time is 16, then the position of the correlation peak is 16, that is, the period of the interference fringes is 16 sampling points.

[0073] Based on the correlation peak positions of the interference fringes within the non-absorption interval, the phase values ​​of the interference fringes at each sampling point within the non-absorption interval are calculated, specifically as follows:

[0074] The angular frequency of the interference fringes is calculated based on the position of the relevant peak. The angular frequency is 2π divided by the period corresponding to the position of the relevant peak. It is assumed that the interference fringes are a cosine oscillation signal, which includes the amplitude and the local phase that varies with the sampling point.

[0075] The phase of each sampling point is extracted by Hilbert transform. Hilbert transform can convert a real signal into an analytic signal, and then the local phase value of each sampling point can be obtained by arctangent operation.

[0076] For example: if the position of the correlation peak is 16, then the angular frequency is π / 8. The phase of the sampling point with the number 10 is calculated to be π / 4 and the phase of the sampling point with the number 20 is π / 2 through Hilbert transform.

[0077] The phase values ​​of the interference fringes at each sampling point within the non-absorption interval are smoothed to remove abnormal phase points. Specifically:

[0078] The smoothing process is performed using the moving average method, with the sliding window size set to 5-9 points. The smoothed phase value of a sampling point is obtained by calculating the average phase value of a sampling point and several adjacent points, and the boundary points are padded with zeros.

[0079] The criteria for determining abnormal phase points are as follows:

[0080] If the difference between the original phase and the smoothed phase of a certain sampling point is greater than 2π / 3, then the point is determined to be an abnormal phase point, and linear interpolation of the phases of two adjacent points is used to replace it.

[0081] For example, if the sliding window has 5 points, the original phase of a certain sampling point is 3π / 2, and the smoothed phase is π / 2. The difference is greater than 2π / 3. Then, the phase of the two adjacent points (both π / 2) is used for interpolation to replace the phase, and the corrected phase value is π / 2.

[0082] The local phase information of the interference fringes in the non-absorption region is output, specifically as follows:

[0083] After smoothing and removing outliers, the local phase values ​​of each sampling point are output to form a complete phase sequence. This phase sequence reflects the phase distribution of the interference fringes in the non-absorption region and includes information on the phase nonlinear drift caused by temperature fluctuations.

[0084] For example, the phase sequence of the non-absorption region on the left side of the output is a continuous sequence from 0 to 2π, clearly showing the phase change pattern of the interference fringes.

[0085] In this implementation scheme, the correlation peak positions of the interference fringes are obtained by performing autocorrelation calculation on the data in the non-absorption interval, thereby determining the periodic characteristics of the interference fringes and providing an accurate frequency reference for phase extraction. The real signal is converted into an analytic signal by Hilbert transform and the local phase value of each sampling point is obtained by arctangent operation, so that the phase distribution of the interference fringes can be accurately quantified. The moving average smoothing process, combined with the abnormal phase point discrimination and linear interpolation replacement mechanism, effectively eliminates the phase jump caused by noise disturbance, ensuring the continuity and reliability of the output phase sequence.

[0086] Specifically, the steps to obtain the baseline shape data for the entire region are as follows:

[0087] The local phase information of the interference fringes within the non-absorption region is converted into fitting weight coefficients for each sampling point within the non-absorption region. Specifically, sampling points at the peaks and troughs of the interference fringes receive high fitting weight coefficients, while sampling points at locations with steep slopes receive low fitting weight coefficients.

[0088] Calculate the phase change rate of each sampling point, which is the absolute value of the phase difference between two adjacent sampling points. The phase change rate of the boundary point is consistent with that of the previous sampling point.

[0089] The smaller the phase change rate, the gentler the slope of the interference fringes (the phase change rate is smallest at the peaks and troughs); the larger the phase change rate, the steeper the slope of the interference fringes.

[0090] The formula for calculating the weighting coefficient is:

[0091] ;

[0092] Where k is the weight adjustment coefficient (k=10-20), which is used to amplify the influence of the phase change rate on the weight;

[0093] For example: if k=15, the phase change rate of the sampling point at the peak is 0.05, then the weighting coefficient is 0.8 (high weighting).

[0094] If the phase change rate of the sampling point at the steep slope is 0.5, then the weighting coefficient is 0.117 (low weight).

[0095] The fitted weight coefficients of each sampling point within the non-absorption interval are multiplied one by one with the data in the non-absorption interval to obtain the weighted data in the non-absorption interval, specifically as follows:

[0096] The original filtered data of each non-absorption interval sampling point is multiplied by the corresponding fitting weight coefficient to obtain the weighted data;

[0097] By weighting, the contribution of sampling points at peaks and troughs to baseline fitting is enhanced, while the interference of sampling points at steep slopes (which are more affected by phase drift) is weakened.

[0098] For example: if the original data of a sampling point in a non-absorption interval is 2.6V and the weighting coefficient is 0.8, then the weighted data is 2.08V;

[0099] The original data for another sampling point is 2.7V, and the weighting coefficient is 0.117, so the weighted data is approximately 0.316V;

[0100] A third-order polynomial was fitted to the weighted data in the non-absorption region to obtain the fitted polynomial curve, which is as follows:

[0101] A third-order polynomial is used to fit the weighted data. The third-order polynomial contains constant terms, linear terms, quadratic terms, and cubic terms. The coefficients of each term of the polynomial are solved by the weighted least squares method. The goal is to minimize the sum of squares of the deviations between the weighted data and the polynomial calculation values, thereby obtaining the fitted polynomial curve.

[0102] For example: after fitting, the constant term is 2.48, and the coefficient of the linear term is 3.2 × 10⁻⁶. -4 The coefficient of the quadratic term is -1.5 × 10⁻⁶. -7 The coefficient of the cubic term is 2.1 × 10⁻⁶. -11 Based on this, a complete fitting polynomial curve is formed;

[0103] Extrapolating the fitted polynomial curve to the entire region yields the baseline linearity data for the entire region, specifically:

[0104] The entire region includes all sampling points. Substitute the index of each sampling point into the above third-order polynomial to calculate the baseline value of the corresponding sampling point. The baseline values ​​of all sampling points constitute the baseline line data of the entire region.

[0105] During extrapolation, the polynomial coefficients are kept constant to ensure the continuity of the baseline shape and avoid abrupt changes at the boundary between the absorbing and non-absorbing regions.

[0106] For example, by substituting the sampling point with serial number 400 (within the absorption interval) into the polynomial, the baseline value of the sampling point is calculated to be approximately 2.60V.

[0107] In this implementation, the local phase information of the interference fringes in the non-absorption region is converted into fitting weight coefficients for each sampling point. This results in higher weights for regions with gentle phase changes, such as peaks and troughs, while areas with steep slopes are given lower weights due to greater phase drift interference. The weighted data is then fitted with a third-order polynomial, which effectively reduces the constraint of phase mismatch caused by temperature fluctuations on baseline estimation. The resulting polynomial curve is closer to the actual slow-changing background trend. When this curve is extrapolated to the entire region, the polynomial coefficients remain unchanged, ensuring that the baseline at the boundary between the absorption region and the two non-absorption regions is continuous and without abrupt changes.

[0108] Specifically, the steps to obtain the Lorentz-fitted signal data are as follows:

[0109] Using the center frequency parameter and the full width at half maximum (FWHM) parameter as prior constraints for fitting, the tolerance ranges for the variation of the center frequency parameter and the FWHM parameter are set as follows:

[0110] The tolerance range for the center frequency parameter is ±0.5% to ±2%, meaning the lower limit is the center frequency parameter minus 0.5%-2%, and the upper limit is the center frequency parameter plus 0.5%-2%.

[0111] The tolerance range for the half-width parameter is ±5% to ±15%, meaning that the lower limit is the half-width parameter minus 5%-15% of itself, and the upper limit is the half-width parameter plus 5%-15% of itself.

[0112] For example, if the center frequency parameter is 1570nm and the variation ratio is 0.01, then the center frequency variation tolerance range is 1568.43nm to 1571.57nm.

[0113] If the full width at half maximum (FWHM) parameter is 0.01 nm and the variation ratio is 0.1, then the FWHM variation tolerance range is 0.009 nm to 0.011 nm.

[0114] Construct the Lorentz fitting objective function with prior constraints, specifically as follows:

[0115] The basic model for Lorentz fitting is:

[0116] ;

[0117] in, The signal is fitted by Lorentz. To fit the maximum intensity of the absorption peak, The sampling point number corresponding to the fitted center frequency. To fit the number of sampling points corresponding to the full width at half maximum (FWHM), Baseline data for the entire region;

[0118] The objective function with prior constraints is to minimize the sum of squares of the deviations between the sampled data and the fitted signal within the absorption interval, plus the weighted sum of squares of the deviations of the center frequency, half width at half maximum, and prior parameters.

[0119] The weights are all set to 10. 3 This is used to penalize fitted parameters that exceed the tolerance range;

[0120] An iterative algorithm is used to perform Lorentz fitting on the data in the absorption interval; specifically, the center frequency parameter is restricted to a variation tolerance range based on the center frequency parameter in the prior constraints, and the half-width at half-maximum (WHM) parameter is also restricted to a variation tolerance range based on the WHM parameter in the prior constraints.

[0121] The Levenberg-Marquardt (LM) iterative algorithm was used for fitting, and the fitting deviation was gradually reduced by continuously adjusting the fitting parameters (fitting absorption peak maximum intensity, fitting center frequency, fitting half width at half maximum).

[0122] After each iteration, it is determined whether the fitted center frequency and the fitted half-width are within the preset tolerance range. If they are outside the range, they are corrected to the tolerance boundary before the next iteration is performed.

[0123] For example: in the initial parameters of the iteration, the maximum intensity of the fitted absorption peak is 0.5V, the fitted center frequency is 1570.1nm (exceeding the upper tolerance limit of 1571.57nm, so it is corrected to 1571.57nm), and the fitted half-width is 0.012nm (exceeding the upper tolerance limit of 0.011nm, so it is corrected to 0.011nm), and then the next iteration is performed;

[0124] The iteration stops when the fitted residual converges to a preset convergence threshold, yielding the Lorentz fitted signal data, specifically:

[0125] The preset convergence threshold is 10. -6 -10 -5 The criterion for judging the convergence of the iteration is that the ratio of the fitting residuals of two adjacent iterations is less than the threshold (the residual is the deviation between the sampled data in the absorption interval and the fitted signal).

[0126] After stopping the iteration, the index of each sampling point in the absorption interval is substituted into the fitted Lorentz model to obtain the Lorentz fitted signal value of each sampling point, which constitutes the Lorentz fitted signal data.

[0127] For example: when the convergence threshold is 5 × 10 -6 After the 20th iteration, the ratio of the residuals between two consecutive iterations is 3.2 × 10⁻⁶. -6 If the value is less than the convergence threshold, stop the iteration and output the Lorentz-fitted signal data.

[0128] In this implementation scheme, the center frequency parameter and full width at half maximum (FWHM) parameter obtained by fitting a known standard gas are set as prior constraints and a reasonable tolerance range is specified. This allows the Lorentz fitting process to search for the optimal parameters within a reliable interval, which not only allows the line shape to undergo slight adaptive adjustments according to actual working conditions, but also eliminates the possibility of the fitting parameters deviating from the physical reality due to noise interference. The iterative algorithm combined with the boundary correction mechanism further improves the stability of the fitting convergence, avoiding common problems such as divergence or getting trapped in local optima when measuring low concentrations. The final output Lorentz fitting signal data closely follows the true absorption profile.

[0129] Specifically, the steps for phase extrapolation compensation of the interference fringe residuals in the absorption region using the local phase information of the interference fringes extracted from the non-absorption region are as follows:

[0130] The local phase information of the interference fringes in the non-absorption region is extrapolated to the absorption region to obtain the predicted residual oscillation phase in the absorption region, which is as follows:

[0131] Phase extrapolation is performed using a linear extrapolation method. The phase sequences of the left and right non-absorption intervals are used to extend linearly to both sides of the absorption interval. The phase extrapolation value of the left non-absorption interval is used to the left of the midpoint of the absorption interval, and the phase extrapolation value of the right non-absorption interval is used to the right of the midpoint.

[0132] When extrapolating, the phase change trend should be kept consistent with the non-absorption region;

[0133] For example: the phases of the last two points in the non-absorbing interval on the left are 7π / 4 and 2π, and the starting number of the absorbing interval is 320. Then the extrapolated phase corresponding to the starting number is 9π / 4, and the extrapolated phase corresponding to the next number is 5π / 2.

[0134] Based on the predicted residual oscillation phase within the absorption interval, the predicted residual of the interference fringes within the absorption interval is generated, specifically as follows:

[0135] The model for predicting residuals from interference fringes is as follows:

[0136] ;

[0137] in, The residual amplitude is determined by the mean amplitude of the interference fringes in the non-absorption interval (i.e., the average of the absolute values ​​of the differences between the data at each sampling point in the non-absorption interval and the baseline value). The angular frequency of the interference fringes in the non-absorption region. The extrapolated predicted residual oscillation phase;

[0138] For example: the average amplitude of the interference fringes in the non-absorption region is 0.12V, the angular frequency is π / 8, and the extrapolated phase of a certain sampling point is 9π / 4. Then the estimated residual of that sampling point is approximately 0.085V.

[0139] Subtracting the predicted residual of the interference fringes within the absorption interval from the initial absorption signal yields the compensated absorption signal, which is as follows:

[0140] The preliminary absorption signal is the difference between the Lorentz fitting signal and the baseline line data within the absorption interval. By subtracting the estimated residual of the corresponding sampling point from the preliminary absorption signal, the interference fringe residual caused by phase mismatch within the absorption interval can be eliminated, and a pure gas absorption signal can be obtained.

[0141] For example: if the Lorentz fitting signal at a sampling point in a certain absorption interval is 2.9V, the baseline value is 2.6V, and the estimated residual is 0.085V, then the initial absorption signal is 0.3V, and the compensated absorption signal is 0.215V.

[0142] In this implementation, the local phase information of the interference fringes extracted from the non-absorption region is linearly extrapolated to the absorption region to predict the oscillation phase of the interference fringe residuals in the absorption region. Based on this, an interference fringe prediction residual that matches the real-time background is generated. After subtracting the prediction residual from the preliminary absorption signal, the residual oscillation component caused by the phase mismatch between the background template and the current measurement signal due to temperature fluctuations is effectively eliminated. This method no longer relies on a fixed background subtraction template, but uses the phase state captured in real time in the non-absorption region to dynamically correct the absorption region, so that the output absorption signal more purely reflects the true absorption profile of the target gas.

[0143] Specifically, the steps for integrating the compensated absorption signal to obtain the spectral absorption area are as follows:

[0144] The lower and upper limits of integration of the compensated absorbed signal within the absorption interval are determined as follows:

[0145] The lower limit of integration corresponds to the starting sampling point number of the absorption interval, and the upper limit of integration corresponds to the ending sampling point number of the absorption interval.

[0146] The integration frequency range is consistent with the absorption range, ensuring that the integration range completely covers the gas absorption signal and does not include noise components in the non-absorption range.

[0147] For example, if the absorption interval index ranges from 320 to 580, then the lower limit of integration is 320 and the upper limit of integration is 580, corresponding to a frequency range of approximately 31.25kHz to 55.66kHz.

[0148] Numerical integration is performed on the compensated absorption signal within the interval from the lower limit to the upper limit of integration, specifically as follows:

[0149] Numerical integration is performed using the trapezoidal integral method, with the integration step size being the frequency domain resolution (i.e., the sampling frequency divided by the total number of sampling points). The integration process involves averaging the compensated absorption signals of two adjacent sampling points, multiplying them by the integration step size, and finally summing the calculation results of all adjacent sampling points to obtain the spectral absorption area.

[0150] The trapezoidal integral method is simple to calculate and has high accuracy, and can accurately reflect the integral area of ​​the absorption signal (positively correlated with gas concentration).

[0151] For example: the sampling frequency is 200kHz, the total number of sampling points is 2048, the integration step size is about 97.66Hz, there are 261 sampling points in the integration interval, and the spectral absorption area is about 12.3V·Hz calculated by trapezoidal integration.

[0152] The spectral absorption area of ​​the gas at the concentration to be measured is obtained, specifically as follows:

[0153] The result of the above numerical integration is taken as the spectral absorption area of ​​the gas at the concentration to be measured, and 6 significant figures are retained.

[0154] If the integral result is negative (due to a small deviation in the baseline fitting), its absolute value is taken as the final absorption area.

[0155] For example, if the integral result is 12.3056 V·Hz, then the output spectral absorption area is 12.3056 V·Hz; if the integral result is -12.3056 V·Hz, then the absolute value is taken and the output is 12.3056 V·Hz.

[0156] In this implementation scheme, the upper and lower limits of integration strictly correspond to the start and end boundaries of the absorption interval, ensuring that the numerical integration range only covers the area where the gas absorption signal is located and excludes noise in the non-absorption interval, thus avoiding interference from irrelevant components on the area calculation results. The trapezoidal integration method is used in conjunction with the frequency domain resolution as the step size for calculation, which can accurately restore the total area corresponding to the absorption profile while keeping the calculation process simple. The absolute value processing is set for possible small negative results, which enhances the stability of the area output. The spectral absorption area obtained thus closely reflects the actual absorption intensity of the target gas.

[0157] Specifically, the steps for calculating the concentration of the gas to be measured based on the spectral absorption area are as follows:

[0158] The calibration correspondence between known standard gas concentrations and spectral absorption areas is retrieved, specifically as follows:

[0159] The calibration correspondence is a linear calibration curve obtained in advance through experiments. In the calibration experiment, 3-5 standard gases with known concentrations (the concentration range is consistent with the concentration range of the gas to be measured) are selected, and their spectral absorption areas are measured using this method to obtain multiple sets of calibration data (standard gas concentration and corresponding spectral absorption area).

[0160] The calibration correspondence formula is obtained through linear fitting:

[0161] ;

[0162] Where u is the calibration coefficient and b is the intercept;

[0163] For example, three standard gases were selected, and the calibration data are shown in Table 1.

[0164] Table 1 Examples of 3 Standard Gases

[0165] Standard gas concentration (ppm) Spectral absorption area (V·Hz) 100 5.12 200 10.25 300 15.38

[0166] The calibration coefficients obtained through linear fitting are 0.0196 (ppm·Hz / V) and the intercept is 0.08 (ppm). Therefore, the calibration correspondence is as follows: ;

[0167] Substituting the spectral absorption area of ​​the gas to be measured into the known calibration correspondence between the concentration of the standard gas and its spectral absorption area, the initial concentration value of the gas to be measured is calculated, specifically as follows:

[0168] Substitute the spectral absorption area of ​​the gas to be measured into the calibration formula above to calculate the initial concentration value, and retain two decimal places;

[0169] For example, if the spectral absorption area of ​​the gas to be measured is 12.3056 V·Hz, substituting it into the calibration formula, the initial concentration value is approximately 0.32 ppm.

[0170] The initial concentration value of the gas to be measured is corrected for systematic error, and the final concentration of the gas to be measured is output, specifically as follows:

[0171] The system error mainly originates from optical loss and detector noise in the TDLAS system, and the correction formula is:

[0172] ;

[0173] in, The system error correction coefficient (ranging from -0.05 to 0.05) is determined through calibration experiments (by selecting a standard gas of known concentration, calculating the deviation between the measured value and the true value, and obtaining the correction coefficient).

[0174] For example, if the systematic error correction factor is 0.02 and the initial concentration is 0.32 ppm, then the final concentration will be approximately 0.33 ppm (rounded to two decimal places).

[0175] In this implementation scheme, a linear calibration correspondence between known standard gas concentrations and spectral absorption areas is established in advance. The spectral absorption area of ​​the gas to be measured is substituted into this correspondence to obtain an initial concentration value. The initial concentration is then compensated based on a system error correction coefficient, resulting in a more accurate output of the gas concentration to be measured. This process directly maps the intermediate physical quantity of spectral absorption area to the dimension of concentration, eliminating the accumulation of deviations caused by system response fluctuations between multiple measurements. The calibration curve is established based on the measured data of multiple sets of standard gases within the same concentration range. The error correction stage compensates for inherent interferences such as optical loss and detector background, ensuring that the concentration calculation results can truly reflect the actual content of the target gas.

[0176] Please see Figure 7This invention provides a technical solution: a high-precision spectral noise cancellation system based on TDLAS technology, comprising: a model parameter acquisition unit, used to acquire sampling data of known standard gas concentrations and perform Lorentz model fitting to obtain center frequency parameters and full width at half maximum (FWHM) parameters; an adaptive filtering unit, used to acquire sampling data of the gas concentration to be measured, determine an adaptive cutoff frequency based on the center frequency parameters and FWHM parameters, and perform digital low-pass filtering to obtain filtered sampling data of the gas concentration to be measured; an interval division unit, used to divide the filtered sampling data of the gas concentration to be measured into absorption intervals and non-absorption intervals located on both sides of the absorption intervals; and a baseline fitting unit, used to extract the baseline within the non-absorption intervals. The local phase information of the interference fringes is used to perform a weighted third-order polynomial fitting on the data in the non-absorption region to obtain the baseline line shape data of the entire region. The absorption fitting compensation unit is used to perform Lorentz fitting on the data in the absorption region under the constraint of the center frequency parameter and the full width at half maximum parameter to obtain the Lorentz fitting signal data. The baseline line shape data of the entire region and the Lorentz fitting signal data are subtracted in the absorption region to obtain the preliminary absorption signal. The local phase information of the interference fringes extracted in the non-absorption region is used to perform phase extrapolation compensation on the interference fringe residuals in the absorption region. The compensated absorption signal is integrated to obtain the spectral absorption area. The concentration of the gas to be measured is calculated based on the spectral absorption area.

[0177] Each functional unit can be implemented using an FPGA chip. The clock frequency of the FPGA chip is 200MHz, which can meet the real-time operation requirements of each algorithm step and ensure that the response time of the entire system does not exceed 100ms.

[0178] The model parameter acquisition unit also includes a fitting verification module, used to verify the fitting results of the Lorentz model. When the fitting residual is greater than 10... -6 Then, refit the data to ensure the accuracy of the center frequency parameter and the full width at half maximum (FWHM) parameter.

[0179] The adaptive filtering unit includes a Fourier transform module and an FIR filtering module, which respectively implement frequency domain transformation and low-pass filtering of the sampled data;

[0180] The interval division unit includes a threshold recognition module, which is used to automatically identify the start and end positions of gas absorption characteristics;

[0181] The baseline fitting unit includes a phase extraction module and a weighted fitting module, which respectively realize local phase extraction of interference fringes and third-order polynomial weighted fitting.

[0182] The absorption fitting compensation unit includes a Lorentz fitting module, a residual compensation module, and an integral calculation module, which respectively realize constrained Lorentz fitting, interference fringe residual compensation, and spectral absorption area calculation.

[0183] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0184] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A high-precision spectral noise reduction method based on TDLAS technology, characterized in that, Includes the following steps: Sampling data of known standard gas concentrations were obtained and fitted using the Lorentz model to obtain the center frequency parameter and the full width at half maximum (FWHM) parameter. The sampling data of the gas concentration to be measured is acquired, and the adaptive cutoff frequency is determined based on the center frequency parameter and the half width at half maximum parameter and digital low-pass filtering is performed to obtain the filtered sampling data of the gas concentration to be measured. The filtered sampling data of the gas concentration to be measured is divided into an absorption interval and non-absorption intervals located on both sides of the absorption interval; Local phase information of interference fringes in the non-absorption region is extracted, the phase change rate is calculated based on the local phase information, and the fitting weight coefficients of each sampling point in the non-absorption region are generated according to the phase change rate. The data in the non-absorption region are fitted with a weighted third-order polynomial using fitting weight coefficients, and the fitted polynomial curve is extrapolated to the entire region to obtain the baseline data of the entire region. The center frequency parameter and the full width at half maximum (FWHM) parameter are used as prior constraints for fitting. The variation tolerance ranges of the center frequency parameter and the FWHM parameter are set. Under the prior constraints for fitting, Lorentz fitting is performed on the data in the absorption interval to obtain Lorentz fitted signal data. Within the absorption range, the Lorentz-fitted signal data is interpolated with the baseline linear data of the entire region to obtain the preliminary absorption signal; The local phase information of the interference fringes in the non-absorption interval is extrapolated to the absorption interval to obtain the estimated residual oscillation phase in the absorption interval. Based on the estimated residual oscillation phase and the estimated residual amplitude determined by the amplitude level of the interference fringes in the non-absorption interval, the estimated residual of the interference fringes in the absorption interval is generated. The estimated residual of the interference fringes is then subtracted from the preliminary absorption signal to obtain the compensated absorption signal. The spectral absorption area is obtained by integrating the compensated absorption signal, and the concentration of the gas to be measured is calculated based on the spectral absorption area.

2. The high-precision spectral noise reduction method based on TDLAS technology according to claim 1, characterized in that, The specific steps for determining the adaptive cutoff frequency and performing digital low-pass filtering based on the center frequency parameter and the full width at half maximum (FWHM) parameter are as follows: Perform a Fourier transform on the sampling data of the gas to be measured to obtain the spectral distribution of the sampling data of the gas to be measured. The theoretical upper limit of the bandwidth of the gas absorption signal in the frequency domain is calculated based on the center frequency parameter and the full width at half maximum (FWHM) parameter. The adaptive cutoff frequency is obtained by extending the theoretical upper limit of the gas absorption signal in the frequency domain outward by a preset protection bandwidth ratio. The sampling data of the gas concentration to be measured is digitally low-pass filtered using an adaptive cutoff frequency to obtain the filtered sampling data of the gas concentration to be measured.

3. The high-precision spectral noise reduction method based on TDLAS technology according to claim 1, characterized in that, The specific steps for dividing the filtered sampling data of the gas concentration to be measured into absorption intervals and non-absorption intervals located on both sides of the absorption interval are as follows: The starting and ending positions of gas absorption characteristics in the sampling data of the gas concentration to be measured after localization filtering; The interval from the start position to the end position of the gas absorption characteristics is defined as the absorption interval; The continuous interval to the left of the absorption interval without gas absorption characteristics is defined as the non-absorption interval; The continuous interval to the right of the absorption interval that has no gas absorption characteristics is defined as the non-absorption interval.

4. The high-precision spectral noise reduction method based on TDLAS technology according to claim 1, characterized in that, The specific steps for extracting the local phase information of interference fringes in the non-absorption region are as follows: Autocorrelation calculations were performed on the data in the non-absorption region to obtain the positions of the correlation peaks of the interference fringes in the non-absorption region. Based on the position of the correlation peak of the interference fringes in the non-absorption interval, calculate the phase value of the interference fringes at each sampling point in the non-absorption interval; The phase values ​​of the interference fringes at each sampling point in the non-absorption interval are smoothed to remove abnormal phase points; Output the local phase information of the interference fringes in the non-absorption region.

5. The high-precision spectral noise reduction method based on TDLAS technology according to claim 1, characterized in that, The specific steps for generating the fitting weight coefficients for each sampling point within the non-absorption interval based on the phase change rate are as follows: The absolute value of the phase difference between adjacent sampling points is calculated as the phase change rate. Based on the inverse relationship between the phase change rate and the fitting weight coefficient, sampling points with smaller phase change rates are assigned higher fitting weight coefficients, while sampling points with larger phase change rates are assigned lower fitting weight coefficients.

6. The high-precision spectral noise reduction method based on TDLAS technology according to claim 1, characterized in that, The specific steps for Lorentz fitting of the data in the absorption interval under the fitting prior constraints are as follows: Construct a Lorentz fitting objective function with a center frequency parameter tolerance term and a half-width parameter tolerance term; An iterative algorithm is used to fit the data within the absorption interval; The iteration stops when the fitting residual converges to the preset convergence threshold, and the Lorentz-fitted signal data is obtained.

7. The high-precision spectral noise reduction method based on TDLAS technology according to claim 1, characterized in that, Based on the predicted residual oscillation phase and the predicted residual amplitude determined by the amplitude level of the interference fringes in the non-absorption interval, the specific steps for generating the predicted residual of the interference fringes in the absorption interval are as follows: Calculate the absolute value of the difference between the data at each sampling point within the non-absorption interval and the corresponding baseline value; The average of the absolute values ​​of each difference is used to obtain the mean amplitude of the interference fringes in the non-absorption region; The mean amplitude is used as the estimated residual amplitude within the absorption interval; The residual is estimated by generating interference fringes within the absorption range based on the estimated residual amplitude and the estimated residual oscillation phase.

8. The high-precision spectral noise reduction method based on TDLAS technology according to claim 1, characterized in that, The specific steps for integrating the compensated absorption signal to obtain the spectral absorption area are as follows: Determine the lower and upper limits of integration of the compensated absorbed signal within the absorption interval; Numerical integration is performed on the compensated absorption signal within the interval from the lower limit of integration to the upper limit of integration; The spectral absorption area of ​​the gas at the concentration to be measured is obtained.

9. The high-precision spectral noise reduction method based on TDLAS technology according to claim 1, characterized in that, The specific steps for calculating the concentration of the gas to be measured based on the spectral absorption area are as follows: Retrieve the calibration correspondence between known standard gas concentrations and spectral absorption areas; Substitute the spectral absorption area of ​​the gas to be measured into the calibration correspondence to calculate the initial concentration value of the gas to be measured. The initial concentration value of the gas to be measured is corrected for systematic error, and the final concentration of the gas to be measured is output.

10. A high-precision spectral noise cancellation system based on TDLAS technology, employing the high-precision spectral noise cancellation method based on TDLAS technology as described in any one of claims 1-9, characterized in that, include: The model parameter acquisition unit is used to acquire sampling data of known standard gas concentrations and perform Lorentz model fitting to obtain the center frequency parameter and the full width at half maximum (FWHM) parameter. An adaptive filtering unit is used to acquire sampling data of the gas concentration to be measured. Based on the center frequency parameter and the full width at half maximum parameter, an adaptive cutoff frequency is determined and digital low-pass filtering is performed to obtain the filtered sampling data of the gas concentration to be measured. The interval division unit is used to divide the filtered sampling data of the gas concentration to be measured into an absorption interval and non-absorption intervals located on both sides of the absorption interval; The baseline fitting unit is used to extract the local phase information of the interference fringes in the non-absorption region, calculate the phase change rate based on the local phase information, generate the fitting weight coefficients of each sampling point in the non-absorption region according to the phase change rate, use the fitting weight coefficients to perform weighted third-order polynomial fitting on the data in the non-absorption region, and extrapolate the fitted polynomial curve to the whole region to obtain the baseline line data of the whole region. The absorption fitting compensation unit is used to use the center frequency parameter and the full width at half maximum (FWHM) parameter as fitting prior constraints, set the variation tolerance range of the center frequency parameter and the variation tolerance range of the FWHM parameter, and perform Lorentz fitting on the data in the absorption interval under the fitting prior constraints to obtain Lorentz fitted signal data; perform difference processing on the Lorentz fitted signal data and the baseline line shape data of the entire region in the absorption interval to obtain the preliminary absorption signal; extrapolate the local phase information of the interference fringes in the non-absorption interval to the absorption interval to obtain the estimated residual oscillation phase in the absorption interval; generate the estimated residual of the interference fringes in the absorption interval based on the estimated residual oscillation phase and the estimated residual amplitude determined by the amplitude level of the interference fringes in the non-absorption interval; and subtract the estimated residual of the interference fringes from the preliminary absorption signal to obtain the compensated absorption signal. The spectral absorption area is obtained by integrating the compensated absorption signal, and the concentration of the gas to be measured is calculated based on the spectral absorption area.

Images