A filtering method, device and system for TDLAS-WMS signals

Abstract

A filtering method, apparatus, and system for TDLAS-WMS signals are disclosed, relating to the field of spectral signal processing technology. The aim is to address the shortcomings of existing TDLAS-WMS signal filtering techniques, which cannot simultaneously address noise smoothing, baseline drift suppression, and non-stationary spike noise suppression. The method includes adaptive wavelet threshold denoising of the original TDLAS-WMS second harmonic signal to obtain a denoised signal; SG smoothing filtering of the denoised signal to obtain a filtered signal; extraction of the real-time baseline from the filtered signal; and dynamic subtraction of the real-time baseline from the filtered signal. In the system, a signal acquisition module acquires the original TDLAS-WMS second harmonic signal, which is then acquired by a data acquisition card and preprocessed by a signal preprocessing module. A filtering module for TDLAS-WMS signals performs filtering on the preprocessed signal according to the above method. The processing result is output and displayed by a data output and display module.

Description

Technical Field

[0001] This invention relates to the field of spectral signal processing technology, and in particular to filtering technology for TDLAS-WMS signals. Background Technology

[0002] The TDLAS-WMS system acquires the second harmonic signal of gas absorption spectra through wavelength modulation. This second harmonic signal can be used to retrieve the concentration of gaseous components such as NO and CO2. The TDLAS-WMS second harmonic signal (hereinafter referred to as the second harmonic signal) is highly susceptible to signal distortion under conditions of strong vibration, high dust, and high temperature. Therefore, the acquired second harmonic signal needs to be filtered to obtain data that more accurately reflects the characteristics of the second harmonic signal.

[0003] Conventional methods for processing second harmonic signals include single or simple combinations of SG smoothing filtering, wavelet denoising, and fixed baseline correction, but these methods all have certain drawbacks. For example, while SG filtering with fixed parameters can smooth noise, it cannot suppress baseline drift and has poor suppression capabilities for non-stationary spike noise. The denoising + fixed baseline correction method uses a baseline obtained by polynomial static fitting, which cannot adapt to dynamic real-time drift caused by environmental factors such as temperature and vibration. Adaptive denoising, without synergy with peak-preserving SG smoothing, easily leads to distortion of absorption peak amplitude, affecting the accuracy of concentration measurement.

[0004] In summary, existing TDLAS-WMS signal filtering techniques cannot simultaneously address issues such as noise smoothing, baseline drift suppression, and non-stationary spike noise suppression, resulting in unsatisfactory filtering effects and impacting the accuracy and stability of gas concentration measurements. Summary of the Invention

[0005] The purpose of this invention is to solve or alleviate the above problems by providing a filtering method, apparatus and system for TDLAS-WMS signals.

[0006] A filtering method for TDLAS-WMS signals according to the present invention includes: performing adaptive wavelet threshold denoising on the original TDLAS-WMS second harmonic signal to obtain a denoised signal; performing SG smoothing filtering on the denoised signal to obtain a filtered signal; extracting a real-time baseline from the filtered signal; and dynamically subtracting the real-time baseline from the filtered signal.

[0007] Optionally, the adaptive wavelet thresholding denoising process for the original TDLAS-WMS second harmonic signal includes: performing wavelet decomposition on the original WMS second harmonic signal; adaptively calculating the threshold based on the signal energy, and removing non-stationary spikes and high-frequency random noise based on the threshold; and reconstructing the signal to retain the effective absorption peak profile.

[0008] Optionally, the window used for the SG smoothing filter is a second harmonic adaptive window.

[0009] Optionally, the SG smoothing filter is fitted using a second- or third-order polynomial.

[0010] Optionally, an iterative weighted least squares method combined with a Whittaker function is used to extract the real-time baseline from the filtered signal.

[0011] Optionally, extracting the real-time baseline from the filtered signal includes: calculating the local noise variance of the filtered signal; determining initial weights based on the local noise variance and performing weighted Whittaker smoothing to obtain the initial baseline and residuals; extracting candidate peak regions from the residuals; calculating the confidence level of each candidate peak region and constructing a peak region mask; using a Tukey double weighting function combined with the peak region mask and peak region boundaries for weight transition; iteratively updating the weights and refitting the baseline to obtain the real-time baseline.

[0012] Optionally, the SG smoothing filtering of the denoised signal includes: real-time scanning of the denoised signal to identify two types of regions: peak region and baseline non-absorbing region; reducing the SG smoothing window and decreasing the polynomial order in the peak region to perform SG smoothing filtering, increasing the SG window and increasing the polynomial order in the baseline non-absorbing region to perform SG smoothing filtering; and adjusting the threshold of the adaptive wavelet threshold denoising process in real time using the noise statistical variance of the non-absorbing peak region.

[0013] The present invention also provides a filtering module for TDLAS-WMS signals, comprising: a noise reduction unit adapted to perform adaptive wavelet threshold noise reduction processing on the original TDLAS-WMS second harmonic signal to obtain a noise-reduced signal; a filtering unit adapted to perform SG smoothing filtering on the noise-reduced signal to obtain a filtered signal; a real-time baseline extraction unit adapted to extract a real-time baseline from the filtered signal; and a baseline subtraction unit adapted to dynamically subtract the real-time baseline from the filtered signal.

[0014] The present invention also provides a filtering system for TDLAS-WMS signals based on the above-mentioned modules, comprising: a signal acquisition module for acquiring the original TDLAS-WMS second harmonic signal; a data acquisition card connected to the signal acquisition module for sending the acquired signal to a signal preprocessing module; a signal preprocessing module connected to the data acquisition card for preprocessing the signal from the data acquisition card; a filtering module for TDLAS-WMS signals connected to the signal preprocessing module for filtering the preprocessed signal; a control and configuration module connected to the filtering module for TDLAS-WMS signals for adaptively performing dynamic threshold adjustment for the filtering module for TDLAS-WMS signals; and a data output and display module connected to the filtering module for TDLAS-WMS signals for displaying the data output by the filtering module for TDLAS-WMS signals.

[0015] The present invention also provides an electronic device, comprising: a memory; a processor; and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the filtering method for TDLAS-WMS signals of the present invention.

[0016] This invention achieves peak-preserving, adaptive, and real-time baseline correction filtering of TDLAS-WMS second harmonics by cascading three steps—adaptive wavelet threshold denoising, SG smoothing filtering, and baseline correction—in a specific order. This results in high signal-to-noise ratio, low distortion, and zero drift processing of weakly absorbed signals, thereby improving the accuracy and stability of gas concentration measurement. Attached Figure Description

[0017] Figure 1 This is a schematic flowchart of a filtering method for TDLAS-WMS signals according to an embodiment of this application;

[0018] Figure 2 These are the input and output waveforms of a filtering method for TDLAS-WMS signals according to embodiments of this application;

[0019] Figure 3 This is a schematic diagram of a filtering system for TDLAS-WMS signals according to an embodiment of this application;

[0020] Figure 4 This is a schematic diagram of the structure of a filtering module for TDLAS-WMS signals according to an embodiment of this application;

[0021] Figure 5 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0022] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.

[0023] As used herein, the singular forms “a,” “an,” and “the” are intended to include the plural forms as well, unless the context indicates otherwise. It should be further understood that the terms “comprising,” “including,” indicate the presence of the stated feature, operation, element, component, item, kind, and / or group, but do not preclude the presence, occurrence, or addition of one or more other features, operations, elements, components, items, kinds, and / or groups. The terms “or” and “and / or” as used herein are interpreted as inclusive, or mean any one or any combination thereof. Thus, “A, B, or C” or “A, B, and / or C” means “any one of: A; B; C; A and B; A and C; B and C; A, B, and C.” Exceptions to this definition occur only when combinations of elements, functions, or operations are inherently mutually exclusive in some manner.

[0024] To address the problem that existing TDLAS-WMS signal filtering techniques cannot simultaneously achieve noise smoothing, baseline drift suppression, and non-stationary spike noise suppression, resulting in unsatisfactory filtering effects and affecting the accuracy of gas concentration measurement, this invention provides a filtering method for TDLAS-WMS signals. This method can effectively solve the signal distortion problem of TDLAS-WMS under conditions of strong vibration, high dust, and high temperature, and achieve high signal-to-noise ratio, low distortion, and zero baseline drift processing for weak absorption signals.

[0025] A filtering method for TDLAS-WMS signals according to an embodiment of this application includes: performing adaptive wavelet threshold denoising on the original TDLAS-WMS second harmonic signal to obtain a denoised signal; performing SG smoothing filtering on the denoised signal to obtain a filtered signal; extracting a real-time baseline from the filtered signal; and dynamically subtracting the real-time baseline from the filtered signal.

[0026] Figure 1 This is a schematic flowchart illustrating a filtering method for TDLAS-WMS signals according to an embodiment of this application. Figure 1 As shown, a filtering method for TDLAS-WMS signals according to an embodiment of this application begins with step S1.

[0027] In step S1, the original TDLAS-WMS second harmonic signal is subjected to adaptive wavelet threshold denoising to obtain a denoised signal.

[0028] This step specifically includes steps S11 to S13.

[0029] Step S11: Perform wavelet decomposition on the original WMS second harmonic signal. Preferably, use the db4 wavelet and decompose the signal into 3 to 4 layers.

[0030] Step S12: For the signal obtained by wavelet decomposition, the threshold is adaptively calculated according to the signal energy, and non-stationary spikes and high-frequency random noise are removed according to the threshold.

[0031] Step S13: Reconstruct the signal after removing non-stationary spikes and high-frequency random noise to obtain a denoised signal. The denoised signal retains the effective absorption peak profile.

[0032] Next, in step S2, the noise-reduced signal is subjected to SG smoothing filtering to obtain the filtered signal.

[0033] In this embodiment, a second-harmonic adaptive window is used for local least-squares fitting smoothing during the SG smoothing process. For the TDLAS-WMS second-harmonic signal, the window length of the second-harmonic adaptive window is 15 to 25 points, and the fitting polynomial is of order 2 to 3. The above SG smoothing filtering process does not weaken the signal peak, does not change the peak shape of the signal wave, and does not produce phase shift. The resulting filtered signal can be considered a distortion-free signal.

[0034] Next, in step S3, a real-time baseline is extracted from the filtered signal.

[0035] In one implementation, this step uses an iterative weighted least squares method combined with a Whittaker function to smoothly extract the real-time baseline from the filtered signal obtained in the previous step.

[0036] Next, in step S4, the real-time baseline is dynamically subtracted from the filtered signal.

[0037] This step dynamically subtracts the real-time baseline extracted in step S3 from the smoothing filtering result of step S2 to eliminate slow drift caused by factors such as temperature, vibration, and optical contamination, and finally outputs a standard drift-free second harmonic signal for concentration inversion.

[0038] The filtering method for TDLAS-WMS signals in this application is not a simple cascade of adaptive wavelet threshold denoising, SG smoothing filtering, and dynamic baseline correction (steps S3 and S4 are collectively referred to as dynamic baseline correction). Instead, it is a processing logic designed specifically for the interference physical characteristics of the WMS second harmonic signal. Wavelet denoising targets non-stationary spike noise (dust particles, vibration shock), SG smoothing filtering targets high-frequency random noise and needs to preserve peak shape, and dynamic baseline correction targets low-frequency slow drift (temperature, optical pollution). If SG smoothing is performed before denoising, the non-stationary spike noise will expand into multiple outliers, affecting the denoising effect; if baseline correction is performed before denoising, the baseline estimation will be severely affected by noise interference, introducing additional errors.

[0039] The following is a verification experiment to demonstrate the filtering effect of the filtering method for TDLAS-WMS signals according to the embodiments of this application. During the detection of NO gas concentration in the flue gas of a gas turbine, the original WMS second harmonic signal collected on-site has the following waveform: Figure 2 As shown, the signal length is 1024 points, and the sampling rate is 1 MS / s. By implanting sensors in the flue, it can be determined that the original WMS second harmonic signal contains high-frequency random noise, non-stationary spike interference, and low-frequency baseline drift. Among them, the amplitude of the high-frequency random noise is ±0.05au; the number of non-stationary spike interference is 3, the amplitude of which is 2 to 3 times that of the normal signal, and the width is 2 to 3 points; the baseline drift is a sinusoidal drift with a frequency of 0.2Hz and an amplitude of 15% of the absorption peak height.

[0040] Four different methods were used to filter this original WMS second harmonic signal. The processing order of the four methods is shown in Table 1. Among them, method A is the filtering method for TDLAS-WMS signals according to an embodiment of this application.

[0041] Table 1. Signal processing order of the four methods

[0042]

[0043] The filtering results of the four methods are shown in Table 2. It is evident that changing the order of the three steps—denoising, smoothing filtering, and baseline correction—affects the filtering effect to varying degrees, especially in method B, where smoothing filtering first directly leads to denoising failure. The input and output waveforms of method A are shown below. Figure 2 As shown.

[0044] The results shown in Table 2 are statistically analyzed, and the results are shown in Table 3. It can be seen that Method A shows significant advantages in terms of signal-to-noise ratio (SNR), peak error, and baseline residual.

[0045] Table 2. Results of the four methods

[0046]

[0047] Table 3. Statistical analysis of the results from the four methods.

[0048]

[0049] Compared with conventional second harmonic signal processing methods, the filtering method for TDLAS-WMS signals in this application has the following advantages:

[0050] 1. Strong anti-interference capability: It can simultaneously suppress high-frequency noise, spike interference and low-frequency drift, making it suitable for harsh industrial environments;

[0051] 2. High signal fidelity: SG smoothing filter can completely preserve the amplitude and shape of the absorption peak, so that the concentration inversion error is ≤3%FS;

[0052] 3. Strong dynamic adaptability: It can achieve real-time baseline correction without manual calibration, and the long-term stability is greatly improved;

[0053] 4. High robustness: The WMS second harmonic design, designed for high-interference scenarios such as gas turbines and flue gas, performs stably in various harsh environments;

[0054] 5. Lightweight Algorithm: This method is implemented using a lightweight computer program, which can run in real time within an FPGA / microcontroller, meeting the requirements for online detection.

[0055] To address the problems of existing baseline correction methods, such as difficulty in effectively distinguishing residual interference and absorption peaks after cascaded preprocessing and uncontrollable iterative convergence, this application provides the following preferred embodiments. As a preferred embodiment of this application, step S3 includes steps S31 to S33.

[0056] Step S31: Calculate the local noise variance of the filtered signal, determine the initial weights based on the local noise variance, and perform weighted Whittaker smoothing to obtain the initial baseline and residuals;

[0057] Step S32: Extract candidate peak regions from the residuals, calculate the confidence level of each candidate peak region, and construct a peak region mask;

[0058] Step S33: Use Tukey's dual weighting function and combine it with peak region mask and peak region boundary to perform weight transition, iteratively update the weights and refit the baseline to obtain the real-time baseline.

[0059] For the scenario of three specific sequential cascaded steps in the filtering method for TDLAS-WMS signals in the embodiments of this application, adaptive modeling is performed based on the statistical characteristics of the residuals. The model mainly includes two parts: a statistical model of the cascaded residuals and a scoring model. The statistical model of the cascaded residuals is used to drive the initial baseline estimation, and the scoring model is used to drive peak region identification.

[0060] In step S31, the local noise variance of the filtered signal is calculated. The statistical model of the cascaded residuals uses a sliding window to statistically analyze the local fluctuation level of the local noise variance segment by segment. Regions with large fluctuations indicate that there is still residual interference after the previous stage (step S1) processing, and the confidence level is low, so they should be given low weights. Regions with small fluctuations are clean, have high confidence, and are given high weights. This adaptive level of confidence guides Whittaker smoothing, fitting the signal in clean areas and relaxing the fit in suspicious areas, forming a safe initial baseline and avoiding residual interference being mistakenly treated as part of the baseline. Based on this principle, the embodiments of this application use the reciprocal of the local noise variance as the initial weight for weighted Whittaker smoothing to obtain the initial baseline and residuals (also known as cascaded residuals).

[0061] In step S32, candidate peak regions are extracted from the residuals. The three indices—the theoretical half-width at half-maximum (HWHM) determined by the WMS signal modulation depth, the symmetry determined by the Voigt line shape, and the degree of adjacent peak interference—are combined into a physical conformity scoring system for the scoring model. The width index has the highest weight because it has the most direct physical relationship with the modulation depth; symmetry is next, reflecting the physical constraints of the Voigt line shape; and the degree of adjacent peak interference, also known as isolation, has the lowest weight and can be used as an auxiliary criterion. Preferably, the combined weight of the three physical conformities is: theoretical HWHM 0.5, symmetry 0.3, and adjacent peak interference 0.2. The allocation of these three weights can be adjusted according to the actual application scenario. The scoring model outputs the confidence score for each absorption peak. Next, a mask is constructed based on the absorption peak confidence scores to constrain the baseline estimation, in order to distinguish residual interference from the true absorption peak.

[0062] In step S33, the weights of the three steps—driving initial baseline estimation, driving peak region identification, and mask-constrained baseline fine estimation—are iteratively updated and the baseline is refitted to achieve active peak region avoidance. During the iteration process, a Tukey dual-weight function combined with a mask and peak region boundaries is used to smoothly transition the weights. Specifically, a gradient band with a width of 3 to 5 points is set on each side of the peak region boundary to smoothly transition the weight from 1 to 0. This step in this embodiment uses a fixed-point iteration strategy, preferably with a fixed number of iterations or early convergence, outputting a drift-free signal. During the iterative weighted baseline fitting process, the absorption peak confidence map or mask determined by the physical model actively excludes regions identified as absorption peaks from the reference data for baseline fitting, so that the baseline estimation only depends on high-confidence non-absorbing region data. This scheme combines cascaded residual adaptive modeling, physical model-guided peak region identification, fixed-point iteration strategy, and active peak region avoidance to achieve accurate orthogonal separation of the baseline and absorption peaks.

[0063] The residual interference after wavelet denoising and SG smoothing has unique statistical characteristics (such as residual spikes being broadened into brief anomalous segments and high-frequency noise being transformed into colored noise). Conventional weighting functions (such as adaptive iterative reweighted penalized least squares) are not designed for this type of residual, making it difficult to effectively distinguish residual interference from useful signals and limiting the accuracy of baseline estimation. Pure mathematical peak-finding algorithms are prone to confusing residual spikes with real absorption peaks under strong industrial interference conditions, leading to missed or incorrect judgments. The variable number of iterations and the involvement of high-order matrix inversion result in excessive computation, making it difficult to meet real-time processing requirements on FPGA / microcontroller platforms. In addition, existing methods do not utilize the physical priors of WMS second harmonic signals in terms of modulation depth, linewidth, and peak shape symmetry, wasting the natural advantage of improving separation accuracy.

[0064] This preferred embodiment fully leverages the unique statistical characteristics of residual interference and the inherent advantages of the WMS second harmonic signal in terms of modulation depth, linewidth, and peak symmetry, thus achieving precise orthogonal separation of the baseline and absorption peak. Furthermore, this preferred embodiment employs a fixed-point iteration strategy, and the entire method can be implemented using lightweight code, meeting real-time processing requirements. On an FPGA, three iterations take less than 50 μs, fully satisfying the 1 MS / s real-time processing requirement.

[0065] To address the issues that fixed-parameter SG smoothing filters cannot simultaneously achieve conformal preservation of absorption peaks and denoising of non-absorption peaks, and that cascaded chains lack adaptive response to environmental noise, this application provides another preferred embodiment. This embodiment can be used in combination with the previous embodiment or implemented independently. As another preferred embodiment of this application, step S2 above includes steps S21 to S23.

[0066] In step S21, the noise reduction signal is scanned in real time to identify two types of regions: peak region and baseline non-absorption region.

[0067] The peak region refers to the absorption peak region, while the baseline non-absorption region is a flat background. The purpose of identifying these two types of regions is to differentiate them in subsequent steps based on their characteristics.

[0068] In step S22, the SG smoothing window is reduced and the polynomial order is decreased in the peak region to perform SG smoothing filtering, and the SG window is increased and the polynomial order is increased in the baseline non-absorption region to perform SG smoothing filtering.

[0069] Specifically, the SG smoothing window is narrowed to 7-11 points in the peak region, and the polynomial order is reduced to the second order to maximize the preservation of the original amplitude and peak shape of the absorption peak; in the baseline non-absorption region, the SG window is increased to 25-35 points, and the polynomial order is increased to the third order to fully suppress residual high-frequency noise.

[0070] In step S23, the threshold of the adaptive wavelet threshold denoising process is adjusted in real time using the noise statistical variance of the non-absorbing peak region. This step establishes an inter-stage adaptive feedback mechanism, which uses the noise statistical variance of the baseline non-absorbing region to fine-tune the threshold parameter of the previous wavelet denoising stage in real time. When the noise statistical variance increases, the threshold used in the adaptive wavelet threshold denoising process in step S1 is increased; when the noise statistical variance decreases, the threshold is decreased. This enables the entire cascade chain to automatically adapt to changes in environmental noise, achieving collaborative optimization of "peak region shape preservation, non-peak region denoising, and inter-stage linkage".

[0071] For TDLAS-WMS second harmonic signals, the characteristics of the absorption peak region and the non-absorption peak region are distinctly different. The absorption peak region needs to preserve the integrity of its amplitude and peak shape as much as possible, while the non-absorption peak region needs to suppress residual noise as much as possible. In existing TDLAS-WMS signal processing, SG smoothing filtering with a fixed window length and a fixed polynomial order is typically used, applying the same level of smoothing to the entire signal. However, fixed parameters cannot simultaneously meet the different needs of the absorption and non-absorption peak regions: a large window length flattens the peak shape, leading to errors in concentration inversion; a small window length results in insufficient smoothing of the noise region, limiting the improvement in signal-to-noise ratio. Therefore, this signal processing method cannot simultaneously achieve both "peak shape preservation" and "noise reduction." Furthermore, in existing TDLAS-WMS signal processing, the threshold for wavelet denoising in the preceding stage is usually a fixed value, unable to adaptively adjust according to real-time changes in signal quality. The entire cascaded chain lacks the ability to perceive and respond to changes in environmental noise.

[0072] This preferred embodiment upgrades fixed-parameter filtering to a partitioned processing strategy that automatically adapts to signal morphology characteristics. Simultaneously, it introduces inter-stage feedback to form a closed-loop adaptive adjustment, significantly reducing amplitude loss in the absorption peak region, effectively ensuring concentration inversion accuracy, and fully suppressing noise in the non-absorption peak region, further improving the overall signal-to-noise ratio. The entire algorithm can automatically adjust parameters according to changes in industrial noise levels without manual calibration, making it suitable for long-term online continuous operation under harsh conditions such as gas turbine flue gas.

[0073] This application also provides a filtering system for TDLAS-WMS signals. For example... Figure 3 As shown, the system includes: a signal acquisition module 1, a data acquisition card 2, a signal preprocessing module 3, a filtering module 4 for TDLAS-WMS signals, and a data output and display module 5.

[0074] The signal acquisition module 1 includes a laser driver and a photodetector. The output signals of the laser driver and photodetector are acquired by the data acquisition card 2. The acquired data is then processed by the signal preprocessing module 3, which sequentially performs DC component filtering and normalization. The preprocessed data enters the filtering module 4 for TDLAS-WMS signals. The filtering module 4 processes the input data according to the filtering method for TDLAS-WMS signals in the embodiments of this application. The processed waveform is output and displayed by the data output and display module 5. In addition, the data output and display module 5 can also perform concentration calculations based on the signal output by the filtering module 4 for TDLAS-WMS signals and output the calculation results.

[0075] like Figure 4 As shown, the filtering module 4 for TDLAS-WMS signals in the above system includes:

[0076] Noise reduction unit 41 is suitable for performing adaptive wavelet threshold noise reduction processing on the original TDLAS-WMS second harmonic signal to obtain a noise-reduced signal;

[0077] Filtering unit 42 is adapted to perform SG smoothing filtering on the noise reduction signal to obtain a filtered signal;

[0078] Real-time baseline extraction unit 43 is adapted to extract a real-time baseline from the filtered signal;

[0079] The baseline subtraction unit 44 is adapted to dynamically subtract the real-time baseline from the filtered signal.

[0080] In a preferred embodiment of this application, the noise reduction unit 41 includes:

[0081] Wavelet decomposition subunits are suitable for wavelet decomposition of the original WMS second harmonic signal;

[0082] The non-stationary spike and high-frequency random noise removal subunit is adapted to adaptively calculate the threshold based on the signal energy and remove non-stationary spikes and high-frequency random noise according to the threshold.

[0083] Re-cells are used to reconstruct the signal while preserving the effective absorption peak profile.

[0084] In a preferred embodiment of this application, the window used for the SG smoothing filter is a second harmonic adaptive window.

[0085] In a preferred embodiment of this application, the SG smoothing filter is fitted using a 2nd to 3rd order polynomial.

[0086] In a preferred embodiment of this application, the real-time baseline extraction unit 43 uses an iterative weighted least squares method combined with a Whittaker function to extract the real-time baseline from the filtered signal.

[0087] In a preferred embodiment of this application, the real-time baseline extraction unit 43 includes:

[0088] The initial baseline and residual calculation subunit is adapted to calculate the local noise variance of the filtered signal, determine the initial weights based on the local noise variance and perform weighted Whittaker smoothing to obtain the initial baseline and residuals;

[0089] A mask construction subunit is adapted to extract candidate peak regions from the residual, calculate the confidence level of each candidate peak region, and construct a peak region mask.

[0090] The iterative subunit is suitable for using the Tukey double weight function and combining peak region mask and peak region boundary for weight transition, iteratively updating the weights and refitting the baseline to obtain the real-time baseline.

[0091] In a preferred embodiment of this application, the filtering unit 42 includes:

[0092] The scanning subunit is adapted to scan the noise-reduced signal in real time to identify two types of regions: peak region and baseline non-absorption region.

[0093] The partitioned smoothing subunit is adapted to reduce the SG smoothing window and decrease the polynomial order in the peak region to perform SG smoothing filtering, and to increase the SG window and increase the polynomial order in the baseline non-absorption region to perform SG smoothing filtering.

[0094] The threshold fine-tuning subunit is suitable for adjusting the threshold of the adaptive wavelet threshold denoising process in real time using the noise statistical variance of the non-absorption peak region.

[0095] The filtering module 4 for TDLAS-WMS signals in this embodiment can perform the processing steps of the filtering method for TDLAS-WMS signals in this embodiment. Its principle and effect are the same as those of the filtering method for TDLAS-WMS signals in this embodiment, and will not be repeated here.

[0096] The various techniques described herein can be implemented in combination with hardware or software, or a combination thereof. Thus, the methods and apparatus of the present invention, or certain aspects or portions thereof, can take the form of program code (i.e., instructions) embedded in a tangible medium, such as a removable hard disk, USB flash drive, floppy disk, CD-ROM, or any other machine-readable storage medium, wherein when the program is loaded into and executed by a machine such as a computer, the machine becomes an apparatus for practicing the present invention.

[0097] This application also provides an electronic device. Figure 5 A schematic diagram of an electronic device according to an embodiment of the present invention is shown. (e.g.) Figure 5 As shown, the electronic device mainly includes a memory 6 and one or more processors 7. A bus 8 can be used for communication between the processors 7 and the memory 6. The memory 6 stores program instructions capable of executing the aforementioned filtering method for TDLAS-WMS signals, and the processors 7 can read and execute these program instructions from the memory 6.

[0098] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A filtering method for TDLAS-WMS signals, characterized in that, include: Adaptive wavelet threshold denoising was performed on the original TDLAS-WMS second harmonic signal to obtain the denoised signal; The noise-reduced signal is subjected to SG smoothing filtering to obtain the filtered signal; Extract the real-time baseline from the filtered signal; The real-time baseline is dynamically subtracted from the filtered signal.

2. The filtering method as described in claim 1, characterized in that, The adaptive wavelet threshold noise reduction process for the original TDLAS-WMS second harmonic signal includes: Wavelet decomposition is performed on the original WMS second harmonic signal; The threshold is adaptively calculated based on the signal energy, and non-stationary spikes and high-frequency random noise are eliminated according to the threshold. Reconstruct the signal to preserve the effective absorption peak profile.

3. The filtering method as described in claim 1, characterized in that, The window used in the SG smoothing filter is a second harmonic adaptive window.

4. The filtering method as described in claim 3, characterized in that, The SG smoothing filter is fitted using second- to third-order polynomials.

5. The filtering method as described in claim 1, characterized in that, The real-time baseline is extracted from the filtered signal using an iterative weighted least squares method combined with the Whittaker function.

6. The filtering method according to any one of claims 1 to 4, characterized in that, The step of extracting the real-time baseline from the filtered signal includes: Calculate the local noise variance of the filtered signal, determine the initial weights based on the local noise variance, and perform weighted Whittaker smoothing to obtain the initial baseline and residuals; Candidate peak regions are extracted from the residuals, the confidence level of each candidate peak region is calculated, and a peak region mask is constructed. The Tukey dual-weight function is used in conjunction with peak region masking and peak region boundary for weight transition. The weights are iteratively updated and the baseline is refitted to obtain the real-time baseline.

7. The filtering method according to any one of claims 1 to 5, characterized in that, The SG smoothing filter for the noise-reduced signal includes: The noise-reduced signal is scanned in real time to identify two types of regions: peak region and baseline non-absorption region; In the peak region, the SG smoothing window is reduced and the polynomial order is decreased to perform SG smoothing filtering; in the baseline non-absorption region, the SG window is increased and the polynomial order is increased to perform SG smoothing filtering. The threshold of the adaptive wavelet threshold denoising process is adjusted in real time using the noise statistical variance of the non-absorption peak region.

8. A filtering module for TDLAS-WMS signals, characterized in that, include: The noise reduction unit is suitable for adaptive wavelet threshold noise reduction processing of the original TDLAS-WMS second harmonic signal to obtain a noise-reduced signal. The filtering unit is adapted to perform SG smoothing filtering on the noise reduction signal to obtain the filtered signal; A real-time baseline extraction unit is adapted to extract a real-time baseline from the filtered signal; The baseline subtraction unit is adapted to dynamically subtract the real-time baseline from the filtered signal.

9. A filtering system for TDLAS-WMS signals based on the module described in claim 8, characterized in that, include: The signal acquisition module acquires the original TDLAS-WMS second harmonic signal; A data acquisition card is connected to the signal acquisition module and sends the acquired signals to the signal preprocessing module; A signal preprocessing module, connected to the data acquisition card, preprocesses the signals from the data acquisition card; The filtering module for TDLAS-WMS signals is connected to the signal preprocessing module and performs filtering processing on the preprocessed signal. The data output and display module is connected to the filtering module for TDLAS-WMS signals and displays the data output by the filtering module for TDLAS-WMS signals.

10. An electronic device, characterized in that, include: Memory; processor; as well as Computer programs; The computer program is stored in the memory and configured to be executed by the processor to implement the method as described in any one of claims 1 to 7.

Images