Equipment pipeline harmful gas high-precision detection system and method based on multispectral fusion

Abstract

The invention relates to the technical field of gas detection, and discloses an equipment pipeline harmful gas high-precision detection system and method based on multispectral fusion. And then the dynamic background and interference mode online learning and reconstruction module identifies and reconstructs a main background and interference spectrum by using an online learning mechanism. And the target gas signal extraction and enhancement module subtracts the interference spectrum from the preprocessed data to obtain a residual spectrum, and extracts an enhanced target gas characteristic spectrum component through orthogonal projection by using standard spectrum information. Finally, the multi-gas concentration inversion module calculates the target harmful gas concentration based on the characteristic and the standard spectral information. Dynamic interference is removed through online learning, the reliability of signal extraction is improved by combining orthogonal projection and a dynamic calibration spectrum library, the distinguishing precision of mixed gas is improved by adopting a step-by-step strategy, and long-term high precision and environmental adaptability of the system are ensured through a self-adaptive optimization mechanism.

Description

Technical Field

[0001] The present invention relates to the field of gas detection technology, and in particular to a high-precision detection system and method for harmful gases in equipment pipelines based on multi-spectral fusion. Background Art

[0002] Accurate and reliable detection of hazardous gases is of vital importance in many key areas such as industrial safety monitoring, environmental quality assessment, and emergency response. In particular, in equipment and pipeline systems involving the transportation of flammable, explosive, or toxic substances, timely detection and accurate quantification of potential gas leaks are key to preventing accidents, ensuring personnel safety, and protecting the environment. Multispectral analysis technology has been widely studied and applied in the field of gas detection because it can capture rich spectral "fingerprint" information of substances and provides an effective technical approach for the identification and quantitative analysis of multiple gas components. However, in the actual deployment and application of these detection systems based on multispectral analysis, especially in the pursuit of high precision and high robustness, existing technologies still face several challenges that limit their performance in complex dynamic environments.

[0003] Specifically, some existing multispectral gas detection technologies lack the adaptability and processing capabilities to handle the complex and dynamically changing background signals and unknown interferences found in real-world conditions. For example, fluctuations in water vapor concentration, changes in dust particles, and dynamic interference from other non-target coexisting gases can significantly affect the spectral signal of the target gas. If the system cannot effectively identify and eliminate these time-varying interference components, it can easily lead to biased detection results, compromising detection accuracy. Furthermore, most systems rely on pre-established libraries of standard gas spectra for comparison and quantitative analysis. However, these standard spectra are typically measured under ideal laboratory conditions. When field conditions such as temperature and pressure change, or when the optical properties of the instrument itself drift over time, the spectral characteristics of the target gas collected may mismatch with the library's standard spectra. This mismatch directly impacts the accuracy of signal extraction and the reliability of subsequent concentration inversion. Especially for trace gas detection, even small spectral deviations can lead to significant quantitative errors. Furthermore, when multiple gas mixtures exist in the environment being measured and their absorption spectra overlap, traditional spectral analysis models may struggle to completely separate the signals of each component, resulting in cross-interference between the components and reducing the accuracy of quantitative analysis of specific target gases in the mixture. Finally, many existing systems operate in a relatively fixed mode after parameter setting, lacking the ability to self-adjust and optimize based on real-time monitoring results and environmental changes. This can lead to a gradual decline in system performance over the long term and increase the frequency and complexity of manual maintenance and calibration.

[0004] To this end, the present invention proposes a high-precision detection system and method for harmful gases in equipment pipelines based on multi-spectral fusion. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides a high-precision detection system and method for harmful gases in equipment pipelines based on multi-spectral fusion, which solves the problems of low accuracy and poor adaptability of existing gas detection technology based on multi-spectral analysis under dynamic interference, spectral mismatch and multi-component cross-influence.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a high-precision detection system for harmful gases in equipment pipelines based on multi-spectral fusion, comprising:

[0007] The multispectral data acquisition and preprocessing module is used to obtain the original multispectral data of the gas environment to be measured, and preprocess the original multispectral data to output preprocessed multispectral data;

[0008] A dynamic background and interference modality online learning and reconstruction module is used to receive the pre-processed multispectral data and, based on the pre-processed multispectral data, use an online learning mechanism to identify and reconstruct the main interference spectra caused by the dynamic background and unknown interference in real time;

[0009] a target gas signal extraction and enhancement module, configured to receive the preprocessed multispectral data and the main interference spectrum, obtain a residual spectrum by subtracting the main interference spectrum from the preprocessed multispectral data, and perform orthogonal projection processing on the residual spectrum using preset target gas standard spectrum information to extract and enhance the target gas characteristic spectral component;

[0010] A multi-gas concentration inversion module is used to receive the target gas characteristic spectral components and the preset target gas standard spectral information, and based on the target gas characteristic spectral components and the preset target gas standard spectral information, inversely calculate and obtain the concentration of one or more target harmful gases in the gas environment to be tested.

[0011] Preferably, the dynamic background and interference mode online learning and reconstruction module specifically includes:

[0012] An online algorithm execution unit, configured to execute an online learning algorithm on the preprocessed multispectral data to identify and extract dynamic evolution laws or main change components in the data;

[0013] A main dynamic characteristic identification and selection unit, configured to identify and select spectral change modes representing the main dynamic background and unknown interference according to the dynamic evolution law or the characteristics of the main change components output by the online algorithm execution unit;

[0014] The dynamic background and interference spectrum reconstruction unit is used to reconstruct the main interference spectrum using the spectrum change mode and its corresponding real-time amplitude determined by the main dynamic characteristic identification and selection unit.

[0015] Preferably, the online learning algorithm adopted by the online algorithm execution unit is at least one of an online dynamic mode decomposition algorithm or a sliding window-based robust principal component analysis algorithm.

[0016] Preferably, the target gas signal extraction and enhancement module specifically includes:

[0017] a residual spectrum calculation unit, configured to subtract the reconstructed main interference spectrum from the preprocessed multispectral data to obtain the residual spectrum;

[0018] A target gas standard spectrum subspace construction unit is used to construct a target gas characteristic subspace according to a preset target gas standard spectrum database;

[0019] The orthogonal projection and signal enhancement unit is used to perform orthogonal projection on the residual spectrum to the target gas characteristic subspace to extract and enhance the target gas characteristic spectrum component.

[0020] Preferably, the orthogonal projection and signal enhancement unit is further configured to calculate a non-target residual component orthogonal to the target gas characteristic subspace.

[0021] Preferably, the multiple gas concentration inversion module uses at least one of the classical least squares method, ridge regression or partial least squares regression algorithm to perform concentration inversion on the characteristic spectral components of the target gas.

[0022] Preferably, the preprocessing operation in the multispectral data acquisition and preprocessing module includes at least one of dark spectrum subtraction, light source reference correction or spectrum normalization, preliminary baseline correction and noise filtering.

[0023] Preferably, it also includes:

[0024] The system feedback and adaptive optimization module is used to generate a feedback signal according to the statistical characteristics of the non-target residual component or the stability and rationality of the concentration of one or more target harmful gases in the gas environment to be measured, and dynamically adjust the mode selection strategy of the dynamic background and interference mode online learning and reconstruction module or the calibration parameters of the target gas standard spectrum library in the target gas signal extraction and enhancement module, or the parameters of the concentration inversion model in the multiple gas concentration inversion module based on the feedback signal.

[0025] The present invention also provides a high-precision detection method for harmful gases in equipment pipelines based on multi-spectral fusion, which is used in the above-mentioned high-precision detection system for harmful gases in equipment pipelines based on multi-spectral fusion, comprising the following steps:

[0026] Obtaining original multispectral data of the gas environment to be measured;

[0027] Preprocessing the original multispectral data to obtain preprocessed multispectral data;

[0028] Based on the pre-processed multispectral data, an online learning mechanism is used to identify and reconstruct the main interference spectra caused by dynamic background and unknown interference in real time;

[0029] subtracting the reconstructed main interference spectrum from the preprocessed multispectral data to obtain a residual spectrum;

[0030] Furthermore, the residual spectrum is subjected to orthogonal projection processing using preset target gas standard spectrum information to extract and enhance target gas characteristic spectrum components;

[0031] Finally, based on the enhanced target gas characteristic spectrum component and the target gas standard spectrum information, the concentration of one or more target harmful gases in the gas environment to be measured is obtained by inversion calculation.

[0032] Preferably, the one or more target harmful gases include at least one of hydrocarbon gases, sulfur-containing compound gases, nitrogen oxide gases, carbon monoxide, chlorine and ammonia.

[0033] The present invention provides a high-precision detection system and method for harmful gases in equipment pipelines based on multi-spectral fusion. It has the following beneficial effects:

[0034] 1. By adopting online learning technology, the present invention can identify and subtract dynamically changing background and interference spectra from the original data in real time. This method makes the effective signal of the target gas more prominent, thereby improving the accuracy of detection. Compared with the existing technology that is prone to signal misjudgment or omission when dealing with changing backgrounds, the present invention solves the problem of insufficient accuracy in such scenarios.

[0035] 2. The present invention applies an orthogonal projection method in the target gas signal extraction process and designs a dynamic calibration mechanism for the target gas standard spectral library used. Therefore, even if the actual gas spectrum undergoes slight changes due to environmental factors or the instrument state drifts, the system can more accurately capture the characteristics of the target gas, thereby enhancing the reliability of the signal and overcoming the defects of some existing technologies that rely on a fixed spectral library, resulting in reduced recognition accuracy and poor robustness when the environment or instrument state changes.

[0036] 3. The present invention adopts a strategy of removing the main interfering spectra in advance and then using standard spectral information to specifically extract the target gas signal. This effectively improves the ability to resolve different components in a mixed gas. This process reduces the mutual influence between the spectral characteristics of different gases, especially for those gases with similar spectral characteristics that are easily confused. It can achieve more accurate concentration inversion, and solves the problem that the existing technology has limited differentiation ability when facing complex mixed gases, which easily leads to deviation in the inversion results.

[0037] 4. The present invention includes a system feedback and adaptive optimization module that monitors the stability of the final output concentration and the characteristics of non-target residual components, and dynamically adjusts the operating parameters of the front-end spectral processing module based on this information. This mechanism enables the system to self-optimize, maintain high detection performance for a long time, and adapt to changes in environmental conditions. Compared with systems with fixed parameters that require frequent manual calibration, the present invention improves the convenience of operation and the accuracy of continuous detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a framework flow chart of the system of the present invention;

[0039] Figure 2 This is a framework flow chart of the dynamic background and interference mode online learning and reconstruction module of the present invention;

[0040] Figure 3 This is a framework flow chart of the target gas signal extraction and enhancement module of the present invention. DETAILED DESCRIPTION

[0041] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the present specification. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0042] Please see the attached Figure 1 The embodiment of the present invention provides a high-precision detection system for harmful gases in equipment pipelines based on multi-spectral fusion, including:

[0043] The multispectral data acquisition and preprocessing module is used to obtain the original multispectral data of the gas environment to be measured, and preprocess the original multispectral data to output preprocessed multispectral data;

[0044] To achieve high-precision detection of hazardous gases in equipment pipelines based on multispectral fusion, an exemplary system can include a multispectral data acquisition and preprocessing module. This module acquires raw multispectral data of the gas environment to be measured. This raw multispectral data then undergoes preprocessing within the module. This preprocessing aims to eliminate or reduce noise and irrelevant information in the raw data, laying the foundation for subsequent accurate analysis. The module ultimately outputs the preprocessed multispectral data.

[0045] In some embodiments, the preprocessing operations performed by the multispectral data acquisition and preprocessing module may include, but are not limited to, at least one of the following steps. For example, dark spectrum subtraction may be performed first. This step is used to correct for systematic bias introduced by factors such as sensor dark current.

[0046] Furthermore, preprocessing can include light source reference correction or spectral normalization. Light source reference correction uses a reference spectrum to correct for light source intensity fluctuations or sensor response inconsistencies. Spectral normalization adjusts spectral data to a uniform scale, reducing overall intensity variations caused by variations in optical pathlength and sample concentration, and highlighting spectral shape characteristics.

[0047] Preprocessing can also include preliminary baseline correction. This step aims to eliminate or mitigate spectral baseline variations caused by instrument drift, scattering, and other factors. Finally, noise filtering can be included, such as using mean filtering, median filtering, or wavelet denoising algorithms to smooth the spectral curve and improve the signal-to-noise ratio. These preprocessing steps work together to produce the preprocessed multispectral data required for subsequent analysis.

[0048] The system can further include a dynamic background and interference modality online learning and reconstruction module. This module receives preprocessed multispectral data output by the multispectral data acquisition and preprocessing module. Its core function is to use an online learning mechanism to identify and reconstruct the main interference spectra caused by dynamic background and unknown interference in real time.

[0049] The dynamic background and interference modal online learning and reconstruction module, in its specific internal structure, can include an online algorithm execution unit. This unit directly receives preprocessed multispectral data and is responsible for executing an online learning algorithm to identify and extract the dynamic evolution patterns or main change components contained in the input multispectral data sequence.

[0050] Next, the module can also include a primary dynamic characteristics identification and selection unit. This unit is coupled to the online algorithm execution unit. It analyzes and determines the dynamic evolution patterns or characteristics of the primary variation components output by the online algorithm execution unit, such as the energy, frequency, persistence, or contribution to the background of each mode. Through this process, it identifies and selects spectral variation modes that represent the primary dynamic background and unknown interference.

[0051] Finally, the module can also include a dynamic background and interference spectrum reconstruction unit. This unit is coupled to the primary dynamic characteristics identification and selection unit. It uses the spectral variation modes determined by the primary dynamic characteristics identification and selection unit, as well as the real-time amplitudes or coefficients corresponding to these modes. By linearly combining these selected modes and their corresponding weights or other reconstruction methods, it ultimately reconstructs the primary interference spectrum.

[0052] In a specific embodiment, the online learning algorithm used by the online algorithm execution unit may be an online dynamic mode decomposition (OnlineDMD) algorithm. When the online DMD algorithm is used, the pre-processed multispectral data vector obtained at time point k is Where N is the number of spectral channels.

[0053] The online DMD algorithm aims to learn an approximate linear dynamical system operator A such that x k+1 ≈Ax k The operator A can be updated efficiently when each new data point arrives. For example, the estimate of A or its low-rank approximation can be updated by recursive least squares or methods based on incremental singular value decomposition.

[0054] By performing eigenvalue decomposition on the updated operator A, a series of DMD modes Φ=[φ1,φ2,…,φ r ] and the corresponding DMD eigenvalues ​​Λ=diag(λ1,λ2,…,λ r ). Among them, φ i represents a coherent spectral structure, and λ i Characterize its time dynamic characteristics (growth / decay rate and oscillation frequency). These DMD modes and eigenvalues ​​constitute the "dynamic evolution law".

[0055] In another specific embodiment, the online learning algorithm used by the online algorithm execution unit may also be a sliding window-based robust principal component analysis (RPCA) algorithm. RPCA aims to decompose the observation data matrix D into a low-rank matrix L and a sparse matrix S, i.e., D = L + S.

[0056] In this application, the observation data matrix The W pre-processed multispectral data vectors x collected within the sliding time window W are k Each column is a spectral vector. The low-rank matrix L is used to characterize the slowly changing dynamic background and main interference components. The sparse matrix S is used to characterize the burst signal or partial noise of the target gas.

[0057] The RPCA problem is usually implemented by solving the following convex optimization problem:

[0058] stD=L+S, where ||L|| * represents the nuclear norm (sum of singular values) of the matrix L, which is used to constrain the low rank of L. ||S||1 represents the L1 norm (sum of the absolute values ​​of all elements) of the matrix S, which is used to constrain the sparsity of S. γ is a regularization parameter used to balance the importance of low rank terms and sparse terms.

[0059] Sliding window-based RPCA means that as new spectral data continue to arrive, the data matrix DD is updated and the RPCA decomposition is performed again or incrementally. At this time, the column vectors or main components of the low-rank matrix L obtained by decomposition constitute the "main variation components".

[0060] Whether using online DMD or sliding-window RPCA, the main dynamic characteristics identification and selection unit receives the modes or components output by these algorithms. For example, in DMD, those DMD modes with higher energy, lower frequency, or longer duration can be selected as representing background interference. In RPCA, the low-rank matrix L itself or its main principal components can be directly considered as the main components of background and interference.

[0061] Subsequently, the dynamic background and interference spectrum reconstruction unit uses these selected spectral change modes (such as the selected DMD mode {φ j} j∈selected or the low-rank component of RPCA) and its corresponding real-time amplitude or projection coefficient to reconstruct the main interference spectrum x at the current moment interference For example, for DMD, it can be expressed as x interference (t)=Σ j∈selected φ j a j (t), where a j (t) is the modal φ j The amplitude at time t. For RPCA, the reconstructed interference spectrum is the corresponding column in the low-rank matrix L of the current window.

[0062] Through the collaborative work of the above-mentioned multispectral data acquisition and preprocessing module and the dynamic background and interference mode online learning and reconstruction module, the dynamically changing background and interference signals can be effectively separated from the original complex multispectral data, providing a purer data basis for subsequent target gas signal extraction and concentration inversion, thereby helping to improve the accuracy and robustness of the detection system.

[0063] Please see the attached Figure 2, dynamic background and interference mode online learning and reconstruction module, which is used to receive pre-processed multispectral data and use online learning mechanism to identify and reconstruct the main interference spectrum caused by dynamic background and unknown interference in real time based on the pre-processed multispectral data;

[0064] The system in this embodiment of the present invention may also include an online learning and reconstruction module for dynamic background and interference modalities. This module, coupled to the aforementioned multispectral data acquisition and preprocessing module, receives preprocessed multispectral data. Its core function is to use the received preprocessed multispectral data and employ an online learning mechanism to identify and reconstruct, in real time, the primary interference spectra caused primarily by dynamic background and unknown interference.

[0065] In one exemplary embodiment, the dynamic background and interference modality online learning and reconstruction module can include an online algorithm execution unit. This unit is responsible for receiving and processing the preprocessed multispectral data sequence output by the multispectral data acquisition and preprocessing module. Its primary task is to execute an online learning algorithm to identify and extract the dynamic evolution patterns or major changing components inherent in the input multispectral data stream.

[0066] Following this, the module can also include a key dynamic characteristics identification and selection unit. This unit is logically connected to the output of the online algorithm execution unit. Based on the dynamic evolution patterns or specific characteristics of the key changing components output by the online algorithm execution unit, it conducts in-depth analysis and judgment, such as the energy contribution of each mode, its time-frequency characteristics, its persistence, or its correlation with the background. Through this process, the unit can identify and accurately select the spectral change modes that best represent the primary dynamic background and unknown interference.

[0067] Finally, the dynamic background and interference modal online learning and reconstruction module can also include a dynamic background and interference spectrum reconstruction unit. This unit is connected to the primary dynamic characteristic identification and selection unit. It utilizes the spectral variation modes ultimately determined by the primary dynamic characteristic identification and selection unit, as well as the real-time amplitudes, coefficients, or weights corresponding to these selected modes at the current moment. By effectively combining these selected modes and their corresponding weight information, such as through linear superposition or other appropriate reconstruction algorithms, the primary interference spectrum is ultimately accurately reconstructed.

[0068] To explain more specifically, when the online algorithm execution unit adopts the online dynamic mode decomposition (OnlineDMD) algorithm, it is assumed that the preprocessed multispectral data vector obtained at the discrete time point k is expressed as Where N represents the number of channels or wavelength points of the spectral data. The goal of the online DMD algorithm is to learn an approximate linear dynamical system operator So that the observation sequence satisfies xk+1 ≈Ax k .

[0069] The operator A can be estimated at each new data point x k Efficient online updates are performed upon arrival. For example, the Recursive Least Squares (RLS) algorithm or the Incremental SVD-based method can be used to update the estimate of A itself, or to update its low-rank approximation.

[0070] By performing eigenvalue decomposition on the operator A obtained by the current update, that is, AΦ=ΦΛ, a set of DMD modes Φ=[φ1,φ2,…,φ r ] and their corresponding DMD eigenvalues ​​Λ=diag(λ1,λ2,…,λ r ). Here, represents a spectral mode with a coherent structure, and the complex eigenvalue It characterizes the temporal dynamic characteristics of the mode, including its growth / decay rate (denoted by |λ i |determined) and the oscillation frequency (determined by arg(λ i These DMD modes and eigenvalues ​​together constitute the “dynamic evolution law”.

[0071] In another exemplary embodiment, the online learning algorithm used by the online algorithm execution unit can be a sliding window-based robust principal component analysis (RPCA) algorithm. The basic idea of ​​RPCA is to decompose an observation data matrix D into the sum of a low-rank matrix L and a sparse matrix S, that is, satisfying D = L + S.

[0072] In this application scenario, the observation data matrix The W pre-processed multispectral data collected in a sliding time window of fixed length W are transmitted to x k , where each column vector represents a spectral data point. The low-rank matrix L is used to characterize the slowly changing dynamic background and major interfering components in the data, as these components typically exhibit structural similarities across multiple consecutive spectral observations. The sparse matrix S is used to characterize sudden, sparse spectral signals or partially independent noise generated by target gas leaks and other factors.

[0073] The RPCA problem can usually be solved by solving the following convex optimization problem:

[0074] Among them, D is the observation data matrix, L is the low-rank matrix, S is the sparse matrix, ||L|| * =∑ i σi (L) represents the nuclear norm of the low-rank matrix L, that is, its singular value σ i (L), which is a convex approximation of the rank function and is used to force the low-rank matrix L to have a low-rank structure; ||S||1=∑ i,j |S ij | represents the L1 norm of the sparse matrix S, that is, the sum of the absolute values ​​of all elements in the matrix. This norm is used to make the sparse matrix S sparse. γ>0 is a regularization parameter used to balance the relative importance of low-rank terms and sparse terms in the optimization objective. Its value is usually determined based on experience or cross-validation.

[0075] Sliding window RPCA means that as new spectral data enter the window and the oldest spectral data exit the window, the data matrix D is updated accordingly. The RPCA decomposition is then re-executed on the updated D, or an incremental RPCA algorithm is used for efficient updates. During this process, the column vectors of the decomposed low-rank matrix L, or the principal components obtained from the singular value decomposition of L, constitute the "primary components of variation."

[0076] Whether using an online DMD algorithm or a sliding window-based RPCA algorithm, the main dynamic characteristics identification and selection unit receives the output of these algorithms. For example, when using an online DMD, the unit can select DMD modes that mainly represent background and interference based on the energy of each DMD mode (for example, measured by its initial amplitude or projected energy on the data), the size of its eigenvalue (close to the unit circle indicates persistence), and the frequency (low frequency usually corresponds to background).

[0077] When using RPCA, since the original intention of designing the low-rank matrix L is to capture the background and main interference, the low-rank matrix L itself or several main components extracted by further principal component analysis can be directly selected by the main dynamic characteristic identification and selection unit as the spectral change mode representing the main dynamic background and unknown interference.

[0078] Subsequently, the dynamic background and interference spectrum reconstruction unit will use these spectrum change modes determined by the main dynamic characteristics identification and selection unit (for example, a selected set of DMD modes {φ j} j∈selectedindices or the low-rank components obtained by RPCA decomposition) and their corresponding real-time amplitudes or projection coefficients to reconstruct the main interference spectrum at the current moment, recorded as x interference (k).

[0079] Specifically, if the DMD method is used, the main interference spectrum at time k can be reconstructed by a linear combination of the selected background modes:

[0080] xinterference (k)=∑ j∈selected indices φ j b j (k), where b j (k) is the modal φ j The amplitude at time k; x interference (k) is the main interference spectrum vector reconstructed at time k, with a dimension of N×1; φ j is the jth selected DMD (dynamic modal decomposition) mode vector, representing a coherent spectral structure with a dimension of N×1; b j (k) is the DMD mode φ j The amplitude or coefficient at time k; ∑ j∈selected indices represents the sum of all DMD modes selected to represent background and interference. These amplitudes can be estimated online by projecting the raw data onto the DMD modes or other methods.

[0081] If the RPCA method is used, the main interference spectrum at time k can be directly obtained from the data matrix D in the current sliding window k The low-rank matrix L obtained by decomposition k The column vector corresponding to the current moment is:

[0082] x interference (k)=(L k ) :,c , where (L k ) :,c Represents the low-rank matrix L obtained at time k (corresponding to the current sliding window) k The cth column of x interference (k) is the main interference spectrum vector reconstructed at time k.

[0083] Through the precise operation of the dynamic background and interference modal online learning and reconstruction module, the system can effectively separate the main spectral components of the dynamically changing background and unknown interference from continuously acquired multispectral data in real time. This provides a purer data foundation for the subsequent target gas signal extraction and concentration inversion steps, thus significantly improving the detection accuracy, sensitivity, and robustness of the entire hazardous gas detection system in complex environments.

[0084] Please see the attached Figure 3 , a target gas signal extraction and enhancement module is used to receive the pre-processed multi-spectral data and the main interference spectrum, obtain a residual spectrum by subtracting the main interference spectrum from the pre-processed multi-spectral data, and perform orthogonal projection processing on the residual spectrum using the preset target gas standard spectrum information to extract and enhance the target gas characteristic spectral component;

[0085] The system in this embodiment of the present invention may also include a target gas signal extraction and enhancement module. This module's upstream inputs are coupled to the multispectral data acquisition and preprocessing module and the dynamic background and interference modal online learning and reconstruction module. It is configured to receive the preprocessed multispectral data output by the former and the primary interference spectrum output by the latter. Its core function is to extract and enhance the characteristic spectral components of the target gas through specialized processing.

[0086] Specifically, the module first subtracts the main interference spectrum from the preprocessed multispectral data to obtain a residual spectrum. It then performs orthogonal projection processing on this residual spectrum using the preset target gas standard spectrum information. In this way, the module can effectively extract the characteristic spectral signal of the target gas and enhance the intensity of this signal relative to the residual noise and interference to a certain extent, ultimately outputting the target gas characteristic spectral component.

[0087] In an exemplary embodiment, the target gas signal extraction and enhancement module may specifically include a residual spectrum calculation unit. The input end of this unit is connected to the output end of the multispectral data acquisition and preprocessing module and the output end of the dynamic background and interference mode online learning and reconstruction module. It is used to receive the preprocessed multispectral data, denoted as (where k is the time index and N is the number of spectral channels), and the reconstructed main interference spectrum is recorded as

[0088] The operation performed by the residual spectrum calculation unit is to subtract the reconstructed main interference spectrum from the pre-processed multi-spectral data. The difference obtained is the residual spectrum, denoted as x residual (k). Its calculation process can be expressed as:

[0089] x residual (k) = x k -x interference (k), where x residual (k) is the residual spectrum vector calculated at time k, with a dimension of N×1; x k is the preprocessed multispectral data vector at time k, with a dimension of N×1; x interference (k) is the main interference spectrum vector reconstructed at time k.

[0090] This residual spectrum x residual (k) Theoretically, it mainly includes the spectral signal of the target gas and some residual interference and noise that have not been completely removed. This unit outputs the calculated residual spectrum to the subsequent processing unit.

[0091] Furthermore, the target gas signal extraction and enhancement module can also include a target gas standard spectral subspace construction unit. This unit constructs a mathematical subspace that effectively characterizes the spectral characteristics of one or more target gases, known as the target gas characteristic subspace, based on a preset target gas standard spectrum database. This database stores standard reference spectra for various target gases to be detected.

[0092] Assume that the target gas standard spectrum database contains M target gas standard spectra, which can be arranged into a matrix Each column is the standard spectrum vector of the jth target gas. The target gas standard spectrum subspace construction unit uses this matrix S to construct the target gas characteristic subspace.

[0093] A commonly used method is to perform singular value decomposition (SVD) on the standard spectral matrix S: S = U∑V T ,in and is an orthogonal matrix, It is a diagonal matrix whose diagonal elements are singular values. Select P left singular vectors corresponding to larger singular values ​​in U (for example, greater than a threshold or the first P according to energy ratio) to form an orthogonal basis matrix This matrix U S The column vectors of t form the target gas characteristic subspace. Alternatively, Gram-Schmidt orthogonalization or other methods can be used to extract an orthogonal basis from S. This unit outputs the constructed subspace basis U S .

[0094] Finally, the target gas signal extraction and enhancement module can also include an orthogonal projection and signal enhancement unit. The input end of this unit is connected to the output end of the residual spectrum calculation unit and the output end of the target gas standard spectrum subspace construction unit. It receives the residual spectrum x residual (k) and the basis matrix U of the target gas characteristic subspace S .

[0095] The core operation of the orthogonal projection and signal enhancement unit is to convert the received residual spectrum x residual (k) Perform an orthogonal projection onto the target gas characteristic subspace. This projection operation extracts the components of the residual spectrum that are most relevant to the target gas spectral characteristics and suppresses noise and interference components that are orthogonal to the target gas characteristic subspace, thereby achieving the extraction and enhancement of the target gas characteristic spectral components.

[0096] Target gas characteristic spectral component x target The calculation formula for (k) is:

[0097] in, is the orthogonal projection operator projected onto the target gas characteristic subspace; x target (k) is the target gas characteristic spectral component vector extracted and enhanced at time k; U S is the orthogonal basis matrix of the target gas characteristic subspace constructed from the target gas standard spectrum library; residual (k) is the residual spectrum vector at time k; x target (k) is one of the main results output by this unit, which will be passed to the subsequent multiple gas concentration inversion modules.

[0098] In addition, in some embodiments, the orthogonal projection and signal enhancement unit is further configured to calculate a non-target residual component orthogonal to the target gas characteristic subspace. The non-target residual component x non-target (k) represents the part of the residual spectrum that cannot be explained by the target gas characteristic subspace, which mainly contains residual noise and other unknown interferences.

[0099] Non-target residual component x non-target (k) can be calculated by subtracting its projection on the target gas characteristic subspace from the residual spectrum: non-target (k) = x residual (k)-x target (k) Alternatively, it can be obtained by projecting onto the complementary space orthogonal to the target gas characteristic subspace:

[0100] Where I is the N×N identity matrix, x non-target (k) is the non-target residual component vector calculated at time k; x residual (k) is the residual spectrum vector at time k; is an orthogonal projection operator that projects onto the complementary space orthogonal to the target gas characteristic subspace. This non-target residual component x non-target (k) can be output to the system feedback and adaptive optimization module to evaluate system performance and make parameter adjustments.

[0101] Through the coordinated work of the above units, the target gas signal extraction and enhancement module can effectively separate and enhance the characteristic signal of the target gas from the residual spectrum after preliminary interference suppression, while also providing information about non-target components, providing key support for the high-precision and adaptive optimization of the entire detection system.

[0102] The multi-gas concentration inversion module is used to receive the target gas characteristic spectral components and the preset target gas standard spectral information, and based on the target gas characteristic spectral components and the preset target gas standard spectral information, inversely calculate the concentration of one or more target harmful gases in the gas environment to be tested.

[0103] The system in this embodiment of the present invention may also include a multi-gas concentration inversion module. The module's input is coupled to the output of the aforementioned target gas signal extraction and enhancement module. It is configured to receive the target gas characteristic spectral components output by the module and also receive or pre-store preset target gas standard spectral information. Based on these two types of information, the module's core task is to inversely calculate the concentration of one or more target harmful gases in the gas environment being measured.

[0104] The characteristic spectral component of the target gas is recorded as is the spectrum extracted and enhanced at time k, where N is the number of spectral channels. The preset target gas standard spectral information is usually expressed as a standard spectral library matrix Each column s of this matrix j Represents the standard spectral response of the jth target harmful gas at unit concentration (or other known concentration), and there are M target gases in total.

[0105] This module is based on basic principles in spectroscopy, such as the expanded form of the Beer-Lambert law, which states that the absorbance of a mixture (or a spectral characteristic quantity related to it) is the linear superposition of the absorbances of its components. Therefore, the characteristic spectral component x of the target gas is target (k) can be approximately expressed as the standard spectrum library matrix S and the concentration vector of the gas to be measured A linear combination of j (k) is the concentration of the jth target gas at time k.

[0106] In an exemplary embodiment, the multiple gas concentration inversion module can use the Classical Least Squares (CLS) method to perform concentration inversion. The CLS method aims to find a concentration vector c(k) such that the reconstructed spectrum Sc(k) is consistent with the observed target gas characteristic spectrum component x target (k) is the smallest residual square sum between them. That is, solve the following optimization problem:

[0107] Where c(k) is the target harmful gas concentration vector to be solved at time k; x target (k) is the target gas characteristic spectral component vector extracted and enhanced at time k; S is the target gas standard spectral library matrix; is the square of the L2 norm of the vector (i.e. the sum of the squares of the vector elements);

[0108] If the matrix S T If S is reversible, the analytical solution of the above least squares problem is:

[0109] c(k)=(S T S) -1 S T x target (k), where c(k) is the target harmful gas concentration vector calculated at time k; S is the target gas standard spectrum library matrix; S T is the transpose of the matrix S; (S T S) -1 is the matrix S T The inverse matrix of x target (k) is the target gas characteristic spectral component vector extracted and enhanced at time k;

[0110] The concentrations of various target harmful gases can be calculated using the above formula.

[0111] In another exemplary embodiment, when there is a strong linear correlation between the standard spectra (i.e., multicollinearity problem), resulting in S T When S approaches singularity or pathological conditions, the CLS method may be unstable. In this case, the multi-gas concentration inversion module can use the Ridge Regression algorithm. Ridge regression improves the stability of the solution by adding a regularization term (the square of the L2 norm of the concentration) to the least squares objective function. Its optimization goal is:

[0112] Where c(k) is the target harmful gas concentration vector to be solved at time k; x target (k) is the target gas characteristic spectral component vector extracted and enhanced at time k; S is the target gas standard spectral library matrix; is the square of the L2 norm of the concentration vector c(k), α>0 is the ridge parameter or regularization parameter, which is used to control the strength of the regularization term. The solution of ridge regression is:

[0113] c(k)=(S T S+αI) -1 S T x target (k), where I is the M×M identity matrix; c(k) is the target harmful gas concentration vector calculated at time k; S is the target gas standard spectrum library matrix; S T is the transpose of the matrix S; α is the ridge parameter or regularization parameter, a positive scalar; (S T S+αI) -1 is the matrix (S T The inverse matrix of S+αI); xtarget (k) is the target gas characteristic spectral component vector extracted and enhanced at time k. The selection of the ridge parameter α is crucial to model performance and can usually be determined through methods such as cross-validation.

[0114] Furthermore, in more complex scenarios, such as when the spectral data is noisy or contains many collinear variables, the Multi-Gas Concentration Inversion Module can also use the Partial Least Squares Regression (PLS) algorithm. PLS is a multivariate statistical analysis method that establishes a regression model by extracting the underlying structure (i.e., latent variables or factors) between the independent variable (the standard spectral library S or features derived from its transformation) and the dependent variable (the concentration c(k)).

[0115] The PLS algorithm first extracts the spectrum from the standard spectral library S (or the X train ) and the corresponding known concentration Y train Extract several pairs of latent variables (score vectors) from S. Let the latent variable score of S be T = [t1,…,t A ] (A is the number of latent variables), the latent variable score of concentration is U=[u1,…,u A ]. PLS establishes regression relationships in these latent variable spaces, such as U = TB inner +F, where B inner is the internal regression coefficient matrix, and F is the residual.

[0116] For the new target gas characteristic spectral component x target (k), first project it into the latent variable space of S to obtain its score vector t new (k). The concentration score u is then predicted using the established internal regression model new (k), and finally converted into the actual concentration value c(k) through the load of concentration space. The whole process can be summarized as a total regression coefficient matrix B PLS , so that: c(k)=B PLS x target (k), where B PLS It is learned during the calibration phase using the PLS algorithm. The PLS algorithm can effectively process high-dimensional, collinear data and has good prediction performance.

[0117] The multi-gas concentration inversion module ultimately outputs the calculated concentration vector c(k) of one or more target hazardous gases in the gas environment being measured. This concentration information is the final output of the gas detection system and can be used for subsequent leak alarms, environmental assessments, or process control. By employing at least one of the aforementioned inversion algorithms, this module accurately converts enhanced spectral signatures into gas concentrations, making it a key component of the entire high-precision detection system.

[0118] Also includes:

[0119] The system feedback and adaptive optimization module is used to generate feedback signals based on the statistical characteristics of non-target residual components or the stability and rationality of the concentrations of one or more target harmful gases in the gas environment to be measured, and dynamically adjust the mode selection strategy of the dynamic background and interference mode online learning and reconstruction module or the calibration parameters of the target gas standard spectrum library in the target gas signal extraction and enhancement module, or the parameters of the concentration inversion model in the multiple gas concentration inversion module based on the feedback signals.

[0120] The system in the embodiment of the present invention may further include a system feedback and adaptive optimization module. The function of this module is to generate feedback signals based on specific performance indicators during system operation and dynamically adjust the parameters or strategies of other key modules in the system based on these feedback signals, in order to continuously optimize the system's overall detection performance and adaptability to environmental changes.

[0121] In an exemplary embodiment, the system feedback and adaptive optimization module may include a performance evaluation and feedback signal generation unit. The input end of the unit is configured to receive performance indication information from other modules. Specifically, it can receive the non-target residual component output by the target gas signal extraction and enhancement module, denoted as x non-target (k). At the same time, it can also receive the concentration of one or more target harmful gases in the gas environment to be measured calculated by the multiple gas concentration inversion module, which is recorded as c(k).

[0122] The performance evaluation and feedback signal generation unit analyzes the received information. For example, it can calculate the non-target residual component x non-target The statistical properties of (k), such as its energy (e.g. ), variance, or the presence of certain expected or unexpected features in its spectral structure. Higher energies or the presence of specific structures may indicate incomplete background subtraction or the presence of unmodeled interferences.

[0123] The unit also assesses the stability and rationality of the concentration c(k). For example, it monitors whether the concentration value is within the physically acceptable range, whether its rate of change over time is too fast or too slow, or whether there are persistent, unreasonable deviations or oscillations. By comprehensively analyzing these performance indicators, the unit generates a feedback signal. This feedback signal can be one or more scalar values, status codes, or structured data containing specific adjustment instructions.

[0124] The system feedback and adaptive optimization module can also include a parameter adjustment and model update unit. This unit is coupled to the performance evaluation and feedback signal generation unit and is configured to receive the feedback signal generated by the performance evaluation and feedback signal generation unit. Based on this feedback signal, this unit is responsible for dynamically adjusting and optimizing the operating parameters or internal models of one or more other modules in the system.

[0125] Specifically, the parameter adjustment and model update unit can dynamically adjust the mode selection strategy of the dynamic background and interference mode online learning and reconstruction module. For example, if the feedback signal indicates the non-target residual component x non-target If (k) contains significant dynamic components that should be removed by the background reconstruction module, the threshold for selecting dynamic modes can be adjusted. For example, when using online DMD, the energy or frequency threshold used to filter background modes can be adjusted; when using sliding window-based RPCA, the target rank of the low-rank matrix L or the regularization parameter γ can be adjusted to capture more background changes.

[0126] Furthermore, the parameter adjustment and model update unit can also dynamically adjust the calibration parameters of the target gas standard spectrum library in the target gas signal extraction and enhancement module. For example, if the feedback signal indicates that the non-target residual component x non-target (k) still exists in the standard spectrum of a certain target gas s j Components with similar shapes but slightly shifted or deformed, or a gas concentration c j (k) shows a systematic deviation over a long period of time, which may indicate that the corresponding spectrum s in the original standard spectral library S j Calibration required.

[0127] This calibration can be done by comparing the standard spectrum s j Perform small wavelength shifts, peak shape adjustments, or intensity scaling. For example, a calibration model Δs j To correct the standard spectrum: s j,calibrated =f cal (s j ,θ j ), where f cal is the calibration function, θ j is a set of calibration parameters for gas j. The feedback signal can drive the j Adjustments. Updated s j,calibrated Will replace the original s j Used for subsequent target gas characteristic subspace construction and concentration inversion.

[0128] Furthermore, the parameter adjustment and model update unit can dynamically adjust the parameters of the concentration inversion models in various gas concentration inversion modules. For example, if the feedback signal indicates that the calculated concentration c(k) exhibits excessive noise or instability, the regularization parameter α can be appropriately increased when using ridge regression to enhance model stability. Conversely, if the concentration response is too smooth, resulting in insufficient tracking of true changes, α can be reduced.

[0129] If partial least squares regression (PLS) is used, the feedback signal can be used to adjust the number of latent variables A used in the model. For example, if the model is overfitting (manifested by the concentration being sensitive to noise), the number of latent variables can be reduced; if it is underfitting (insufficient response to true concentration changes), the number of latent variables can be increased. In some cases, if the system performance continues to be poor, the feedback signal can even trigger a change in the entire PLS model (i.e., the regression coefficient matrix B). PLS ) is retrained using recently accumulated, verified data.

[0130] Through this closed-loop feedback and adaptive optimization mechanism, the system can continuously self-adjust and improve its internal models and parameters based on actual operating results and environmental changes. This helps maintain and improve gas detection accuracy, robustness, and adaptability to complex and dynamic environments, thereby achieving more reliable and efficient hazardous gas monitoring.

[0131] The high-precision detection method for harmful gases in equipment pipelines based on multi-spectral fusion described below and the high-precision detection system for harmful gases in equipment pipelines based on multi-spectral fusion described above can be referenced to each other.

[0132] The high-precision detection method for harmful gases in equipment pipelines based on multi-spectral fusion includes the following steps:

[0133] Obtaining original multispectral data of the gas environment to be measured;

[0134] Preprocessing the original multispectral data to obtain preprocessed multispectral data;

[0135] Based on the pre-processed multispectral data, an online learning mechanism is used to identify and reconstruct the main interference spectra caused by dynamic background and unknown interference in real time;

[0136] The reconstructed main interference spectrum is subtracted from the preprocessed multispectral data to obtain the residual spectrum;

[0137] Furthermore, the residual spectrum is orthogonally projected using the preset target gas standard spectrum information to extract and enhance the target gas characteristic spectrum components;

[0138] Finally, based on the enhanced target gas characteristic spectral components and target gas standard spectral information, the concentration of one or more target harmful gases in the gas environment to be measured is obtained by inversion calculation.

[0139] The one or more target harmful gases include at least one of hydrocarbon gases, sulfur compound gases, nitrogen oxide gases, carbon monoxide, chlorine and ammonia.

[0140] The method of this embodiment can be used to execute the above system embodiment. Its principles and technical effects are similar and will not be described in detail here.

[0141] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. High-precision detection system for harmful gases in equipment pipelines based on multi-spectral fusion, characterized by: include: The multispectral data acquisition and preprocessing module is used to obtain the original multispectral data of the gas environment to be measured, and preprocess the original multispectral data to output preprocessed multispectral data; A dynamic background and interference modality online learning and reconstruction module is used to receive the pre-processed multispectral data and, based on the pre-processed multispectral data, use an online learning mechanism to identify and reconstruct the main interference spectra caused by the dynamic background and unknown interference in real time; a target gas signal extraction and enhancement module, configured to receive the preprocessed multispectral data and the main interference spectrum, obtain a residual spectrum by subtracting the main interference spectrum from the preprocessed multispectral data, and perform orthogonal projection processing on the residual spectrum using preset target gas standard spectrum information to extract and enhance the target gas characteristic spectral component; A multi-gas concentration inversion module is used to receive the target gas characteristic spectral components and the preset target gas standard spectral information, and based on the target gas characteristic spectral components and the preset target gas standard spectral information, inversely calculate and obtain the concentration of one or more target harmful gases in the gas environment to be tested.

2. The high-precision detection system for harmful gases in equipment pipelines based on multi-spectral fusion according to claim 1 is characterized in that: The dynamic background and interference mode online learning and reconstruction module specifically includes: An online algorithm execution unit, configured to execute an online learning algorithm on the preprocessed multispectral data to identify and extract dynamic evolution laws or main change components in the data; A main dynamic characteristic identification and selection unit, configured to identify and select spectral change modes representing the main dynamic background and unknown interference according to the dynamic evolution law or the characteristics of the main change components output by the online algorithm execution unit; The dynamic background and interference spectrum reconstruction unit is used to reconstruct the main interference spectrum using the spectrum change mode and its corresponding real-time amplitude determined by the main dynamic characteristic identification and selection unit.

3. The high-precision detection system for harmful gases in equipment pipelines based on multi-spectral fusion according to claim 2 is characterized in that: The online learning algorithm adopted by the online algorithm execution unit is at least one of an online dynamic mode decomposition algorithm or a robust principal component analysis algorithm based on a sliding window.

4. The high-precision detection system for harmful gases in equipment pipelines based on multi-spectral fusion according to claim 1 is characterized in that: The target gas signal extraction and enhancement module specifically includes: a residual spectrum calculation unit, configured to subtract the reconstructed main interference spectrum from the preprocessed multispectral data to obtain the residual spectrum; A target gas standard spectrum subspace construction unit is used to construct a target gas characteristic subspace according to a preset target gas standard spectrum database; The orthogonal projection and signal enhancement unit is used to perform orthogonal projection on the residual spectrum to the target gas characteristic subspace to extract and enhance the target gas characteristic spectrum component.

5. The high-precision detection system for harmful gases in equipment pipelines based on multi-spectral fusion according to claim 4 is characterized in that: The orthogonal projection and signal enhancement unit is further configured to calculate a non-target residual component orthogonal to the target gas characteristic subspace.

6. The high-precision detection system for harmful gases in equipment pipelines based on multi-spectral fusion according to claim 1 is characterized in that: The multiple gas concentration inversion module uses at least one of the classical least squares method, ridge regression or partial least squares regression algorithm to perform concentration inversion on the characteristic spectral components of the target gas.

7. The high-precision detection system for harmful gases in equipment pipelines based on multi-spectral fusion according to claim 1 is characterized in that: The preprocessing operation in the multispectral data acquisition and preprocessing module includes at least one of dark spectrum subtraction, light source reference correction or spectrum normalization, preliminary baseline correction and noise filtering.

8. The high-precision detection system for harmful gases in equipment pipelines based on multi-spectral fusion according to claim 5 is characterized in that: Also includes: The system feedback and adaptive optimization module is used to generate a feedback signal according to the statistical characteristics of the non-target residual component or the stability and rationality of the concentration of one or more target harmful gases in the gas environment to be measured, and dynamically adjust the mode selection strategy of the dynamic background and interference mode online learning and reconstruction module or the calibration parameters of the target gas standard spectrum library in the target gas signal extraction and enhancement module, or the parameters of the concentration inversion model in the multiple gas concentration inversion module based on the feedback signal.

9. A high-precision detection method for harmful gases in equipment pipelines based on multi-spectral fusion, characterized in that: The high-precision detection system for harmful gases in equipment pipelines based on multi-spectral fusion as described in any one of claims 1 to 8 comprises the following steps: Obtaining original multispectral data of the gas environment to be measured; Preprocessing the original multispectral data to obtain preprocessed multispectral data; Based on the pre-processed multispectral data, an online learning mechanism is used to identify and reconstruct the main interference spectra caused by dynamic background and unknown interference in real time; subtracting the reconstructed main interference spectrum from the preprocessed multispectral data to obtain a residual spectrum; Furthermore, the residual spectrum is subjected to orthogonal projection processing using preset target gas standard spectrum information to extract and enhance target gas characteristic spectrum components; Finally, based on the enhanced target gas characteristic spectrum component and the target gas standard spectrum information, the concentration of one or more target harmful gases in the gas environment to be measured is obtained by inversion calculation.

10. The high-precision detection method for harmful gases in equipment pipelines based on multi-spectral fusion according to claim 9 is characterized in that: The one or more target harmful gases include at least one of hydrocarbon gas, sulfur compound gas, nitrogen oxide gas, carbon monoxide, chlorine gas and ammonia gas.

Images