Method for detecting hydrogen-fueled engine exhaust gas by terahertz multi-component based on sparse inversion

Abstract

This invention discloses a terahertz multi-component detection method for hydrogen-fired engine exhaust gas based on sparse inversion. The method includes: selecting a predetermined number of discrete frequency points in the terahertz band as observation frequency points, acquiring discrete spectral data, and preprocessing it to obtain an observation data vector; parameterizing the absorption characteristics of multiple gas components in the exhaust gas based on the observation data vector to construct a feature dictionary; utilizing the sparse structure where the number of gas components in the actual exhaust gas is less than the size of the feature dictionary, solving the parameter vector by introducing sparse constraints, identifying the gas type based on the feature atoms corresponding to the significantly non-zero parameter components in the parameter vector, and estimating the concentration of the gas component based on the sum of the significantly non-zero parameter components belonging to the same gas component. This invention, through sparse frequency sampling and sparse inversion, achieves rapid identification and concentration inversion of multiple gas components in hydrogen-fired engine exhaust gas without relying on the complete terahertz spectrum, significantly reducing measurement and computational overhead.

Description

Technical Field

[0001] This invention belongs to the field of terahertz gas absorption spectroscopy analysis and digital signal processing technology, specifically relating to a terahertz spectrum processing method for multi-component analysis of hydrogen-fired engine exhaust gas, and particularly a terahertz multi-component detection method for hydrogen-fired engine exhaust gas based on sparse inversion. Background Technology

[0002] With the development of hydrogen-fired engines and hydrogen-powered aviation propulsion technology, real-time monitoring and accurate analysis of key components in hydrogen-fired engine exhaust have become an important foundation for engine performance evaluation, emission control, and safe operation. Hydrogen-fired engine exhaust typically contains various gaseous components such as water vapor (H2O), nitrogen oxides (NO, NO2), and unburned hydrogen (H2). Accurately obtaining their concentration information is of great significance for environmental impact assessment and combustion status diagnosis.

[0003] The terahertz band contains characteristic rotational absorption lines of various gas molecules, making terahertz absorption spectroscopy a promising method for multi-component gas analysis. However, hydrogen engine exhaust exhibits significantly complex spectral characteristics in the terahertz band: on the one hand, water vapor absorption lines are dense and intense, further broadening and overlapping under high-temperature conditions, forming a strong absorption background; nitrogen oxide absorption lines are few in number and weak in intensity, easily masked by the water vapor background; on the other hand, nitrogen oxide absorption lines in the terahertz band are relatively few in number and weak in intensity, easily masked by the strong absorption background of water vapor; furthermore, hydrogen molecules absorb very weakly in the terahertz band and can be considered a low-absorption component. These characteristics result in a complex terahertz spectrum of hydrogen exhaust gas exhibiting a "strong background, weak features, and significant spectral line overlap," increasing the difficulty of multi-component identification and quantitative inversion.

[0004] Existing terahertz spectral analysis methods mostly employ broadband scanning or terahertz time-domain spectroscopy techniques, requiring the acquisition of relatively complete spectral information and fitting analysis. This results in high measurement and computational costs, making it difficult to balance real-time performance and robustness. When multi-component absorption characteristics overlap or spectral line shifts occur due to temperature perturbations, conventional inversion methods are prone to problems such as unstable identification and large concentration estimation errors.

[0005] Therefore, there is an urgent need for a fast and robust multi-component detection method that can be implemented with limited observation frequencies. Summary of the Invention

[0006] This invention aims to address the challenges of strong water vapor absorption background, weak target component absorption characteristics, and susceptibility to high-temperature broadening and frequency shifts in the terahertz spectrum of hydrogen-fired engine exhaust gases. It provides a sparse inversion-based terahertz multi-component detection method for hydrogen-fired engine exhaust gases. This method acquires discriminative spectral information through sparse frequency sampling, and combines parametric modeling considering operating condition disturbances with sparse-constrained inversion to achieve efficient and robust qualitative and quantitative analysis of multi-component gases. Simultaneously, it reduces measurement and computational costs, adapting to online monitoring requirements.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] A terahertz multi-component detection method for hydrogen-fired engine exhaust gas based on sparse inversion includes the following steps: Within the terahertz band, a preset number of discrete frequency points are selected as observation frequency points. Terahertz absorption or transmission measurements are performed on the exhaust gas of the hydrogen-fired engine to obtain discrete spectral data at the observation frequency points. The discrete spectral data is preprocessed to obtain an observation data vector; Based on the observed data vector, the absorption characteristics of multiple gas components in the exhaust gas of a hydrogen-fired engine are parametrically modeled, and a feature dictionary is constructed to characterize the superposition relationship of multi-component absorption. The feature dictionary contains feature atoms corresponding to candidate absorption states under different temperature and / or pressure conditions. An observation-parameter mapping relationship is established between the observation data vector and the feature dictionary to construct an observation model. Taking advantage of the sparse structure that the number of gas components present in the actual exhaust gas is less than the size of the feature dictionary, the parameter vector is solved by introducing sparse constraints. The gas type is identified based on the feature atoms corresponding to the significantly non-zero parameter components in the parameter vector, and the concentration of the gas component is estimated based on the sum of the significantly non-zero parameter components belonging to the same gas component.

[0009] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention acquires terahertz absorption information through sparse frequency sampling, enabling multi-component gas identification and concentration inversion without requiring high-density scanning of the entire terahertz band. Compared to traditional methods that rely on acquiring the complete spectrum, this approach significantly reduces the amount of measurement data and computational overhead, providing a technical foundation for achieving rapid or even near real-time exhaust gas analysis.

[0010] 2. In view of the characteristics of strong absorption of water vapor and weak absorption of nitrogen oxides in the exhaust gas of hydrogen-fired engines, which are easily masked, this invention utilizes sparse parameter domain modeling and sparse constraint inversion mechanism to selectively activate features corresponding to the actual gas components during the inversion process, thereby effectively suppressing the interference of strong background components on weak absorption features and improving the identification stability of key emissions such as nitrogen oxides.

[0011] 3. In the absorption feature modeling process, this invention incorporates factors such as spectral line broadening and center frequency shift caused by high-temperature environments into a unified parameter representation framework. By combining parameterized modeling with sparse constraint inversion, the algorithm's adaptability to non-ideal spectral morphologies is improved. Compared to traditional frequency domain analysis methods that rely on fixed spectral line positions, this invention exhibits better robustness under complex operating conditions.

[0012] 4. The technical solution of this invention consists of algorithm modules such as sparse frequency point observation, parameterized modeling, and sparse inversion. The overall structure is clear, the calculation process is controllable, and it does not rely on overly complex physical modeling or large-scale training processes. It is easy to implement and integrate in engineering systems and is suitable for applications such as online analysis and condition monitoring of hydrogen fuel cell engine exhaust. Attached Figure Description

[0013] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a schematic flowchart of the terahertz multi-component detection method for hydrogen-fired engine exhaust gas based on sparse inversion, according to an embodiment of the present invention. Detailed Implementation

[0015] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0016] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. This application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0017] This invention provides a terahertz multi-component detection method for hydrogen-fired engine exhaust gas based on sparse inversion, such as... Figure 1 As shown, the method includes the following steps: Step 1: Discrete Frequency Selection and Terahertz Measurement Within the terahertz band, a predetermined number of M discrete frequency points are selected as observation frequency points, where the predetermined number is finite. These observation frequency points are not randomly selected, but rather chosen based on the characteristic absorption regions of the target gas (e.g., H₂O, NO, NO₂) within the terahertz band, the absorption variation characteristics of the water vapor background gas (e.g., flat regions), and the ability of different frequency points to distinguish between multiple components. Specifically, from a pre-determined set of candidate frequency points, the set of frequency points with superior evaluation indicators is selected as the observation frequency points by calculating the absorption difference at each candidate frequency point, the ability to distinguish between strong water vapor backgrounds, and the sensitivity to temperature disturbances. By selecting frequency points with high discriminative information density, the amount of observation data is reduced while retaining the information basis that is significant for discriminating the absorption characteristics of multiple components, providing effective input for subsequent component identification and concentration inversion.

[0018] Terahertz time-domain spectroscopy or continuous-wave terahertz system is used to perform terahertz absorption or transmission measurements on the exhaust gas of a hydrogen-fired engine to obtain discrete spectral data at the M observation frequency points.

[0019] This step improves the reliability of multi-component identification and concentration inversion under high-temperature exhaust gas conditions by selecting a limited number of observation frequency points with high discriminative information density, thereby reducing measurement and computation costs and avoiding the adverse effects of strong water vapor absorption background and spectral line overlap on inversion stability.

[0020] Step 2: Preprocessing and Vector Representation of Observation Data The discrete spectral data obtained in step 1 undergoes preprocessing such as normalization, background compensation, or relative absorbance calculation to reduce the impact of source power fluctuations, system gain changes, and environmental disturbances on subsequent inversion results. Then, the absorbance or equivalent absorbance at M observation frequency points is calculated and uniformly represented as an observation data vector. :

[0021] Where M is the number of observation frequency points, Frequency point The observed values ​​at the specified locations. This observed data vector serves as a unified input for subsequent parametric modeling and component inversion.

[0022] Step 3: Parametric modeling and feature dictionary construction considering operating condition disturbances Based on the observation data vector obtained in step 2 This invention parametrically models the absorption characteristics of multiple gas components in the exhaust gas of a hydrogen-fired engine, constructing a feature dictionary to characterize the superposition relationship of multi-component absorption. In this embodiment, the multiple gas components include water vapor and nitrogen oxides. The theoretical absorption cross section (i.e., gas absorption spectral line) of each gas component under given frequency, temperature, and pressure conditions can be described using either a Lorentz line type or a Voith line type.

[0023] In this embodiment, to adapt to the spectral broadening and center frequency shift of the exhaust gas from a hydrogen-fired engine under different temperature and pressure conditions, multiple candidate absorption states are constructed for each target gas within preset temperature and pressure ranges, and their corresponding absorption responses are represented as characteristic atoms. The preset temperature and pressure ranges are determined based on the measurement location and actual operating conditions; for example, the temperature range can be 300K–1500K, and the pressure range can be 0.5 atm–3 atm. It should be noted that the above ranges are merely exemplary and do not constitute a limitation of the present invention.

[0024] Let the first Each characteristic atom at frequency ,temperature and pressure The continuous absorption response function under the given conditions is: Under conditions where pressure broadening dominates, the continuous absorption response function can be exemplarily expressed using a Lorentz linear form as follows:

[0025] in, For characteristic atom index, For the first Effective spectral line intensity factor of each characteristic atom For reference center frequency, This is the center frequency offset related to temperature and / or pressure. This refers to the broadening factor, which is related to temperature and pressure. In this embodiment, the broadening factor can be exemplarily represented as the half-width at half-maximum (WHM) of the corresponding spectral line.

[0026] In a preferred embodiment, the broadening factor It can be represented as:

[0027] in, For the first The broadening factor of each characteristic atom under reference operating conditions and These are the reference pressure and reference temperature, respectively. This is the corresponding temperature dependence index.

[0028] It should be noted that, in other embodiments, in order to simultaneously consider Doppler broadening and collision broadening, the continuous absorption response function can also be described using the Voith curve, and the present invention is not limited to the Lorentz curve described above.

[0029] Determining discrete observation frequency points Then, the continuous absorption response function is sampled at each observation frequency point to obtain the first... Discrete vector representation of each characteristic atom:

[0030] in, Indicates the relationship with the first Candidate operating condition parameters corresponding to each feature atom. By discretizing the preset temperature and pressure ranges, multiple feature atoms corresponding to different operating conditions can be generated for each target gas. All feature atom vectors are combined to form a feature dictionary.

[0031] in, For the number of observation frequency points, This represents the total number of feature atoms contained in the feature dictionary, and is usually... The feature dictionary is used to cover various absorption response forms that the target gas may exhibit under different operating conditions, thereby enhancing the model's adaptability to complex operating conditions and parameter uncertainties of high-temperature exhaust gases.

[0032] Step 4: Parameter Inversion and Component Identification Based on Sparse Constraints Because the number of gaseous components present in the exhaust gas of an actual hydrogen-fired engine is less than the number of feature atoms N contained in the constructed feature dictionary, and only a few feature atoms match the actual absorption response under the same operating conditions, the parameter vector used to characterize the contribution of each feature exhibits a sparse or nearly sparse structure overall. Therefore, the parameter vector to be estimated... It has a sparse structure.

[0033] Based on the above structural characteristics, the observation data vector obtained in step 2 will be... Establish an observation-parameter mapping relationship with the feature dictionary D constructed in step 3, and construct an observation model:

[0034] in, Let be the sparse parameter vector to be estimated. This is the noise term.

[0035] In this embodiment of the invention, sparse constraints are introduced on the parameter vector. Inversion is performed to effectively activate only the parameter components corresponding to the actual gas components, thereby suppressing the influence of strong water vapor background and irrelevant features on the inversion results and improving the detectability and inversion stability of weakly absorbing gas components.

[0036] In a preferred embodiment, this is achieved by solving the following optimization problem with sparse regularization terms:

[0037] in, The regularization parameter can be determined through cross-validation or the L-curve method. The above solution is the result of parameter vector estimation. It should be noted that the above solution form is only used to illustrate one implementation of sparse constraint inversion, and this invention does not limit the specific constraint form or solution method.

[0038] The parameter vector obtained by solving In this context, the gas types represented by the characteristic atoms corresponding to significantly non-zero parameter components are the identified exhaust gas components. Specifically, if... If the parameter component corresponding to a certain gas component is significantly non-zero, then the gas component is determined to exist in the exhaust gas.

[0039] In this embodiment of the invention, "significantly non-zero" refers to the absolute value of a parameter component being greater than a preset threshold. This threshold can be preset or adaptively determined based on the noise level of the terahertz measurement system, the intensity of the absorption spectral lines, and the convergence tolerance of the numerical solution. It is understood that, due to the presence of noise in actual measurements and the difficulty in obtaining strictly zero solutions using sparse optimization algorithms, threshold determination can accurately identify the actual gas components.

[0040] Step 5: Concentration Inversion and Result Output Based on the parameter vector estimation results obtained in step 4 The concentration of each target gas component is inverted. Let the first... The set of characteristic atom indices corresponding to the target gas is Then the estimated concentration of the target gas is... Represented as:

[0041] in, Indicates the first Estimated concentrations of the gases. parameter vector The Middle A significantly non-zero parameter component, For the first The scaling factor corresponding to the target gas is determined by the optical path length, the system calibration coefficient, and the unit conversion relationship.

[0042] The final output includes the identification results of each target gas component in the exhaust gas of the hydrogen-fired engine and the corresponding concentration estimates. When necessary, the inversion results can be evaluated for consistency or stability to meet the reliability requirements of online monitoring scenarios. The method described in this embodiment can stably identify weakly absorbing components such as nitrogen oxides even under strong water vapor background conditions, and exhibits good robustness to high-temperature spectral broadening and frequency shifts.

[0043] In summary, this invention, through the synergistic effect of sparse frequency point observation, parametric modeling considering operating condition disturbances, and inversion solution based on sparse constraints, achieves rapid identification and concentration inversion of multiple gas components in hydrogen-fired engine exhaust gas without relying on a complete terahertz spectrum scan. This technical solution significantly reduces the data volume and computational complexity of terahertz spectral measurements while ensuring the accuracy of multi-component identification and concentration inversion, and enhances the analytical stability and reliability under complex operating conditions of high-temperature exhaust gas. It is suitable for online monitoring and rapid analysis applications of hydrogen-fired engine exhaust gas.

[0044] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations can be made to the embodiments of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A terahertz multi-component detection method for hydrogen-fired engine exhaust gas based on sparse inversion, characterized in that, Includes the following steps: Within the terahertz band, a preset number of discrete frequency points are selected as observation frequency points. Terahertz absorption or transmission measurements are performed on the exhaust gas of the hydrogen-fired engine to obtain discrete spectral data at the observation frequency points. The discrete spectral data is preprocessed to obtain an observation data vector; Based on the observed data vector, the absorption characteristics of multiple gas components in the exhaust gas of a hydrogen-fired engine are parametrically modeled, and a feature dictionary is constructed to characterize the superposition relationship of multi-component absorption. The feature dictionary contains feature atoms corresponding to candidate absorption states under different temperature and / or pressure conditions. The multiple gas components include water vapor and nitrogen oxides. An observation-parameter mapping relationship is established between the observation data vector and the feature dictionary to construct an observation model. Taking advantage of the sparse structure that the number of gas components present in the actual exhaust gas is less than the size of the feature dictionary, the parameter vector is solved by introducing sparse constraints. The gas type is identified based on the feature atoms corresponding to the significantly non-zero parameter components in the parameter vector, and the concentration of the gas component is estimated based on the sum of the significantly non-zero parameter components belonging to the same gas component.

2. The method according to claim 1, characterized in that, The observation frequencies are selected based on the characteristic absorption region of the target gas in the terahertz band, the absorption variation characteristics of the water vapor background gas, and the ability of different frequencies to distinguish multi-component gases. This reduces the amount of observation data while retaining information that is significant for discriminating the absorption characteristics of multi-component gases.

3. The method according to claim 1, characterized in that, In the parametric modeling, Lorentz or Voith line types are used to describe gas absorption spectra. Temperature and pressure parameters are discretized within a preset temperature and pressure range. Multiple feature atoms corresponding to different operating conditions are constructed for each gas and combined to form a feature dictionary.

4. The method according to claim 3, characterized in that, The Lorentz line type is represented as follows: in, For continuous absorption response function, For characteristic atom index, For frequency, For temperature, For pressure, For the first Effective spectral line intensity factor of each characteristic atom For reference center frequency, This is the center frequency offset related to temperature and / or pressure. The broadening factor is related to temperature and pressure, and the broadening factor is expressed as... in, For the first The broadening factor of each characteristic atom under reference operating conditions and These are the reference pressure and reference temperature, respectively. This is the corresponding temperature dependence index.

5. The method according to claim 1, characterized in that, The observation model is as follows: in, Let D be the observed data vector, and D be the feature dictionary. Let be the parameter vector to be estimated. This is the noise term.

6. The method according to claim 5, characterized in that, The sparse constraint is achieved by solving an optimization problem with sparse regularization terms: in, For regularization parameters, This is the result of parameter vector estimation.

7. The method according to claim 1, characterized in that, The preprocessing includes one or more of the following: normalization, background compensation, or relative absorption calculation.

8. The method according to claim 1, characterized in that, The actual gas components present in the exhaust gas are determined by identifying the gas types represented by the characteristic atoms corresponding to the significantly non-zero parameter components in the parameter vector, and the concentration of the corresponding gas components is calculated based on the significantly non-zero parameter components, optical path length, system calibration coefficient, and unit conversion relationship.

9. The method according to claim 8, characterized in that, The concentration of the gaseous component is calculated according to the following formula: in, Indicates the first Estimated concentrations of the gases. For the first A significantly non-zero parameter component, For the first The scaling factor corresponding to the target gas is determined by the optical path length, the system calibration coefficient, and the unit conversion relationship.

Images