Intelligent fuel identification system based on spectrum identification
By acquiring, preprocessing, extracting and selecting full-spectrum signals, and combining them with an improved support vector machine model, the problems of unstable identification and difficulty in revealing dynamic features in traditional fuel identification systems have been solved, thus achieving accurate fuel type identification and combustion process analysis.
Patent Information
- Application Number
- CN202411390996.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-08
- Publication Date
- 2026-01-23
AI Technical Summary
Traditional fuel identification systems rely on a limited number of spectral features or empirical judgments, which cannot provide detailed information, are difficult to handle noise and background interference, resulting in unstable identification results and difficulty in revealing the dynamic changes and multidimensional characteristics of the flame combustion process.
The system uses a signal acquisition unit to acquire full-spectrum signal data, a signal preprocessing unit to remove noise, a feature extraction unit to extract spectral features, a feature selection unit to select representative features, an improved support vector machine model to identify fuel, and an intelligent display unit to display the results.
It provides detailed combustion flame spectral characteristics, eliminates noise interference, improves identification accuracy, reveals dynamic changes in the combustion process, supports combustion monitoring and fire identification, and enables accurate fuel type identification.
Smart Images
Figure CN121384831A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of fuel identification, and particularly relates to an intelligent fuel identification system based on spectral identification. BACKGROUND
[0002] In the field of fuel, spectral identification specifically refers to using the spectral characteristics generated during fuel combustion to identify the type, combustion state and other related parameters of the fuel. During the combustion process of the fuel, specific spectral signals are generated, which can reflect important information such as the chemical composition, temperature and combustion efficiency of the fuel.
[0003] Traditional systems may only rely on a small number of spectral features or empirical judgments to identify the type of fuel, and cannot provide sufficient detailed information or accurate combustion stage analysis. Moreover, traditional systems may not be able to effectively handle noise and background interference in the signal, resulting in unstable identification results or misjudgments. Furthermore, traditional systems may have difficulty in quickly and accurately extracting key features from flame spectral data, thus being inefficient in fuel identification and combustion process analysis. Moreover, traditional systems may be limited to static feature analysis or analysis of a small number of frequency bands, and cannot effectively reveal the dynamic changes and multi-dimensional characteristics in the flame combustion process. SUMMARY
[0004] The present application aims to solve the problems in the background art and proposes an intelligent fuel identification system based on spectral identification.
[0005] The technical solution of the present application is an intelligent fuel identification system based on spectral identification, which comprises a feature selection unit, a fuel identification unit and an intelligent display unit, and further comprises:
[0006] A signal acquisition unit is configured to acquire spectral signals of the combustion flame of the target fuel at the water analysis stage, the volatile release combustion stage, the fixed carbon combustion stage and the burnout stage by a spectrometer, so as to obtain full-spectrum signal data of the combustion flame at different combustion stages. The full-spectrum signal data comprises OH free radical radiation intensity data, CN free radical radiation intensity data, CH free radical radiation intensity data, C2 free radical radiation intensity data and combustion flame radiation intensity data. The full-spectrum signal data of the combustion flame at different combustion stages is transmitted to a signal preprocessing unit.
[0007] A signal preprocessing unit is configured to receive the full-spectrum signal data of the combustion flame at different combustion stages transmitted by the signal acquisition unit, and remove the envelope line of the full-spectrum signal data of the combustion flame at different combustion stages, so as to obtain standard full-spectrum signal data of the combustion flame at different combustion stages. The standard full-spectrum signal data of the combustion flame at different combustion stages is transmitted to a feature extraction unit.
[0008] a feature extraction unit configured to receive the standard full-spectrum signal data of the combustion flame at different combustion stages transmitted by the signal preprocessing unit, and extract features from the standard full-spectrum signal data of the combustion flame at different combustion stages to obtain a spectrum feature set of the combustion flame at different combustion stages, the spectrum feature set of the combustion flame at different combustion stages including the emissivity of each wave band, the absorption coefficient of each component, the fluctuation rate, the specific variance, the flicker frequency and the uniformity, and transmit the spectrum feature set of the combustion flame at different combustion stages to the feature selection unit.
[0009] Preferably, the feature selection unit is configured to receive the spectrum feature set of the combustion flame at different combustion stages transmitted by the feature extraction unit, calculate the information density and the density difference between each spectrum feature in the spectrum feature set of the combustion flame at different combustion stages, select the spectrum feature with the information density greater than a preset information density threshold and the density difference less than a preset density difference threshold to obtain a standard spectrum feature set of the combustion flame at different combustion stages, and transmit the standard spectrum feature set of the combustion flame at different combustion stages to the fuel identification unit.
[0010] Preferably, the fuel identification unit is configured to receive the standard spectrum feature set of the combustion flame at different combustion stages transmitted by the feature selection unit, input the standard spectrum feature set of the combustion flame at different combustion stages into a trained fuel identification model, and output the fuel category corresponding to the standard spectrum feature set of the combustion flame at different combustion stages, i.e. the fuel category corresponding to the target fuel, by the trained fuel identification model, wherein the fuel identification model is an improved support vector machine model, the support vector machine model is improved by an optimization algorithm to obtain the fuel identification model, and the fuel category corresponding to the target fuel is transmitted to the intelligent display unit.
[0011] Preferably, the intelligent display unit is configured to receive the fuel category corresponding to the target fuel transmitted by the fuel identification unit, and display the fuel category corresponding to the target fuel through an intelligent display terminal.
[0012] Preferably, the full-spectrum signal data of the combustion flame at different combustion stages is envelope-removed to obtain the standard full-spectrum signal data of the combustion flame at different combustion stages, including:
[0013] deriving a first derivative of the full spectrum signal to obtain a point corresponding to a first derivative of the full spectrum signal being zero, and deriving a second derivative of the point corresponding to the first derivative being zero to obtain a point corresponding to a second derivative of the full spectrum signal being zero, if the second derivative is less than zero, the point is a local maximum point, and counting all the local maximum points in the full spectrum signal to obtain a local maximum point set;
[0014] comparing each local maximum point in the local maximum point set to obtain a maximum extreme point, and obtaining a coordinate corresponding to the maximum extreme point, the maximum extreme point calculation formula being as follows:
[0015]
[0016] wherein E represents the maximum extreme point, represents an i-th local maximum point in the local maximum point set, and n represents a total number of local maximum points in the local maximum point set;
[0017] based on a slope corresponding to each point in the full spectrum signal based on the maximum extreme point, the slope corresponding to each point calculation formula being as follows:
[0018]
[0019] wherein K i represents a slope corresponding to an i-th point in the full spectrum signal, S E represents a coordinate corresponding to the maximum extreme point, S i represents a coordinate corresponding to the i-th point in the full spectrum signal.
[0020] Preferably, the full spectrum signal data of the combustion flame at different combustion stages is subjected to envelope removal to obtain standard full spectrum signal data of the combustion flame at different combustion stages, and further comprising:
[0021] calculating a first point set and a second point set of the full spectrum signal, the first point set and the second point set calculation formula being as follows:
[0022]
[0023] wherein Q i represents a tangent point corresponding to a slope corresponding to an i-th point in the first point set on the full spectrum signal, P i represents a tangent point corresponding to a slope corresponding to an i-th point in the second point set on the full spectrum signal;
[0024] An envelope line point set is obtained based on the first point set, the second point set and the maximum extreme point, and an envelope line of the full spectrum signal is removed based on the envelope line point set, and the envelope line point set is as follows:
[0025] L = {P1,..., P i , E, Q E+1 ,..., Q i};
[0026] Wherein, L represents the envelope line point set.
[0027] Preferably, the emissivity expression of each waveband is as follows:
[0028]
[0029] Wherein, α(v) represents the emissivity of each waveband, Rad represents the radiation intensity value, I b (v, T) represents the radiation intensity value under the same condition;
[0030] The absorption coefficient expression of each component is as follows:
[0031] k(v ij ) = ∑S ij ·f L (v ij );
[0032] Wherein, k(v ij ) represents the absorption coefficient of each component, S ij represents the spectral line intensity, f L represents the spectral line shape, and v ij represents the wave number of the characteristic waveband;
[0033] The fluctuation rate expression is as follows:
[0034]
[0035] Wherein, R represents the combustion flame fluctuation rate, X represents the combustion flame radiation intensity, MAX(X) represents the maximum value of the combustion flame radiation intensity, MIN(X) represents the minimum value of the combustion flame radiation intensity, and MEAN(X) represents the average value of the combustion flame radiation intensity;
[0036] The specific variance expression is as follows:
[0037]
[0038] Wherein, B represents the combustion flame specific variance, D represents the variance of the combustion flame radiation intensity, and u represents the mathematical expectation of the combustion flame radiation intensity;
[0039] The flicker frequency expression is as follows:
[0040]
[0041] Wherein, F represents the flicker frequency, P(i) represents the power spectrum value of the i-th point in the full spectrum signal, and f(i) represents the frequency of the i-th point in the full spectrum signal.
[0042] Preferably, the information density between each spectral feature is calculated according to the following formula:
[0043]
[0044] Wherein, ρ M represents the information density between each spectral feature, Y M represents the spectral feature of the M-th data in the spectral feature set of the combustion flame at different combustion stages, H(Y M ) represents the information entropy of the M-th spectral feature, and p represents the total number of spectral features in the spectral feature set of the combustion flame at different combustion stages.
[0045] The density difference between each spectral feature is calculated according to the following formula:
[0046]
[0047] Wherein, J M―N represents the density difference between each spectral feature, Y N represents the spectral feature of the N-th data in the spectral feature set of the combustion flame at different combustion stages, and H(Y N ) represents the information entropy of the N-th spectral feature.
[0048] Preferably, the support vector machine model is improved by an optimization algorithm to obtain the fuel identification model, including:
[0049] Mapping the parameters to be optimized of the support vector machine model to the individual positions of the optimization algorithm, the parameters to be optimized including the parameters of the kernel function and the penalty parameter, taking the classification error rate of the support vector machine model as the fitness value of the optimization algorithm, initializing the individual population and the maximum iteration number of the optimization algorithm;
[0050] Each of the individuals performs exploration operation and development operation, and judges the next behavior operation according to the results of the exploration operation and the development operation, the next behavior operation including exploration operation and development operation, after a round of search, according to the fitness value of each of the individuals obtained, the optimal fitness value is stored in the optimal solution individual position;
[0051] In the iterative optimization process, according to the fitness value of the individual, it is judged whether the iteration number meets the maximum iteration number, if yes, the iteration is terminated, the parameter corresponding to the optimal individual position is output, the optimal to-be-optimized parameter is output, and the support vector machine model is improved based on the optimal to-be-optimized parameter, otherwise, the above operation is repeatedly executed.
[0052] Preferably, the position updating formula of the exploration operation is as follows:
[0053] X(t+1)=X(t)·σ+r·(Xbest―I·X(t));
[0054] Wherein, X(t+1) and X(t) represent the position of the individual at the t+1th and tth iteration of the exploration operation, sigma represents a wavelet function disturbance factor, r represents a random number in the interval [0,1], Xbest represents the optimal solution individual position, and I represents a random number in the interval [0,2];
[0055] The position updating formula of the development operation is as follows:
[0056]
[0057] Wherein, And X(t+1) and X(t) represent the position of the individual at the t+1th and tth iteration of the development operation, ub local And lb local Jth-dimensional variable upper limit and lower limit updated with iteration number, Rmv represents the interaction relationship between the current individual and the random individual, bf represents the benefit factor, and 1 or 2 is randomly selected.
[0058] Compared with the prior art, the above technical scheme of the present application has the following beneficial technical effects:
[0059] 1、The signal acquisition unit is used for collecting full-spectrum signal data, and the full-spectrum signal data includes free radical radiation intensity and combustion flame radiation intensity and other information of each combustion stage, detailed spectral characteristics of the combustion flame in different stages are provided, and the characteristics can reflect the changes of chemical substances and energy release in the combustion process, the envelope line of the full-spectrum signal data of the combustion flame in different combustion stages is removed through the signal preprocessing unit, so that noise and background interference can be removed, the obtained full-spectrum signal data is clearer and more accurate, and the spectral signal data of different combustion stages has higher comparability and repeatability.
[0060] 2、The present application extracts the spectral feature set of the standard full spectrum signal data through the feature extraction unit, so that the energy release of the flame in different combustion stages can be understood by analyzing the emissivity of each wave band, which is crucial for evaluating the combustion efficiency, and the concentration of different chemical substances in the flame and the chemical dynamic change in the combustion process can be revealed, and information about the flame combustion instability and combustion dynamics change can be provided, which helps to understand the oscillation characteristics and stability of the flame, and through the calculation of information density and density difference by the feature selection unit, the most representative and distinctive spectral features can be selected, so that the spectral characteristics of the flame in different combustion stages can be more accurately described, and excessive redundant information and unnecessary complexity can be avoided, and the trained fuel recognition model can accurately identify the fuel type corresponding to the flame, which is crucial for combustion monitoring in industrial production process, fuel identification in fire scene and environmental monitoring. BRIEF DESCRIPTION OF DRAWINGS
[0061] Figure 1 A flowchart of the overall system of an embodiment of the present application.
[0062] Reference signs: 1, signal acquisition unit; 2, signal preprocessing unit; 3, feature extraction unit; 4, feature selection unit; 5, fuel recognition unit; 6, intelligent display unit. DETAILED DESCRIPTION
[0063] Embodiment one, as shown in the figure, the present application proposes an intelligent fuel recognition system based on spectral recognition, which includes a feature selection unit 4, a fuel recognition unit 5 and an intelligent display unit 6, and further includes: Figure 1 The signal acquisition unit 1 is used to collect the spectral signal of the combustion flame of the target fuel in the water analysis stage, the volatile release combustion stage, the fixed carbon combustion stage and the burnout stage through the spectrometer, so as to obtain the full spectrum signal data of the combustion flame in different combustion stages, the full spectrum signal data includes OH free radical radiation intensity data, CN free radical radiation intensity data, CH free radical radiation intensity data, C2 free radical radiation intensity data and combustion flame radiation intensity data, and the full spectrum signal data of the combustion flame in different combustion stages is transmitted to the signal preprocessing unit 2;
[0064] The signal preprocessing unit 2 receives the full spectrum signal data of the combustion flame in different combustion stages transmitted by the signal acquisition unit 1, and removes the envelope line of the full spectrum signal data of the combustion flame in different combustion stages, so as to obtain the standard full spectrum signal data of the combustion flame in different combustion stages, and the standard full spectrum signal data of the combustion flame in different combustion stages is transmitted to the feature extraction unit 3;
[0065]
[0066] The feature extraction unit 3 receives the standard full-spectrum signal data of the combustion flame at different combustion stages transmitted by the signal preprocessing unit 2, and extracts features from the standard full-spectrum signal data of the combustion flame at different combustion stages to obtain a spectrum feature set of the combustion flame at different combustion stages, which includes the emissivity of each wave band, the absorption coefficient of each component, the fluctuation rate, the specific variance, the flicker frequency and the uniformity, and transmits the spectrum feature set of the combustion flame at different combustion stages to the feature selection unit 4.
[0067] In the present application, the water analysis stage is the early stage of fuel combustion, mainly the stage of water analysis, at which the water in the fuel is heated and released, resulting in the appearance of water vapor in the flame; the volatile combustion stage is the second stage of fuel combustion, at which the volatile components in the fuel are volatilized and burned at high temperature, producing combustible gas; the fixed carbon combustion stage is the stage at which the remaining carbon in the fuel begins to burn at a higher temperature after the volatile components in the fuel are burned out, which is called the fixed carbon combustion stage; the burnout stage is the last stage of fuel combustion, at which all combustible materials are burned out and the flame gradually extinguishes; the full-spectrum signal data refers to the complete signal data obtained within the spectral range, including the radiation intensity information at all wavelengths; the free radical is an intermediate in chemical reactions, and the radiation intensity data emitted by the free radical in the combustion process reflects the reaction activity in the combustion process; envelope removal is a process of processing signal data to remove the envelope of the signal, i.e., removing the basic profile or trend of the signal, so as to more clearly analyze the detailed features of the signal.
[0068] In an optional embodiment, the feature selection unit 4 receives the spectrum feature set of the combustion flame at different combustion stages transmitted by the feature extraction unit 3, calculates the information density and density difference between each spectrum feature in the spectrum feature set of the combustion flame at different combustion stages, selects the spectrum features with an information density greater than a preset information density threshold and a density difference less than a preset density difference threshold to obtain a standard spectrum feature set of the combustion flame at different combustion stages, and transmits the standard spectrum feature set of the combustion flame at different combustion stages to the fuel identification unit 5.
[0069] In an optional embodiment, the fuel recognition unit 5 receives the standard spectral feature set of the combustion flame at different combustion stages transmitted by the feature selection unit 4, and inputs the standard spectral feature set of the combustion flame at different combustion stages into the trained fuel recognition model, and outputs the fuel type corresponding to the standard spectral feature set of the combustion flame at different combustion stages, i.e. the fuel type corresponding to the target fuel, by the trained fuel recognition model. The fuel recognition model adopts an improved support vector machine model, which is improved by an optimization algorithm to obtain the fuel recognition model. The fuel type corresponding to the target fuel is transmitted to the intelligent display unit 6.
[0070] In an optional embodiment, the intelligent display unit 6 receives the fuel type corresponding to the target fuel transmitted by the fuel recognition unit 5, and displays the fuel type corresponding to the target fuel through the intelligent display terminal.
[0071] It should be noted that the intelligent display terminal is a display device integrated with intelligent functions, which is usually used for real-time monitoring and displaying various data, information or results.
[0072] In Embodiment Two, the present application provides an intelligent fuel recognition system based on spectral recognition. Compared with Embodiment One, the present embodiment further comprises: removing the envelope of the full-spectrum signal data of the combustion flame at different combustion stages to obtain the standard full-spectrum signal data of the combustion flame at different combustion stages, including:
[0073] Deriving the first-order derivative of the full-spectrum signal to obtain the point corresponding to the zero first-order derivative of the full-spectrum signal, and deriving the second-order derivative of the point corresponding to the zero first-order derivative to obtain the point corresponding to the zero second-order derivative of the full-spectrum signal. If the second-order derivative is less than zero, the point is a local maximum point. All local maximum points in the full-spectrum signal are counted to obtain a local maximum point set;
[0074] Comparing each local maximum point in the local maximum point set to obtain the maximum extreme point, and obtaining the coordinates corresponding to the maximum extreme point. The calculation formula of the maximum extreme point is as follows:
[0075]
[0076] Wherein, E represents the maximum extreme point, represents the i-th local maximum point in the local maximum point set, and n represents the total number of local maximum points in the local maximum point set.
[0077] Based on the slope corresponding to each point in the full-spectrum signal of the maximum extreme point, the calculation formula of the slope corresponding to each point is as follows:
[0078]
[0079] wherein K i represents the slope corresponding to the i-th point of the full spectrum signal, S E represents the coordinate corresponding to the maximum extreme point, S i represents the coordinate corresponding to the i-th point of the full spectrum signal.
[0080] In an optional embodiment, the envelope removal is performed on the full spectrum signal data of the combustion flame in different combustion stages to obtain standard full spectrum signal data of the combustion flame in different combustion stages, and the envelope removal further comprises:
[0081] The first point set and the second point set of the full spectrum signal are calculated, and the calculation formulae of the first point set and the second point set are as follows:
[0082]
[0083] wherein Q i represents the tangent point corresponding to the slope of the i-th point in the first point set on the full spectrum signal, P i represents the tangent point corresponding to the slope of the i-th point in the second point set on the full spectrum signal;
[0084] The envelope point set is obtained based on the first point set, the second point set and the maximum extreme point, and the envelope of the full spectrum signal is removed based on the envelope point set, and the envelope point set is as follows:
[0085] L = {P1,..., P i , E, Q E+1 ,..., Q i};
[0086] wherein L represents the envelope point set.
[0087] In an optional embodiment, the emissivity expression of each wave band is as follows:
[0088]
[0089] wherein a(v) represents the emissivity of each wave band, Rad represents the radiation intensity value, I b (v, T) represents the radiation intensity value under the same condition;
[0090] The absorption coefficient expression of each component is as follows:
[0091] k(v ij ) = ∑S ij ·f L (v ij );
[0092] wherein k(vij represents the absorption coefficient of each component, S ij represents the spectral line intensity, f L represents the spectral line type, v ij represents the wave number of the characteristic wave band;
[0093] The fluctuation rate expression is as follows:
[0094]
[0095] wherein R represents the combustion flame fluctuation rate, X represents the combustion flame radiation intensity, MAX(X) represents the maximum value of the combustion flame radiation intensity, MIN(X) represents the minimum value of the combustion flame radiation intensity, and MEAN(X) represents the average value of the combustion flame radiation intensity;
[0096] The specific variance expression is as follows:
[0097]
[0098] wherein B represents the combustion flame specific variance, D represents the variance of the combustion flame radiation intensity, and u represents the mathematical expectation of the combustion flame radiation intensity;
[0099] The flicker frequency expression is as follows:
[0100]
[0101] wherein F represents the flicker frequency, P(i) represents the power spectrum value of the i-th point in the full spectrum signal, and f(i) represents the frequency of the i-th point in the full spectrum signal.
[0102] In an optional embodiment, the information density calculation formula between each spectral feature is as follows:
[0103]
[0104] wherein p M represents the information density between each spectral feature, Y M represents the spectral feature of the M-th data in the spectral feature set of the combustion flame at different combustion stages, H(Y M ) represents the information entropy of the M-th spectral feature, and p represents the total number of spectral features in the spectral feature set of the combustion flame at different combustion stages;
[0105] The density difference calculation formula between each spectral feature is as follows:
[0106]
[0107] wherein J M―N represents the density difference between each spectral feature, Y NH(YN) represents a spectral feature of the Nth data in a spectral feature set representing spectral features of a combustion flame in different combustion stages. N H(YN) represents a spectral feature of the Nth data in a spectral feature set representing spectral features of a combustion flame in different combustion stages.
[0108] In an optional embodiment, the support vector machine model is improved by an optimization algorithm to obtain a fuel identification model, comprising:
[0109] Mapping the parameters to be optimized of the support vector machine model to the individual positions of the optimization algorithm, the parameters to be optimized including the parameters of the kernel function and the penalty parameter, taking the classification error rate of the support vector machine model as the fitness value of the optimization algorithm, initializing the individual population and the maximum iteration number of the optimization algorithm;
[0110] Each individual performs an exploration operation and a development operation, and judges the next behavior operation according to the results of the exploration operation and the development operation, the next behavior operation including the exploration operation and the development operation, after a round of search, according to the fitness value of each individual obtained, the optimal fitness value is stored in the optimal solution individual position;
[0111] In the iterative optimization process, the iteration is performed according to the fitness value of the individual, it is judged whether the iteration number meets the maximum iteration number, if it meets, the iteration is terminated and the parameters corresponding to the best individual position are output, and the optimal parameter to be optimized is output, and the support vector machine model is improved based on the optimal parameter to be optimized, otherwise, the above operation is repeatedly executed.
[0112] It should be noted that the support vector machine model is a supervised learning algorithm for classification and regression analysis, which finds an optimal hyperplane in the feature space to perform classification, so that the distance of sample points of different categories from the hyperplane is maximized; the optimization algorithm is an algorithm for finding the parameters of the support vector machine model, the purpose is to find the best parameter combination of the support vector machine model in the parameter space; in the optimization algorithm, the individual position refers to a point in the parameter space, which represents a parameter combination of the support vector machine model; in the optimization algorithm, the fitness value is used to evaluate the degree of goodness of each individual (parameter combination); the individual population refers to the set of parameter combinations considered simultaneously in the optimization algorithm; the exploration operation and the development operation are basic operations in the optimization algorithm, the exploration operation refers to exploring new positions in the parameter space, and the development operation refers to finding better solutions near the current position.
[0113] In an optional embodiment, the position update formula of the exploration operation is as follows:
[0114] X(t+1)=X(t)·σ+r·(Xbest―I·X(t));
[0115] wherein X(t+1) and X(t) represent the position of the individual at the (t+1)th and tth iteration of the exploration operation, σ represents a wavelet function perturbation factor, r represents a random number in the interval [0, 1], Xbest represents the position of the optimal solution individual, and I represents a random number in the interval [0, 2];
[0116] The position updating formula of the exploitation operation is as follows:
[0117]
[0118] wherein, and X(t+1) and X(t) represent the position of the individual at the (t+1)th and tth iteration of the exploitation operation, ub local and l blocal ub and lb represent the upper and lower bounds of the jth variable updated with the iteration number, Rmv represents the interaction relationship between the current individual and the random individual, bf represents a benefit factor, and 1 or 2 is randomly selected.
[0119] The embodiments of the present application are described in detail above with reference to the drawings, but the present application is not limited thereto, and various changes can be made within the knowledge of those skilled in the art without departing from the spirit of the present application.
Claims
1. An intelligent fuel identification system based on spectral recognition, comprising a feature selection unit (4), a fuel identification unit (5) and an intelligent display unit (6), characterized in that: a signal acquisition unit (1) is configured to collect spectral signals of a target fuel in a water analysis stage, a volatile emission combustion stage, a fixed carbon combustion stage and a burnout stage by a spectrometer to obtain full-spectrum signal data of the combustion flame in different combustion stages, wherein the full-spectrum signal data comprises OH free radical radiation intensity data, CN free radical radiation intensity data, CH free radical radiation intensity data, C2 free radical radiation intensity data and combustion flame radiation intensity data, and the full-spectrum signal data of the combustion flame in different combustion stages is transmitted to a signal preprocessing unit (2); the signal preprocessing unit (2) receives the full-spectrum signal data of the combustion flame in different combustion stages transmitted by the signal acquisition unit (1), and removes the envelope of the full-spectrum signal data of the combustion flame in different combustion stages to obtain standard full-spectrum signal data of the combustion flame in different combustion stages, and the standard full-spectrum signal data of the combustion flame in different combustion stages is transmitted to a feature extraction unit (3); the feature extraction unit (3) receives the standard full-spectrum signal data of the combustion flame in different combustion stages transmitted by the signal preprocessing unit (2), and extracts features from the standard full-spectrum signal data of the combustion flame in different combustion stages to obtain a spectral feature set of the combustion flame in different combustion stages, wherein the spectral feature set of the combustion flame in different combustion stages comprises emissivity of each waveband, absorption coefficient of each component, fluctuation rate, specific variance, flicker frequency and uniformity, and the spectral feature set of the combustion flame in different combustion stages is transmitted to the feature selection unit (4). the feature selection unit (4) receives the spectral feature set of the combustion flame in different combustion stages transmitted by the feature extraction unit (3), calculates information density and density difference between each spectral feature in the spectral feature set of the combustion flame in different combustion stages, selects the spectral features with the information density greater than a preset information density threshold and the density difference less than a preset density difference threshold to obtain a standard spectral feature set of the combustion flame in different combustion stages, and transmits the standard spectral feature set of the combustion flame in different combustion stages to the fuel identification unit (5).
2. The intelligent fuel identification system based on spectral recognition as claimed in claim 1, wherein: 3. The intelligent fuel identification system based on spectral recognition as claimed in claim 2, wherein: The fuel recognition unit (5) receives the standard spectral feature set of the combustion flame at different combustion stages transmitted by the feature selection unit (4), and inputs the standard spectral feature set of the combustion flame at different combustion stages into the trained fuel recognition model, and outputs the fuel category corresponding to the standard spectral feature set of the combustion flame at different combustion stages through the trained fuel recognition model, that is, the fuel category corresponding to the target fuel, the fuel recognition model adopts an improved support vector machine model, and the support vector machine model is improved through an optimization algorithm to obtain the fuel recognition model, and the fuel category corresponding to the target fuel is transmitted to the intelligent display unit (6).
4. The intelligent fuel identification system based on spectral recognition as claimed in claim 3, wherein: The intelligent display unit (6) receives the fuel category corresponding to the target fuel transmitted by the fuel recognition unit (5), and displays the fuel category corresponding to the target fuel through an intelligent display terminal.
5. The intelligent fuel identification system based on spectral recognition as claimed in claim 1, wherein: The envelope line of the full spectrum signal data of the combustion flame at different combustion stages is removed to obtain the standard full spectrum signal data of the combustion flame at different combustion stages, including: The first derivative of the full spectrum signal is derived to obtain the point corresponding to the zero first derivative of the full spectrum signal, and the second derivative of the point corresponding to the zero first derivative is derived to obtain the point corresponding to the zero second derivative of the full spectrum signal, if the second derivative is less than zero, the point is a local maximum point, and all the local maximum points in the full spectrum signal are counted to obtain a local maximum point set; Each local maximum point in the local maximum point set is compared to obtain the maximum extreme point, and the coordinates corresponding to the maximum extreme point are obtained, and the maximum extreme point calculation formula is as follows: wherein E represents the maximum extreme point, represents the i-th local maximum point in the local maximum point set, and n represents the total number of local maximum points in the local maximum point set. Based on the slope of each point in the full spectrum signal corresponding to the maximum extreme point, the slope calculation formula of each point is as follows: wherein K i represents the slope corresponding to the i-th point in the full spectrum signal, S E represents the coordinate corresponding to the maximum extreme point, S i represents the coordinate corresponding to the i-th point in the full spectrum signal.
6. The intelligent fuel identification system based on spectral recognition as claimed in claim 5, wherein: The envelope line of the full spectrum signal data of the combustion flame at different combustion stages is removed to obtain the standard full spectrum signal data of the combustion flame at different combustion stages, further including: The first point set and the second point set of the full spectrum signal are calculated, and the first point set and the second point set calculation formula is as follows: wherein Q i represents the tangent point corresponding to the slope of the i-th point in the first point set on the full spectrum signal, P i represents the tangent point corresponding to the slope of the i-th point in the second point set on the full spectrum signal. Based on the first point set, the second point set and the maximum extreme point, an envelope point set is obtained, and the envelope line of the full spectrum signal is removed based on the envelope point set, and the envelope point set is as follows: L = {P1,..., P i , E, Q E+1 ,..., Q i}; Wherein, L represents the envelope point set.
7. The intelligent fuel identification system based on spectral recognition as claimed in claim 1, wherein: The emissivity expression of each wave band is as follows: wherein a(v) represents the emissivity of each wavelength band, Rad represents the radiation intensity value, I b (v, T) represents the radiation intensity value under the same conditions; The absorption coefficient expression of each component is as follows: k(v ij ) = ∑S ij ·f L (v ij ) where k(v ij ) represents the absorption coefficient of each component, S ij represents the spectral line intensity, f L represents the spectral line profile, v ij represents the wave number of the characteristic waveband; The fluctuation rate expression is as follows: Wherein, R represents the combustion flame fluctuation rate, X represents the combustion flame radiation intensity, MAX(X) represents the maximum value of the combustion flame radiation intensity, MIN(X) represents the minimum value of the combustion flame radiation intensity, and MEAN(X) represents the average value of the combustion flame radiation intensity; The specific variance expression is as follows: Wherein, B represents the combustion flame specific variance, D represents the variance of the combustion flame radiation intensity, and u represents the mathematical expectation of the combustion flame radiation intensity. The flicker frequency expression is as follows: Wherein, F represents the flicker frequency, P(i) represents the power spectrum value of the i-th point in the full spectrum signal, and f(i) represents the frequency of the i-th point in the full spectrum signal.
8. The intelligent fuel identification system based on spectral recognition as claimed in claim 2, wherein: The information density calculation formula between the respective spectral features is as follows: wherein p M represents the information density between each spectral feature, Y M represents the spectral feature of the Mth data in the spectral feature set of the combustion flame at different combustion stages, H(Y M ) represents the information entropy of the Mth spectral feature, and p represents the total number of spectral features in the spectral feature set of the combustion flame at different combustion stages. The density difference calculation formula between the respective spectral features is as follows: where J M―N represents the density difference between each spectral feature, Y N represents the spectral feature of the Nth data in the spectral feature set of the combustion flame at different combustion stages, H(Y N ) represents the information entropy of the Nth spectral feature.
9. The intelligent fuel identification system based on spectral recognition as claimed in claim 3, wherein: The support vector machine model is improved by an optimization algorithm to obtain the fuel identification model, including: The parameters to be optimized of the support vector machine model are mapped to the individual positions of the optimization algorithm, the parameters to be optimized include the parameters of the kernel function and the penalty parameter, the classification error rate of the support vector machine model is taken as the fitness value of the optimization algorithm, the individual population and the maximum iteration number of the optimization algorithm are initialized; Each individual performs exploration operation and development operation, and judges the next behavior operation according to the results of the exploration operation and the development operation, the next behavior operation includes exploration operation and development operation, after a round of search, according to the fitness value of each individual obtained, the optimal fitness value is stored in the optimal solution individual position; In the iterative optimization process, the iteration is performed according to the fitness value of the individual, whether the iteration number meets the maximum iteration number is judged, if it meets, the iteration is terminated and the parameters corresponding to the best individual position are output, the optimal parameter to be optimized is output, and the support vector machine model is improved based on the optimal parameter to be optimized, otherwise, the above operation is repeatedly executed.
10. The intelligent fuel identification system based on spectral recognition as claimed in claim 9, wherein: The position update formula of the exploration operation is as follows: X(t+1) = X(t) * sigma + r(Xbest - I * X(t)); Wherein, X(t+1) and X(t) represent the positions of the individual at the t+1th and tth iterations of the exploration operation, sigma represents the wavelet function disturbance factor, r represents a random number in the interval [0, 1], Xbest represents the optimal solution individual position, and I represents a random number in the interval [0, 2]; The position update formula of the development operation is as follows: wherein, and denote the position of the individual at the (t+1)th and tth iteration of the development operation, ub local and lb local denote the upper and lower bounds of the jth dimension variable updated with the iteration number, Rmv denotes the interaction between the current individual and the random individual, bf denotes the benefit factor, randomly selected 1 or 2.