CN113705100A - Gas detection method based on combination of temperature modulation detection and PSO-SVM algorithm - Google Patents
Gas detection method based on combination of temperature modulation detection and PSO-SVM algorithm Download PDFInfo
- Publication number
- CN113705100A CN113705100A CN202111008440.8A CN202111008440A CN113705100A CN 113705100 A CN113705100 A CN 113705100A CN 202111008440 A CN202111008440 A CN 202111008440A CN 113705100 A CN113705100 A CN 113705100A
- Authority
- CN
- China
- Prior art keywords
- gas
- temperature modulation
- pso
- signal
- svm algorithm
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 68
- 238000001514 detection method Methods 0.000 title claims abstract description 56
- 238000000034 method Methods 0.000 claims abstract description 36
- 238000012545 processing Methods 0.000 claims abstract description 30
- 238000007781 pre-processing Methods 0.000 claims abstract description 17
- 238000000605 extraction Methods 0.000 claims abstract description 16
- 230000008569 process Effects 0.000 claims abstract description 13
- 239000007789 gas Substances 0.000 claims description 171
- 230000004044 response Effects 0.000 claims description 49
- 238000000354 decomposition reaction Methods 0.000 claims description 29
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 claims description 27
- 238000010438 heat treatment Methods 0.000 claims description 10
- 238000001914 filtration Methods 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 9
- 239000004065 semiconductor Substances 0.000 claims description 9
- 238000001228 spectrum Methods 0.000 claims description 8
- UGFAIRIUMAVXCW-UHFFFAOYSA-N Carbon monoxide Chemical compound [O+]#[C-] UGFAIRIUMAVXCW-UHFFFAOYSA-N 0.000 claims description 7
- 229910002091 carbon monoxide Inorganic materials 0.000 claims description 7
- 238000006243 chemical reaction Methods 0.000 claims description 7
- 238000012216 screening Methods 0.000 claims description 6
- 239000000126 substance Substances 0.000 claims description 4
- 238000013499 data model Methods 0.000 claims description 3
- 239000000463 material Substances 0.000 claims description 3
- 238000000513 principal component analysis Methods 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 238000013461 design Methods 0.000 claims description 2
- 238000003909 pattern recognition Methods 0.000 abstract description 8
- 238000005457 optimization Methods 0.000 description 8
- 238000004458 analytical method Methods 0.000 description 7
- 230000009466 transformation Effects 0.000 description 7
- 238000012360 testing method Methods 0.000 description 6
- 230000008859 change Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 239000002245 particle Substances 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- LFQSCWFLJHTTHZ-UHFFFAOYSA-N Ethanol Chemical compound CCO LFQSCWFLJHTTHZ-UHFFFAOYSA-N 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 238000013519 translation Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012847 principal component analysis method Methods 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000001179 sorption measurement Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/0004—Gaseous mixtures, e.g. polluted air
- G01N33/0009—General constructional details of gas analysers, e.g. portable test equipment
- G01N33/0062—General constructional details of gas analysers, e.g. portable test equipment concerning the measuring method or the display, e.g. intermittent measurement or digital display
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Computing Systems (AREA)
- Medical Informatics (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Biology (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Molecular Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Combustion & Propulsion (AREA)
- Food Science & Technology (AREA)
- Medicinal Chemistry (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Investigating Or Analyzing Materials By The Use Of Fluid Adsorption Or Reactions (AREA)
Abstract
A gas detection method based on the combination of temperature modulation detection and a PSO-SVM algorithm belongs to the field of gas detection methods. The algorithm involved in the existing gas detection process has the disadvantages of large operation amount caused by high feature extraction dimensionality and low gas detection accuracy. A gas detection method based on the combination of temperature modulation detection and PSO-SVM algorithm is characterized in that a temperature modulation system is designed, and a gas dynamic signal is collected by the temperature modulation system; preprocessing the acquired gas dynamic signal, and extracting gas characteristic parameters; and processing the characteristic parameters by adopting a PSO-SVM algorithm to obtain the component and concentration information of the gas to be detected. The invention reduces the dimension of the input sample of the subsequent pattern recognition algorithm, reduces the operation amount and has high detection accuracy.
Description
Technical Field
The invention relates to a gas detection method based on the combination of temperature modulation detection and a PSO-SVM algorithm.
Background
Although various intelligent algorithms such as artificial intelligence and mode recognition are continuously improved and optimized in recent years, and are used for gas sensor signal processing to overcome sensor cross sensitivity and improve sensor selectivity, the sensor data processing algorithm generally has the defects of high complexity, large limitation, high application cost and the like, and the algorithm usually needs to analyze a large number of sensor experiment data samples, thereby causing the complexity of the experiment process. Therefore, how to adopt a simpler, more convenient and more effective data processing algorithm of the gas sensor, improve the selectivity of the sensor, and realize accurate detection of gas components and content is still a difficult problem to overcome.
The algorithm involved in the existing gas detection process has the problems of large operation amount caused by high feature extraction dimensionality and low gas detection accuracy.
Disclosure of Invention
The invention aims to solve the problems of large operation amount caused by high feature extraction dimensionality and low gas detection accuracy of an algorithm involved in the existing gas detection process, and provides a gas detection method based on the combination of temperature modulation detection and a PSO-SVM algorithm.
A gas detection method based on temperature modulation detection and PSO-SVM algorithm combination is realized by the following steps:
designing a temperature modulation system, and acquiring a gas dynamic signal by using the temperature modulation system;
preprocessing the acquired gas dynamic signal, and extracting gas characteristic parameters;
and thirdly, processing the characteristic parameters by adopting a PSO-SVM algorithm to obtain the component and concentration information of the gas to be detected.
Preferably, the designing of the temperature modulation system of the first step includes: the device comprises a sensor module, a gas distribution module, a temperature modulation module and a signal acquisition module; wherein the content of the first and second substances,
the sensor module is used for realizing the reaction between the gas-sensitive material and gas and generating information to be detected;
the gas distribution module is used for controlling input parameters of the gas to be detected and setting the concentration of the gas to be detected and mixed gas parameters;
the temperature modulation module is used for controlling any numerical value of heating voltage, heating waveform and heating period through the programmable power supply to realize the setting of various temperature modulation modes;
and the signal acquisition module comprises a data acquisition card and an acquisition display module, and is used for acquiring the response voltage value of the sensor module through the data acquisition card, converting the response voltage value into digital quantity and transmitting the digital quantity to the acquisition display module for real-time display.
Preferably, the second step of preprocessing the collected gas dynamic signal, and the process of extracting the gas characteristic parameters comprises,
preprocessing the acquired gas dynamic response signal data, wherein the preprocessing comprises mean value filtering, normalization and feature extraction; the mean value filtering processing is adopted, so that the response curve of the sensor is smoother, and the interference caused by background noise is removed or reduced; defining the response value of the sensor module between [0, 1] through normalization processing; the characteristic extraction processing is realized by adopting a wavelet decomposition method, a time-frequency spectrum of a signal is obtained through wavelet decomposition, and then partial wavelet coefficients are extracted to be used as characteristic parameters.
Preferably, the process of processing the characteristic parameters by using the PSO-SVM algorithm to obtain the component and concentration information of the gas to be detected in the step three comprises the following steps of,
determining an optimal temperature modulation mode;
and simultaneously, performing dimensionality reduction selection on the extracted characteristic parameters by utilizing principal component analysis, constructing a data model for the selected characteristic parameters, and selecting a PSO-SVM algorithm to perform qualitative identification and quantitative regression on the gas to be detected to obtain the component and concentration information of the gas to be detected.
Preferably, the wavelet decomposition process to obtain the time-frequency spectrum of the signal includes performing multi-layer decomposition on different gas response signals, extracting signal wavelet coefficients as the primary features of the gas, reconstructing the signal to eliminate partial noise, and screening the low-frequency wavelet coefficients extracted after decomposition by PCA, wherein only principal elements with the first three contribution rates are selected as the features of the gas during screening; and then processing the different preprocessed gas data in each period, wherein the obtained amplitude ratio is the initial characteristic coefficient of the gas.
Preferably, the modulation value controlled by the temperature modulation module is that the temperature modulation frequency range is 0-100 Hz, the output signal amplitude range is 0-5V, and the output power range is 0-2W.
Preferably, the method is applicable to the identification of methane and carbon monoxide gases.
Preferably, the sensor module is an MQ series semiconductor sensor.
The invention has the beneficial effects that:
the invention designs a system specially used for detecting gas, and can effectively improve the selectivity and stability of a semiconductor gas sensor and reduce the power consumption of the system by adding a temperature modulation technology, but the optimal temperature modulation modes of different detected gases are different and are controlled by parameters such as waveforms, frequencies, amplitudes and offsets of the different detected gases. The method comprises the steps of carrying out multiple temperature modulation mode tests on a semiconductor gas sensor, carrying out feature extraction on dynamic signals by using an amplitude ratio and discrete wavelet transformation, and carrying out mode identification on extracted information to realize qualitative identification and quantitative regression of gas.
The method comprises the steps of carrying out multi-layer decomposition on different gas response signals, then carrying out signal reconstruction to eliminate partial noise, screening low-frequency wavelet coefficients extracted after decomposition through PCA, and only selecting principal elements with the first three contribution rates as the characteristics of the gas, so that the uncertainty of gas characteristic selection is solved, the dimension of input samples of a subsequent pattern recognition algorithm is reduced, and the operation amount is reduced.
And the prediction accuracy is improved by multi-algorithm fusion and selection of proper hyper-parameters for the mode identification algorithm.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a flow chart of the process of the present invention for methane and carbon monoxide gas identification;
FIG. 3 is a graph of the real component of the dynamic signal of the methane gas sensor after FFT conversion under the modulation of rectangular wave with a period of 10S according to the present invention;
FIG. 4 is a graph of the real component of the dynamic signal of the methane gas sensor after FFT conversion under the modulation of rectangular wave with a period of 20S according to the present invention;
FIG. 5 is a graph of the real component of the dynamic signal of the methane gas sensor after FFT conversion under the rectangular wave modulation with the period of 40S according to the present invention;
FIG. 6 is a schematic diagram of a six-layer wavelet decomposition tree according to the present invention;
FIG. 7 is a flow chart of an SVM algorithm optimized using PSO in accordance with the present invention.
Detailed Description
The first embodiment is as follows:
in the present embodiment, a gas detection method based on temperature modulation detection combined with a PSO-SVM algorithm is implemented by the following steps, as shown in fig. 1:
the method is realized by the following steps:
designing a temperature modulation system, and acquiring a gas dynamic signal by using the temperature modulation system;
preprocessing the acquired gas dynamic signal, and extracting gas characteristic parameters;
and thirdly, processing the characteristic parameters by adopting a PSO-SVM algorithm to obtain the component and concentration information of the gas to be detected.
The second embodiment is as follows:
different from the first embodiment, in the first embodiment, the gas detection method based on the combination of temperature modulation detection and PSO-SVM algorithm includes: the device comprises a sensor module, a gas distribution module, a temperature modulation module and a signal acquisition module; wherein the content of the first and second substances,
the sensor module is used for realizing the reaction between the gas-sensitive material and gas and generating information to be detected and is the most main part of the temperature modulation system;
the gas distribution module is used for controlling input parameters of the gas to be detected and setting the concentration of the gas to be detected and mixed gas parameters;
the temperature modulation module is used for controlling any numerical value of heating voltage, heating waveform and heating period through the programmable power supply to realize the setting of various temperature modulation modes;
and the signal acquisition module comprises a data acquisition card and an acquisition display module, and is used for acquiring the response voltage value of the sensor module through the data acquisition card, converting the response voltage value into digital quantity and transmitting the digital quantity to the acquisition display module for real-time display. The data acquisition card acquires and stores the dynamic response signals of the sensor, and the sampling time interval of the signals of the sensor is 0.5 s.
The third concrete implementation mode:
different from the first specific embodiment, in the gas detection method based on the combination of temperature modulation detection and PSO-SVM algorithm of the present embodiment, the second step of preprocessing the acquired gas dynamic signal and extracting the gas characteristic parameters includes,
preprocessing the acquired gas dynamic response signal data, wherein the preprocessing comprises mean value filtering, normalization and feature extraction; the mean value filtering processing is adopted, so that the response curve of the sensor is smoother, and the interference caused by background noise is removed or reduced; through normalization processing, the response value of the sensor module is limited between [0 and 1], so that the convergence analysis of the algorithm is facilitated, the complexity is reduced, and the algorithm efficiency is improved; the characteristic extraction processing is realized by adopting a wavelet decomposition method, a time-frequency spectrum of a signal is obtained through wavelet decomposition, then partial wavelet coefficients are extracted to be used as characteristic parameters, and meanwhile, the amplitude ratio is selected to be used as the characteristic parameters to be compared with a wavelet characteristic extraction algorithm.
The fourth concrete implementation mode:
different from the first embodiment, in the gas detection method based on the combination of temperature modulation detection and PSO-SVM algorithm of the present embodiment, the process of processing the characteristic parameters by using the PSO-SVM algorithm to obtain the component and concentration information of the gas to be detected in the third step is,
determining an optimal temperature modulation mode, and realizing optimization of the temperature modulation mode of the semiconductor sensor;
and simultaneously, performing dimensionality reduction selection on the extracted characteristic parameters by utilizing principal component analysis, constructing a data model for the selected characteristic parameters, and selecting a PSO-SVM algorithm to perform qualitative identification and quantitative regression on the gas to be detected to obtain the component and concentration information of the gas to be detected.
The fifth concrete implementation mode:
different from the specific embodiment, the gas detection method based on the combination of temperature modulation detection and the PSO-SVM algorithm in the embodiment is characterized in that the process of obtaining the time-frequency spectrum of signals by wavelet decomposition comprises the steps of performing multi-layer decomposition on different gas response signals, extracting signal wavelet coefficients to serve as preliminary characteristics of gas, reconstructing the signals to eliminate partial noise, screening the low-frequency wavelet coefficients extracted after decomposition through PCA, and selecting only principal elements with the first three contribution rates as characteristics of the gas during screening; the method not only solves the uncertainty of gas characteristic selection, but also reduces the dimension of the input sample of the subsequent pattern recognition algorithm and reduces the operation amount. And then processing the different preprocessed gas data in each period, wherein the obtained amplitude ratio is the initial characteristic coefficient of the gas.
The sixth specific implementation mode:
different from the first specific embodiment, in the gas detection method based on the combination of temperature modulation detection and PSO-SVM algorithm of the present embodiment, the modulation value controlled by the temperature modulation module is a temperature modulation waveform output circuit in which the temperature modulation frequency range is 0 to 100Hz, the output signal amplitude range is 0 to 5V, the output power range is 0 to 2W, and parameters such as the amplitude, the frequency, the square wave duty ratio and the like of the waveform are adjustable.
The seventh embodiment:
different from the specific embodiment, the gas detection method based on temperature modulation detection and the PSO-SVM algorithm is applicable to the identification of methane and carbon monoxide gas.
The specific implementation mode is eight:
different from the specific implementation way, in the gas detection method based on the combination of the temperature modulation detection and the PSO-SVM algorithm, the sensor module is an MQ series semiconductor sensor, and the MQ series semiconductor sensor is produced by Chinese weisheng and has excellent performance, so that the method is widely applied to the field of electronic nose research at present. The resistance value and the concentration of the gas to be detected are in an exponential relation, the chemical and physical reaction between the sensitive layer and the gas to be detected is reversible, the adsorption time is fast, the recovery time is short, the stability is good, and the gas test can be continuously carried out.
Example (b):
the method can be used for determining the optimal temperature modulation mode suitable for the carbon monoxide and the methane gas, and further optimizing the temperature modulation mode of the semiconductor sensor.
Firstly, the method of the invention is summarized as follows:
the method comprises three steps of designing a temperature modulation system, and preprocessing and identifying the acquired gas dynamic signal. The temperature modulation system comprises a sensor module, a gas distribution module, a temperature modulation module and a signal acquisition module; the working temperature of the gas sensor is modulated in a certain temperature mode, then the collected dynamic response signals are simply preprocessed, and then the characteristic parameters of the dynamic gas information are extracted, the characteristic values are further extracted and sent to a mode recognition system, and the type and concentration information of the gas is obtained. As shown in fig. 2.
1. A temperature modulation module:
the hardware circuit system of the temperature modulation module mainly comprises a temperature modulation waveform output circuit, a sensor signal detection circuit and a human-computer interaction interface. The temperature modulation waveform circuit outputs four modulation waveforms of sine wave, square wave, triangular wave and sawtooth wave. The frequency range of the modulation voltage required by the semiconductor gas sensor is 0.001 Hz-100 Hz, and the modulation voltage can meet the requirement of heating power consumption required by the gas sensor. Therefore, the temperature modulation waveform output circuit is designed, wherein the temperature modulation frequency is 0-100 Hz, the amplitude of an output signal is 0-5V, the output power is 0-2W, and parameters such as the amplitude, the frequency and the square wave duty ratio of a waveform can be adjusted.
The sensor signal detection circuit is mainly used for collecting, processing and storing response signals of the gas sensor under temperature modulation so as to further analyze the signal characteristics of data.
The man-machine interaction mainly comprises a matrix keyboard parameter input circuit, a system data liquid crystal display circuit and an upper computer control and display interface. The bridge function between the temperature modulation system and a user is mainly realized, and the operability and convenience of the system are enhanced.
2. Preprocessing the acquired gas dynamic signal:
after the initial sample is obtained, preprocessing of the data is also required, which is a necessary step prior to pattern recognition. The preprocessing comprises mean filtering, normalization and feature extraction. The mean filtering can make the response curve of the sensor smoother and remove or reduce the interference caused by background noise. The normalization can limit the sensor response value between [0, 1], thereby being beneficial to the convergence analysis of the algorithm, reducing the complexity and improving the efficiency of the algorithm. The characteristic extraction is to extract characteristic points which can represent gas modes most from a large number of sensor response characteristics as the input of a subsequent mode recognition model, so that the accuracy and the efficiency of mode recognition are improved.
The gas data is first carded. The gas response data under the same waveform modulation mode (same temperature, waveform, period) is classified by period. When the gas to be detected is just introduced into the gas testing cavity, the amplitude change of the gas response curve is large, so that the gas response characteristics are effectively extracted, stable gas response data obtained by testing are taken to be processed after the gas sensor responds stably, and wavelet decomposition is selected to preprocess dynamic response signals. The wavelet decomposition is used for decomposing different gas data of each period after filtering and normalization, and a signal wavelet coefficient is extracted to serve as the initial characteristic of the gas. And then processing the different preprocessed gas data in each period, wherein the obtained amplitude ratio is the initial characteristic coefficient of the gas.
3. Processing the characteristic parameters by adopting a PSO-SVM algorithm to obtain the component and concentration information of the gas to be detected
After the characteristics of the dynamic response signals of the gas after temperature modulation are extracted, a principal component analysis method is selected to reduce the dimension of the characteristic parameters, and the extracted characteristic values are sent to a pattern recognition system to carry out classification recognition on the gas. The method comprises the steps of respectively selecting SVM, PSO-SVM and PSO-SVR to analyze and predict gas characteristic samples extracted after temperature modulation, comparing an amplitude ratio with a time-frequency domain characteristic extraction method (DWT), and finally realizing optimization of optimal temperature modulation modes of methane and carbon monoxide.
Preprocessing an acquired gas dynamic signal, and extracting gas characteristic parameters by adopting a fast Fourier transform omega; the sensor dynamic response signals in the mixed gas are preprocessed and different from the response signals in the clean air, and the difference of the waveforms is difficult to describe in a time domain view, so that the difference can be analyzed by seeking the representation of the difference in a transform domain. Fourier transform is the most common and simple frequency analysis method. Therefore, dynamic processing generally converts the signal in the time domain into the frequency domain, examines the frequency domain characteristics of the signal, and extracts effective characteristic parameters in the frequency domain as the input of pattern recognition, thereby achieving the purpose of gas detection and recognition. The specific principle is as follows:
let the one-dimensional discrete signal x (N) be a sequence of N bit periods whose fourier transform is:
in the formula, X (k ω)0) A Fourier transform representing the signal x (n);
n represents a time series;
k represents a frequency discrete variable;
kω0digital frequency representing k harmonicsThe ratio of the total weight of the particles,
the fourier transform is a frequency domain representation corresponding to the time domain waveform of the signal, and expresses the change rule of the signal in the frequency domain structure. The fourier spectrum of a signal can be expressed by amplitude characteristics and phase characteristics in a polar coordinate system, and also by a cosine component as a real part of a complex vector and a sine component as an imaginary part, i.e., by a direct coordinate system.
The fourier transform is a tool for converting time domain and frequency domain into each other, and in a physical sense, the essence of the fourier transform is to say that f (t) this waveform is decomposed into a superposition sum of a plurality of sinusoids of different frequencies. Thus, we convert the study of the primitive function f (t) into a study of its weight coefficients, i.e. spectral values. The standard basis of the fourier transform is composed of a sine wave and its higher harmonics, so it is localized in the frequency domain. The fourier transform is able to relate the time and frequency domain features of a signal, viewed from the time and frequency domain of the signal, respectively, but is not able to combine them organically.
In order to deeply analyze the dynamic response signal, the dynamic response of the gas sensor is processed by applying fast Fourier transform to obtain a frequency domain representation corresponding to the dynamic response signal, and the change rule of the frequency spectrum structure is analyzed according to a response frequency function, so that effective characteristics are extracted. This discussion is based on the real and imaginary parts of the signal.
FIG. 3 is a graph of real components of a methane gas sensor dynamic signal modulated by a square wave with a period of 10S after FFT; since the sampling frequency is 2Hz, the frequency range of the calculated signal components is between-1, 1 Hz. The frequency interval is 0.1, and the real and imaginary values of f ═ 0, 0.1, 0.2, 0.3, 0.4Hz are extracted. Fig. 4 is a real component diagram of a methane gas sensor dynamic signal under 20S square wave modulation, which is subjected to FFT transformation, the frequency interval is 0.05, and real and imaginary values of 0, 0.05, 0.15, 0.25 and 0.35Hz are extracted. Fig. 5 is a real component diagram of a methane gas sensor dynamic signal under rectangular wave modulation with a period of 40S after FFT transformation, the frequency interval is 0.025, and real and imaginary values of 0, 0.025, 0.05, 0.075, and 0.1Hz are extracted.
Wavelet transformation:
wavelet Transform (WT) is one of the most common signal processing tools at present, and is widely applied to multiple subjects such as mathematics, physics, computer science, signal and information processing, image processing, seismic exploration and the like. It can analyze the local characteristics of the time (spatial) and frequency (scale) domains of a signal, called a signal "microscope". The specific principle is as follows:
the WT may transform the original signal to the wavelet domain and thus the information contained in the original signal data may be represented by a series of wavelet coefficients. Wherein each wavelet coefficient is composed of wavelet subfunctions, and the wavelet subfunctions are converted from wavelet basis functions. The wavelet basis function is also referred to as the wavelet mother function ψ (t). In short, a wavelet subfunction is obtained by expanding and translating a wavelet mother function, which can be expressed as:
wherein a and b are translation factor and scale factor respectively. By adjusting the scaling factor a and the translation factor b, wavelets with different time-frequency widths can be obtained to match any local position in the original signal, so that the purpose of analyzing and researching the time-frequency localization of the signal is achieved. However, in actual engineering operation, the WT is operated by a computer, and therefore, it is necessary to perform discretization on information data before processing. The expression form of the discretization wavelet transformation coefficient is defined as follows:
discrete Wavelet Transform (DWT) is carried out on signals, and not only is the detail information contained in the signals not lost, but also the problem of redundancy between two points of the signals in a Wavelet space caused by WT is eliminatedTo give a title. Because of the orthogonality of the wavelet basis functions, the signal transformation has smaller calculation error, so that the transformation result time-frequency function can better reflect the self characteristics of the original signal. An effective way to perform the discrete wavelet transform is to use a filter, i.e. the Mallat algorithm. FIG. 6 shows a six-layer wavelet decomposition tree diagram, where S is an input signal, the input signal passes through two complementary filter banks (one filter is a low-pass filter and one filter is a high-pass filter), each node of the tree structure represents a coefficient of the signal after discrete wavelet transform, where aiIs a low frequency coefficient of the signal, diThe decomposition relationship for the high frequency coefficients of the signal (i is the number of decomposition levels) is: s-a 6+ d1+ d2+ d3+ d4+ d5+ d 6.
To analyze the gas dynamic response signal in depth, the gas dynamic response signal is DWT. As can be seen from the above section, the first,
the discrete wavelet transform can effectively decompose an original signal into a high-frequency part and a low-frequency part, and a plurality of low-frequency wavelet coefficients can be extracted from a gas dynamic response signal DWT by extracting the high-frequency wavelet coefficient or the low-frequency wavelet coefficient of the gas signal for analysis, wherein the method comprises the following specific steps:
step 1: the gas data is first carded. The gas response data under the same temperature modulation mode (same temperature, waveform, period) is classified by period. When the gas to be detected is just introduced into the gas testing cavity, the amplitude change of the gas response curve is large, so that the gas response characteristics are effectively extracted, and only the last 200s gas response data obtained by testing is taken for processing after the gas sensor has stable response.
Step 2: and (5) wavelet decomposition. And performing wavelet decomposition and reconstruction on the preprocessed different gas data of each period in turn, and extracting a signal wavelet coefficient as the initial characteristic of the gas, wherein the selected last 200s data is the stable response data of the gas, so the main energy of the gas response signal is a low-frequency signal, and the wavelet coefficient extracted herein is the low-frequency wavelet coefficient of the signal. The wavelet function is selected from Daubechies4(db4) wavelets with orthogonality, which can not only decompose signals to non-overlapped sub-bands, but also can efficiently perform discrete wavelet transform, and simultaneously has better smoothness and numerical stability, thereby being beneficial to subsequent wavelet analysis.
Analysis based on the pso-svm algorithm:
in the CV sense, grid search (grid search) is used to find the optimal parameters, and although the highest classification accuracy rate in the CV sense, i.e. the global optimal solution, can be found by using grid search, sometimes if it is time-consuming to find the optimal parameters in a larger range, a heuristic algorithm can be used to find the global optimal solution without traversing all parameter points in the grid. This time, the particle swarm optimization algorithm is selected for optimization.
Particle Swarm Optimization (PSO) is another Swarm intelligence-based Optimization algorithm besides ant colony algorithm in the field of computational intelligence, which was originally proposed in 1995 by Kennedy Eber art, and its basic concept is derived from the research on artificial life and bird Swarm predation behavior. PSO is an evolution calculation technology based on group intelligence, compared with GA, PSO has no operations of selection, intersection and variation, and is searched by following an optimal example in a solution space through particles.
And taking the accuracy rate of the training set in the CV sense as a fitness function value in the PSO, and then using the PSO to optimize SVM parameters as an overall algorithm process shown in the figure. FIG. 7 is a flow chart of an SVM algorithm optimized using PSO;
each gas sample is subjected to DWT, low-frequency wavelet coefficients are extracted as features, and 13 wavelet coefficients are selected. Table 1 shows the PSO-SVM recognition rate and training rate of the sensor under sine wave modulation at three modulation temperatures (2-5V, 3-5V and 4-5V) in 3 periods (T is 10s, 20s and 40s) for three gases of methane, carbon monoxide and ethanol.
TABLE 1 SVM Algorithm recognition accuracy under Small wavelength decomposition
As can be seen from Table 2, the SVM algorithm after PSO optimization has higher accuracy and more stable recognition, wherein the prediction rate of the gas under the temperature modulation modes of 2-5V, 20S and 40S is 93.75% at most.
TABLE 2 SVM algorithm recognition accuracy under small wavelength decomposition
As can be seen from Table 2, the SVM algorithm after PSO optimization has higher accuracy and more stable recognition, but the accuracy of wavelet decomposition extraction characteristic value combined with algorithm prediction gas is higher than the accuracy of fast Fourier transform extraction characteristic value combined with algorithm prediction gas in comprehensive comparison.
Second, feasibility analysis
CO and CH4 gas are used as the gas to be detected, the voltage is 4.5-5V, the signal acquired by the gas sensor after square wave modulation is used as a dynamic response signal, the acquired dynamic response is subjected to wavelet decomposition, the extracted characteristic is subjected to pattern recognition by using an SVM algorithm, the accuracy rate reaches 87.5%, and the qualitative recognition of the gas is completed. And the amplitude ratio is taken as a characteristic value, and the pattern recognition is carried out by combining an SVM algorithm, so that the accuracy rate reaches over 90 percent.
The above is only a preferred embodiment of the present invention, and is not intended to limit the present invention, and various modifications and changes may be made to the present invention by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (8)
1. A gas detection method based on the combination of temperature modulation detection and a PSO-SVM algorithm is characterized in that: the method is realized by the following steps:
designing a temperature modulation system, and acquiring a gas dynamic signal by using the temperature modulation system;
preprocessing the acquired gas dynamic signal, and extracting gas characteristic parameters;
and thirdly, processing the characteristic parameters by adopting a PSO-SVM algorithm to obtain the component and concentration information of the gas to be detected.
2. The method of claim 1, wherein the method comprises a temperature modulation based detection method combined with a PSO-SVM algorithm, wherein the method comprises: the design temperature modulation system of step one includes: the device comprises a sensor module, a gas distribution module, a temperature modulation module and a signal acquisition module; wherein the content of the first and second substances,
the sensor module is used for realizing the reaction between the gas-sensitive material and gas and generating information to be detected;
the gas distribution module is used for controlling input parameters of the gas to be detected and setting the concentration of the gas to be detected and mixed gas parameters;
the temperature modulation module is used for controlling any numerical value of heating voltage, heating waveform and heating period through the programmable power supply to realize the setting of various temperature modulation modes;
and the signal acquisition module comprises a data acquisition card and an acquisition display module, and is used for acquiring the response voltage value of the sensor module through the data acquisition card, converting the response voltage value into digital quantity and transmitting the digital quantity to the acquisition display module for real-time display.
3. A gas detection method based on temperature modulation detection combined with PSO-SVM algorithm according to claim 1 or 2, characterized in that: the second step is that the collected gas dynamic signal is preprocessed, the process of extracting the gas characteristic parameters is that,
preprocessing the acquired gas dynamic response signal data, wherein the preprocessing comprises mean value filtering, normalization and feature extraction; the mean value filtering processing is adopted, so that the response curve of the sensor is smoother, and the interference caused by background noise is removed or reduced; defining the response value of the sensor module between [0, 1] through normalization processing; the characteristic extraction processing is realized by adopting a wavelet decomposition method, a time-frequency spectrum of a signal is obtained through wavelet decomposition, and then partial wavelet coefficients are extracted to be used as characteristic parameters.
4. A gas detection method based on temperature modulation detection combined with PSO-SVM algorithm as claimed in claim 3 wherein: the process of processing the characteristic parameters by adopting the PSO-SVM algorithm to obtain the component and concentration information of the gas to be detected comprises the following steps,
determining an optimal temperature modulation mode;
and simultaneously, performing dimensionality reduction selection on the extracted characteristic parameters by utilizing principal component analysis, constructing a data model for the selected characteristic parameters, and selecting a PSO-SVM algorithm to perform qualitative identification and quantitative regression on the gas to be detected to obtain the component and concentration information of the gas to be detected.
5. A gas detection method based on temperature modulation detection combined with PSO-SVM algorithm as claimed in claim 3 wherein: the wavelet decomposition is carried out to obtain the time-frequency spectrum of the signal, multilayer decomposition is carried out on different gas response signals, a signal wavelet coefficient is extracted to serve as the primary characteristic of the gas, then signal reconstruction is carried out to eliminate partial noise, the low-frequency wavelet coefficient extracted after decomposition is screened through PCA, and only principal elements with the first three contribution rates are selected to serve as the characteristic of the gas during screening; and then processing the different preprocessed gas data in each period, wherein the obtained amplitude ratio is the initial characteristic coefficient of the gas.
6. A gas detection method based on temperature modulation detection combined with PSO-SVM algorithm as claimed in claim 2, characterized in that: the temperature modulation module controls the modulation value of 0-100 Hz of temperature modulation frequency range, 0-5V of output signal amplitude range and 0-2W of output power range.
7. A method for gas detection based on temperature modulation detection combined with PSO-SVM algorithm according to claim 4, 5 or 6, characterized in that: the method is suitable for the identification of methane and carbon monoxide gases.
8. The method of claim 7, wherein the method comprises a combination of temperature modulation detection and PSO-SVM algorithm, and further comprising: the sensor module adopts MQ series semiconductor sensors.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111008440.8A CN113705100A (en) | 2021-08-31 | 2021-08-31 | Gas detection method based on combination of temperature modulation detection and PSO-SVM algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111008440.8A CN113705100A (en) | 2021-08-31 | 2021-08-31 | Gas detection method based on combination of temperature modulation detection and PSO-SVM algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113705100A true CN113705100A (en) | 2021-11-26 |
Family
ID=78657310
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111008440.8A Pending CN113705100A (en) | 2021-08-31 | 2021-08-31 | Gas detection method based on combination of temperature modulation detection and PSO-SVM algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113705100A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114166773A (en) * | 2021-12-08 | 2022-03-11 | 中煤科工集团重庆研究院有限公司 | Particle swarm optimization-support vector machine-based NOx measurement method |
| US20220091083A1 (en) * | 2020-09-23 | 2022-03-24 | Ut-Battelle, Llc | Chemical detection system with at least one electronic nose |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170356936A1 (en) * | 2016-05-11 | 2017-12-14 | Mcmaster University | Enhanced system and method for conducting pca analysis on data signals |
| CN111476339A (en) * | 2020-04-20 | 2020-07-31 | 山东师范大学 | Rolling bearing fault feature extraction method, intelligent diagnosis method and system |
| CN113219130A (en) * | 2021-04-16 | 2021-08-06 | 中国农业大学 | Calibration method and test platform of multi-parameter gas sensor |
-
2021
- 2021-08-31 CN CN202111008440.8A patent/CN113705100A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170356936A1 (en) * | 2016-05-11 | 2017-12-14 | Mcmaster University | Enhanced system and method for conducting pca analysis on data signals |
| CN111476339A (en) * | 2020-04-20 | 2020-07-31 | 山东师范大学 | Rolling bearing fault feature extraction method, intelligent diagnosis method and system |
| CN113219130A (en) * | 2021-04-16 | 2021-08-06 | 中国农业大学 | Calibration method and test platform of multi-parameter gas sensor |
Non-Patent Citations (8)
| Title |
|---|
| ZHAO, WJ: "Optimized Low Frequency Temperature Modulation for Improving the Selectivity and Linearity of SnO2Gas Sensor", IEEE SENSORS JOURNAL, vol. 20, no. 18, pages 10433 - 10443, XP011804894, DOI: 10.1109/JSEN.2020.2993055 * |
| 于洋;赵文杰;王欣;王暄;施云波;: "半导体气体传感器动态温度调制***设计及检测方法研究", 传感技术学报, no. 09, pages 1365 - 1371 * |
| 余道洋;戚功美;瞿顶军;李民强;刘锦淮;: "基于SVM和PCA的痕量多组分气体检测方法", 模式识别与人工智能, no. 08, pages 720 - 727 * |
| 李宁;薛亚许;: "电子鼻信号处理方法综述", 电子世界, no. 01 * |
| 李熙;何秀丽;李建平;张阳;: "基于单个SnO_2传感器的CO/H_2混合气体定量分析", 传感技术学报, no. 10 * |
| 葛海峰;林继鹏;刘君华;丁晖;: "基于支持向量机和小波分解的气体识别研究", 仪器仪表学报, no. 06 * |
| 金翠云;崔瑶;王颖;: "粒子群优化的SVM算法在气体分析中的应用", 电子测量与仪器学报, no. 07 * |
| 马泽亮;国婷婷;殷廷家;王志强;杨方旭;李彩虹;李钊;袁文浩;: "基于电子鼻***的白酒掺假检测方法", 食品与发酵工业, no. 02 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20220091083A1 (en) * | 2020-09-23 | 2022-03-24 | Ut-Battelle, Llc | Chemical detection system with at least one electronic nose |
| US11958251B2 (en) | 2020-09-23 | 2024-04-16 | Ut-Battelle, Llc | Additive manufacturing system with at least one electronic nose |
| US11975491B2 (en) * | 2020-09-23 | 2024-05-07 | Ut-Battelle, Llc | Chemical detection system with at least one electronic nose |
| US12005649B2 (en) | 2020-09-23 | 2024-06-11 | Ut-Battelle, Llc | Aroma detection systems for food and beverage and conversion of detected aromas to natural language descriptors |
| CN114166773A (en) * | 2021-12-08 | 2022-03-11 | 中煤科工集团重庆研究院有限公司 | Particle swarm optimization-support vector machine-based NOx measurement method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Omitaomu et al. | Empirical mode decomposition technique with conditional mutual information for denoising operational sensor data | |
| Perunicic et al. | Power quality disturbance detection and classification using wavelets and artificial neural networks | |
| Huang | Hilbert-Huang transform and its applications | |
| Frei et al. | Intrinsic time-scale decomposition: time–frequency–energy analysis and real-time filtering of non-stationary signals | |
| CN113705100A (en) | Gas detection method based on combination of temperature modulation detection and PSO-SVM algorithm | |
| Rao et al. | Hilbert-Huang transform analysis of hydrological and environmental time series | |
| CN102288843B (en) | Power quality disturbance signal detection method | |
| Giannakis et al. | Nonlinear Laplacian spectral analysis: capturing intermittent and low‐frequency spatiotemporal patterns in high‐dimensional data | |
| Kang et al. | Detecting and classifying events in noisy time series | |
| Ekambaram et al. | Wavelet-regularized graph semi-supervised learning | |
| KR102149495B1 (en) | Optimization apparatus for training conditions of environmental prediction model and operating thereof | |
| CN109270345A (en) | A kind of detection method of mains by harmonics signal | |
| CN112215405A (en) | Non-invasive type residential electricity load decomposition method based on DANN domain adaptive learning | |
| Székely et al. | Extraction and predictability of coherent intraseasonal signals in infrared brightness temperature data | |
| Bellisario et al. | Contributions of MIR to soundscape ecology. Part I: Potential methodological synergies | |
| Zhu et al. | Complex disturbances identification: A novel PQDs decomposition and modeling method | |
| Chen et al. | Causality analysis in process control based on denoising and periodicity-removing CCM | |
| Jin et al. | Bearing fault diagnosis based on VMD fuzzy entropy and improved deep belief networks | |
| Li et al. | A new denoising method based on decomposition mixing of hydro-acoustic signal | |
| Li et al. | Magnetotelluric noise suppression via convolutional neural network | |
| Lei et al. | High voltage shunt reactor acoustic signal denoising based on the combination of VMD parameters optimized by coati optimization algorithm and wavelet threshold | |
| CN117634905A (en) | Power grid business data feature extraction method | |
| Safavian et al. | Aquantitative comparison of different mother wavelets for characterizing transients in power systems | |
| Hooten et al. | Shifts in the spatio-temporal growth dynamics of shortleaf pine | |
| van Jaarsveldt et al. | Package AdvEMDpy: Algorithmic variations of empirical mode decomposition in Python |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |