CN110426421B - Device and method for detecting multi-component harmful gas in kitchen environment - Google Patents
Device and method for detecting multi-component harmful gas in kitchen environment Download PDFInfo
- Publication number
- CN110426421B CN110426421B CN201910848479.7A CN201910848479A CN110426421B CN 110426421 B CN110426421 B CN 110426421B CN 201910848479 A CN201910848479 A CN 201910848479A CN 110426421 B CN110426421 B CN 110426421B
- Authority
- CN
- China
- Prior art keywords
- sensor
- voltage
- gas
- detection
- module
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
- G01D21/00—Measuring or testing not otherwise provided for
- G01D21/02—Measuring two or more variables by means not covered by a single other subclass
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N27/00—Investigating or analysing materials by the use of electric, electrochemical, or magnetic means
- G01N27/02—Investigating or analysing materials by the use of electric, electrochemical, or magnetic means by investigating impedance
- G01N27/04—Investigating or analysing materials by the use of electric, electrochemical, or magnetic means by investigating impedance by investigating resistance
- G01N27/12—Investigating or analysing materials by the use of electric, electrochemical, or magnetic means by investigating impedance by investigating resistance of a solid body in dependence upon absorption of a fluid; of a solid body in dependence upon reaction with a fluid, for detecting components in the fluid
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Health & Medical Sciences (AREA)
- Pure & Applied Mathematics (AREA)
- Chemical & Material Sciences (AREA)
- Health & Medical Sciences (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Software Systems (AREA)
- Computational Mathematics (AREA)
- Databases & Information Systems (AREA)
- Electrochemistry (AREA)
- Biochemistry (AREA)
- Probability & Statistics with Applications (AREA)
- Operations Research (AREA)
- Analytical Chemistry (AREA)
- Immunology (AREA)
- Algebra (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Chemical Kinetics & Catalysis (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Pathology (AREA)
- Investigating Or Analyzing Materials By The Use Of Fluid Adsorption Or Reactions (AREA)
Abstract
The invention discloses a device and a method for detecting multi-component harmful gas in a kitchen environment. Under the condition that air is introduced into the air chamber at a constant speed from the outside, the sensor can quickly stabilize and respond and recover a voltage baseline, the output voltage response can be received and processed by the single chip microcomputer, and the concentration detection results of three components such as carbon monoxide, methane and formaldehyde in the mixed gas are output to the upper computer. The device has the advantages of perfect function, small volume, convenient operation, good accuracy, stability and repeatability, and can be used as a core device to be put into production and manufacture of harmful gas detection instruments in the kitchen environment.
Description
Technical Field
The invention relates to the field of harmful gas detection, in particular to a device and a method for detecting multi-component harmful gas in a kitchen environment.
Background
For a long time, the concentration of toxic and harmful gases in the indoor environment is likely to be far above the safety standard in a short time due to various factors such as gas, decoration, geographical location, ventilation habits and the like. The high-concentration toxic and harmful gas has extremely adverse effect on human body, causes discomfort and damage to health if the gas is light, and causes serious accidents such as fire and explosion if the gas is heavy.
The detection of the toxic and harmful gas is usually applied to the industry and outdoor occasions such as tunnels, mines, factories and the like at present, and often only provides the function of alarming, but can not detect the concentration of a plurality of single component gases in the complex gas mixture, so that the increasing requirements of people are difficult to meet. Commercial instruments on the market are not suitable for long-term air quality detection in daily life due to the defects of large volume, high price, inconvenience in operation and the like. Therefore, it is necessary to provide a fast, cheap, convenient and miniaturized toxic and harmful gas detection device and detection technology to further realize intelligent determination of the concentration of each target gas component.
Disclosure of Invention
The invention aims to provide a device and a method for detecting multi-component harmful gas in a kitchen environment, aiming at overcoming the defects of complex operation, high price, large instrument, incapability of multi-component measurement and the like in a common detection means.
The purpose of the invention is realized by the following technical scheme: a multi-component harmful gas detection device in a kitchen environment is characterized by comprising a sample injection module, a power supply module, a gas chamber and sensor module, a signal processing module, an AD conversion module, a single chip microcomputer minimum system module and a serial port module;
the sample injection module comprises an air pump, a drying and filtering pipe and a heating chamber and is used for introducing dried, filtered and temperature-controlled gas into the air chamber and the sensor module at a constant speed; the air pump is used for enabling the flow rate of gas in the sample introduction process and the sensor cleaning process to be consistent and unchanged, and ensuring the relative stability of voltage signals output by the sensor before and after the sensor cleaning process; the interior of the drying filter tube is filled with allochroic silica gel and is used for filtering particulate impurities and water vapor in the sample gas and indicating the performance condition of the silica gel through color change; the air pump, the drying filter pipe and the heating chamber are connected in sequence;
the power supply module is used for providing different required voltages for the sample introduction module and the singlechip minimum system module and supplying power by a +12V switching power supply;
the gas chamber and sensor module comprises a gas chamber and a sensor array, the main body of the gas chamber is of a cuboid structure, a gas detection chamber is arranged above the inner part of the gas chamber, a gas inlet hole is formed in the side surface of the gas detection chamber, the gas inlet hole is connected with a gas inlet pipe, 6 MOS sensor circular grooves are formed in the bottom of the gas chamber and used for fixing and replacing a sensor, a bottom empty groove with the height of 3mm is formed in the bottom of the gas chamber, a gas outlet groove is formed in the side, opposite to the gas inlet hole, of the bottom of the gas chamber and communicated with the gas outlet groove, a temperature and humidity sensor site is formed in the bottom of the gas outlet groove, a gas outlet hole is formed in the side surface of the gas outlet groove and connected with a gas outlet pipe, and gas to be detected is discharged through; the sensor array comprises 6 MOS sensors and 1 temperature and humidity sensor; the MOS sensor is arranged in the MOS sensor circular groove, and the head of the MOS sensor is positioned in the gas detection chamber and is consistent with the height of the gas inlet; the temperature and humidity sensor is arranged at the position of the temperature and humidity sensor at the bottom of the air outlet groove and used for detecting the actual temperature and humidity of the heated air flowing through the sensor array.
The signal processing module adopts a differential amplification circuit and RC filtering and is used for carrying out differential amplification and filtering processing on the output signals of the sensor array, so that the output voltage of the sensor array is stabilized below 1.0V after pure air with a constant flow rate of 1000 ml/min is introduced, and a larger change interval of the output voltage of the sensor is allowed;
the AD conversion module is used for rapidly completing AD conversion of the voltage signal of the sensor after differential amplification and filtering processing, and inputting the voltage signal into the minimum system module of the single chip microcomputer;
the minimum system module of the single chip microcomputer is provided with an algorithm model for judging the concentration of the gas to be detected, and the algorithm model is used for directly carrying out analysis and operation on the received signal on a lower computer to obtain a final result and sending the final result to the upper computer through a serial port module;
the serial port module adopts a USART serial port to transmit a detection result to the upper computer through a serial port-to-USB signal line
Further, the air pump is a miniature vacuum pump for a laboratory, flow rate adjustment is realized through PWM waves, and the air pump runs all the time in the whole detection process; the air pump can also be replaced by an air distribution instrument or other sample injection instruments.
Furthermore, the heating chamber is composed of a copper pipe, a heating plate, heat conducting glue, a fixing frame and the like; the heating chamber is formed by 3D printing of high-performance heat-resistant nylon, and 4 copper pipes with the outer diameter of 6mm and the inner diameter of 4mm can be filled in the heating chamber.
Furthermore, the power supply module adopts voltage stabilizing chips UA7805 and REG1117-3.3 to generate +5V voltage and +3.3V voltage respectively, and two patch LED lamps are connected in series with a resistor and then connected between the output of the voltage stabilizing chip and the ground to indicate whether the voltage conversion is normal or not.
Furthermore, the gas chamber is made of resin, future 8000 of resin is used in the gas chamber to fill most of useless space, so that the gas volume in the gas chamber is further reduced, and the time for cleaning the sensor is shortened; the intake pipe and the outlet duct size of air chamber are external diameter 6mm, internal diameter 4mm, and the silicone tube size as gas circuit connection is external diameter 8mm, internal diameter 4mm to guarantee good gas tightness.
Further, one combination of the sensor arrays is: MP-9, MP-4, MP503, TGS821, TGS816, TGS2602 and SHT20 digital temperature and humidity sensors.
Further, the harmful gases detected by the sensor array are carbon monoxide, methane and formaldehyde in the kitchen environment, and the detected interference gases are hydrogen and ethanol.
A method for detecting multi-component harmful gases in a kitchen environment by using the detection device is characterized by comprising the following steps:
(1) sensor preheating
The sensor of the sensing device needs to be warmed up for at least half an hour without intake air to allow the sensor to be sufficiently warmed up.
(2) Sensor cleaning
And (3) switching on an air pump in the sample injection module by the singlechip minimum system module through a relay, introducing standard air into an air chamber at a constant flow rate of 1000 ml/min, cleaning the sensor array, removing impurity interference, continuing for at least 30 minutes until the output voltage of the sensor array is stable and the voltage of each sensor is reduced to be lower than a preset voltage baseline value, entering a state capable of sample injection and detection, and executing the step (3).
(3) Sample introduction and detection, specifically comprising the following steps:
(3.1) sampling and collecting a sensor characteristic value, specifically: introducing gas to be detected which is dried, filtered and controlled in a sample introduction module into a gas chamber for detection, wherein the gas to be detected lasts for 5 minutes, during which a singlechip minimum system module samples voltage signals output by a sensor array, the voltage signals are sampled at equal time intervals, the voltage signals are converted into digital signals through a signal processing module and an AD conversion module, and then characteristic value data of the sensor are obtained and stored, the characteristic value data of the sensor comprise a baseline value before sample introduction, a maximum voltage positive slope, a voltage response peak value, a voltage peak area, a recovery voltage baseline time, a temperature characteristic value and a humidity characteristic value, and the singlechip minimum system module is loaded with a BP artificial neural network;
the pre-sampling baseline value is a sensor output voltage value subjected to differential amplification processing before a single detection process and is marked as Bi, i is 1,2,3,4,5 and 6; the voltage response peak value is the maximum voltage value output by the sensor after differential amplification processing in a single detection process and is marked as Pi, i is 1,2,3,4,5 and 6; the voltage peak area is the accumulated sum of voltage difference values of voltage response rise and output voltage of the sensor relative to a baseline value before sample injection in the single detection sample injection process; the maximum positive slope of the voltage is the maximum slope of a response curve when the voltage response rises in the sample introduction process of single detection; the recovery voltage baseline time is counted from sampling points in the cleaning process of single detection, when the voltage is reduced to the voltage of the baseline before detection, the cleaning is considered to be complete, and the number of the sampling points at the moment is taken as the recovery voltage baseline time; the temperature characteristic value and the humidity characteristic value are as follows: and carrying out three-point averaging on the temperature and humidity sensor numerical value extracted at the peak value of the sensor and the temperature and humidity sensor numerical values of two adjacent sampling points before and after the peak value to obtain a final temperature characteristic value and a final humidity characteristic value.
Processing the sampled pre-sampling baseline value and voltage response peak value to obtain the ratio of the i-th sensor resistance at the response peak value moment to the sensor resistance at the pre-sampling baseline, namely the resistance ratio, which is recorded as RiThe formula is as follows:
(3.2) training the BP artificial neural network by taking the characteristic values collected and processed by the sensor as a training set, specifically comprising the following steps: 26 neurons of an input layer, 9 neurons of a middle hidden layer, 3 neurons of an output layer and 3 neurons of the output layer of the BP artificial neural network respectively represent gas concentration values of carbon monoxide, methane and formaldehyde, initial weights and threshold values of each neuron of the input layer and the middle layer are randomly generated, a loss function adopts a mean square error, and the sum of the mean square errors of calculated values and actual values of the concentrations of three detection gases is adopted. The method for determining the number of the neurons in the middle hidden layer comprises the following steps: determined from empirical formulas, the formulas are as follows:
wherein n is the number of input neurons, m is the number of output neurons, l is the number of intermediate layer neurons, and a is a constant in the range of [1,10 ]. Selecting 9 conditions with the number of the intermediate neurons being 6-14, calculating the sum of the mean square errors of the concentration calculation values of the three detection gases and the actual value for each condition for a plurality of times, and averaging the sum of the mean square errors, wherein the number of the calculated mean square errors for each condition is the same, so as to obtain 9 average values, wherein the number of the intermediate neurons corresponding to the minimum value is 9, and the number is the optimal number of the intermediate hidden layer neurons.
The input layer-to-mid layer activation function is tansig, so the output mid _ output of each mid layer neuron can be calculated by the following equation. Wherein n represents the result obtained by accumulating the multiple inputs of a single neuron after calculating the weight and the threshold of each input.
mid_output＝tansig(n)＝2/(1+exp(-2*n))-1
While the mid-layer-to-output-layer activation function is purelin, the output of each output layer neuron can be calculated by the following equation.
output＝purelin(n)＝n
Inputting the ratio of the sensor resistance at the 6 voltage response peak moments to the sensor resistance at the baseline before sample injection, the 6 voltage maximum positive slopes, the 6 voltage peak areas, the 6 voltage recovery baseline time, the 1 temperature characteristic value and the 1 humidity characteristic value into a BP artificial neural network loaded by a singlechip minimum system module as a training set, training the BP artificial neural network by a Bayesian regularization algorithm, wherein the loss function is less than 10-3And stopping training, and finally determining the weight and the threshold of each neuron of the input layer and the middle layer to obtain the trained BP artificial neural network.
(3.3) inputting 26 characteristic values of the gas to be detected, which is acquired by the sensor, into the BP artificial neural network, calculating by the BP artificial neural network, outputting the concentrations of the 3 gases to be detected, and sending the calculation result to an upper computer through a serial port module.
(4) Cleaning state
After the detection is finished, the sample injection air pump introduces standard air into the detection air chamber at a constant flow rate of 1000 ml/min, the sensor array is cleaned for more than 10 minutes until each sensor recovers below a voltage baseline value, and then the steps (3.3) and (4) are repeated to continue sample injection and detection until the detection is stopped artificially.
Further, the single chip microcomputer minimum system module can also be loaded with a BP artificial neural network optimized by a classical principal component analysis algorithm, the classical principal component analysis algorithm carries out pre-optimization processing on 26 input characteristic values of the BP artificial neural network, an original input covariance matrix is calculated, the characteristic root of the original input covariance matrix is the principal component variance, and the characteristic root is enlarged from big to bigAnd (3) small arrangement, wherein feature vectors corresponding to the first 4 feature roots are taken as coefficients of principal components, the cumulative variance contribution rate of the coefficients reaches more than 85%, and most information of the original input feature values can be explained. From the coefficients and the input 26 eigenvalues, 4 principal components F can be obtainedi＝ai0+ai1x1+ai2x2+…+ai26x26I ═ 1,2,3,4 as new characteristic values, where aijFor the calculated coefficients, i is 1,2,3,4, j is 1,2 … 26, x1～x26Is the 26 input eigenvalues. At the moment, the BP artificial neural network comprises 4 neurons of an input layer, 9 neurons of a middle hidden layer and 3 neurons of an output layer, 4 main components are used as a training set, the BP artificial neural network is trained through a Bayesian regularization algorithm, and the weight and the threshold of each neuron of the input layer and the middle layer are determined, so that the trained BP artificial neural network is obtained and used for sample injection and detection.
Further, the singlechip minimum system module can also be loaded with a BP artificial neural network optimized by a classical partial least square algorithm, 26 characteristic values in a training set are input as independent variables, the actual concentration outputs of 3 gases to be detected are output as dependent variables, the classical partial least square algorithm extracts components t1 and u1 from the independent variables and the dependent variables respectively, and t1 is the independent variable xiLinear combination of, i.e. t1＝a0+a1x1+a2x2+…+a26x26U1 is a linear combination of dependent variables, i.e. u1＝β0+β1x1+β2x2+…+β26x26The covariance of both t1 and u1 is maximized. Let t1 and u1 be the first component. Continuing to extract the second component with the residual error of the first component, repeating the step, and taking the first four components, which have cross validity
Coefficient a between ti, available for each component extraction, and the respective input argumentijI is 1,2,3,4, j is 1,2 … 26. Four groups of coefficients obtained according to classical partial least square method for BP artificial neural networkThe 26 input characteristic values are optimized in advance to calculate four components Fi＝ai0+ai1x1+ai2x2+…+ai26x26I ═ 1,2,3,4 as new characteristic values, where x1～x26Is the 26 input eigenvalues. At the moment, the BP artificial neural network comprises 4 neurons of an input layer, 9 neurons of a middle hidden layer and 3 neurons of an output layer, the 4 components are used as a training set, the BP artificial neural network is trained through a Bayesian regularization algorithm, and the weight and the threshold of each neuron of the input layer and the middle layer are determined, so that the trained BP artificial neural network is obtained and used for sample introduction and detection.
The invention has the beneficial effects that:
the multi-component gas synchronous detection device has a multi-component gas synchronous detection function, and can still accurately and rapidly detect the multi-component gas of the target gas particularly under the condition that specific interference gas exists in a kitchen environment, so that the practicability of the device is enhanced.
Secondly, the unique structural design of the detection air chamber. The use of future 8000 resin inside the chamber fills most of the dead space, further reducing the maximum gas volume and the time to clean the sensor.
Thirdly, designing a dynamic sample injection detection process. Static diffusion often cannot meet the requirement of rapid detection, and the measured data cannot be ensured to be accurate enough, so the method adopts a fixed flow dynamic detection mode, optimizes the cleaning and detection time, and well overcomes the problem.
Fourthly, the device is miniaturized and intelligent. The instrument has the advantages of simple structure of each module, optimized connection and layout and small volume, can be embedded into a complex algorithm model at a lower computer for data operation, realizes detection of harmful gas in a kitchen, and uploads the result to the upper computer or other display equipment. The whole instrument is simple and convenient to operate, complete in function and high in intelligent level.
And fifthly, two combined algorithms of the principal component analysis algorithm and the BP artificial neural network and the partial least square algorithm and the BP artificial neural network are provided, and the principal component analysis algorithm and the partial least square algorithm are optimized aiming at the input characteristic value part of the BP artificial neural network.
Drawings
FIG. 1 is a schematic diagram of a circuit configuration;
FIG. 2 is a structural view of the gas path heating device;
FIG. 3 is a left and right perspective bottom view of the air chamber;
FIG. 4 is a graph of the temperature of the gas chamber as a function of time;
FIG. 5 is a graph of the humidity of the chamber as a function of time.
FIG. 6 is a diagram showing the raw characteristic values of the output signals of the gas sensor
In the figure, 1, a sample injection module; 2. a power supply module; 3. an air chamber and a sensor module; 4. a signal processing module; 5, an AD conversion module; 6. a single chip microcomputer minimum system module; 7. a serial port module; 8. an air inlet pipe; 9. an air inlet; a MOS sensor circular groove; 11. a temperature and humidity sensor site; 12. a bottom empty slot; 13. an air outlet groove; 14. an air outlet; 15. and an air outlet pipe.
Detailed Description
The invention is described in further detail below with reference to the figures and specific examples.
As shown in fig. 1, the schematic circuit structure of the device for detecting multi-component harmful gas in a kitchen environment of the present invention includes a sample introduction module 1, a power supply module 2, a gas chamber and sensor module 3, a signal processing module 4, an AD conversion module 5, a drying tube 5, a minimum system module 6 of a single chip microcomputer, and a serial port module 7. The sample injection module 1 comprises an air pump, a filter screen, a drying filter tube and a heating chamber. The air pump is driven by the relay, the air which is pumped at a constant flow rate and filtered by the filter screen is sequentially introduced into the drying filter tube and the heating chamber and finally reaches the air chamber, and the air pump can be replaced by a gas distributor or other sample injection instruments. The heating chamber is formed by 3D printing of high-performance heat-resistant nylon, and 4 copper pipes with the outer diameter of 6mm and the inner diameter of 4mm can be filled in the heating chamber. The power supply module is powered by a +12V switching power supply, then +5V voltage and +3.3V voltage are respectively generated by adopting voltage stabilizing chips UA7805 and REG1117-3.3 to provide different required voltages for the detection device, and the two patch LED lamps are connected in series with a resistor and then connected between the output of the voltage stabilizing chip and the ground to indicate whether voltage conversion is normal or not. The air chamber and sensor module 3 comprises an air chamber and a sensor array, the air chamber is formed by adopting future 8000 resin 3D printing, and the air chamber has the advantages of high smoothness, stable performance and the like. The intake pipe and the outlet duct size of air chamber are external diameter 6mm, internal diameter 4mm, and the silicone tube size as gas circuit connection is external diameter 8mm, internal diameter 4mm to guarantee good gas tightness. A plurality of sensors can be arranged at the bottom of the gas chamber to form a sensor array, and the sensor array adopts 6 screened commercial metal oxide semiconductor sensors and can make cross response to complex gas components, including harmful gases and harmless interference gases. In addition, a temperature and humidity sensor SHT20 is added to measure the temperature and humidity in the air chamber at regular time. The signal processing module 4 adopts a differential amplifying circuit and RC filtering, and can preprocess the output voltage of the sensor and flexibly adjust the change range of the output voltage. The AD conversion module adopts an off-chip ADC128S052 chip, has the advantages of low power consumption, multiple channels and the like, and can quickly complete AD conversion of 6-channel sensor channel voltage signals. The single-chip microcomputer minimum system adopts an MSP430F169 chip as a microcontroller, has the advantages of low power consumption, high operation speed, rich in-chip resources and the like, and an algorithm model is loaded on the single-chip microcomputer minimum system module 6, so that a measured signal can be directly analyzed and operated on a lower computer to obtain a final result, and concentration detection data can be sent to the upper computer through a UART serial port. The serial port module 7 is connected with an upper computer through a serial port-USB signal line by adopting a USART serial port, and can send detection progress conditions and final detection data to the upper computer in the whole detection process.
Referring to fig. 2, a schematic diagram of a heating chamber of a multi-component harmful gas detection device in a kitchen environment according to the present invention is shown. This heating chamber is printed by the heat-resisting nylon 3D of high performance and forms, and 4 hollow copper pipes of external diameter 6mm internal diameter 4mm can be placed to middle recess for transmit the heat of heating plate for inside air current fast and fully. The heating plate is fixed in the square groove above, and the bottom is contacted with the copper pipe and the heating device through heat conduction silica gel, so that the contact area and the heat conductivity coefficient are increased, and good heat conduction is realized.
As shown in fig. 3, a gas chamber structure diagram of a multi-component harmful gas detection apparatus in a kitchen environment according to the present invention. The bottom of the air chamber is connected with a circuit board sensor area through 705 transparent RTV silicon rubber, so that good adhesion and air tightness of the air chamber are guaranteed.
When the detection device is started for detection, the sensor of the detection device needs to be preheated for at least half an hour under the condition of no air intake, so that the sensor is fully preheated. And then introducing the gas into a gas chamber at a constant flow rate of 1000 ml/min, cleaning the sensor array, removing impurity interference, continuing for at least 30 minutes until the output voltage of the sensor array is stable, and the voltage of each sensor is reduced to be below a preset voltage baseline value, entering a state capable of sample introduction and detection, and detecting the concentration of the gas to be detected, wherein the detection device is not required to be stopped after being started and can continuously detect the gas to be detected. When the continuous detection is carried out, the time for cleaning the sensor array can be shortened, and the voltage of each sensor can be ensured to be reduced to be lower than the preset voltage baseline value in about 10 minutes.
During the detection, the mixed gas after drying and particle filtering enters the air chamber through the air inlet pipe 8, and the air is continuously introduced for 5 minutes at the flow rate of 1000 ml/minute. If the sample injection time is too short, the sensor response still changes violently, and if the sample injection time is too long, energy waste and time consumption are caused; the air inlet 9 is equal to the head of the sensor in height, so that air flow can rapidly flow through each sensor; the MOS sensor circular groove 10 can stably place a sensor; the air flow can flow through the surface of the sensor, passes through the bottom hollow groove 12, passes through the temperature and humidity sensor sites 11 in the air outlet groove 13, and is discharged from the air outlet holes 14 and the air outlet pipe 15, and the temperature and humidity sensor sites 11 are positioned behind the sensor array, namely the tail end of an air path in the air chamber, so that the measured data can accurately reflect the overall temperature and humidity change conditions of the air flow near the sensor array, as shown in fig. 4 and 5. During detection, a single-chip microcomputer minimum system module 6 samples voltage signals output by a sensor array at regular time, the voltage signals are converted into digital signals through a signal processing module 4 and an AD conversion module 5, and then characteristic value data of the sensor and temperature and humidity data of a gas chamber are obtained and stored, the original characteristic value data of the sensor comprise a sample introduction front baseline value, a voltage maximum positive slope, a voltage response peak value, a voltage peak area, voltage recovery baseline time, temperature and humidity and the like, the sample introduction front baseline value is a sensor output voltage value subjected to differential amplification processing before a single detection process and is recorded as Bi, i is 1,2,3,4,5, 6; the voltage response peak value is the maximum voltage value output by the sensor after differential amplification processing in a single detection process and is marked as Pi, i is 1,2,3,4,5 and 6; the voltage peak area is the accumulated sum of voltage difference values of voltage response rise and output voltage of the sensor relative to a baseline value before sample injection in the single detection sample injection process; the maximum positive slope of the voltage is the maximum slope of a response curve when the voltage response rises in the sample introduction process of single detection; the recovery voltage baseline time is counted from sampling points in the cleaning process of single detection, when the voltage is reduced to the voltage of the baseline before detection, the cleaning is considered to be complete, and the number of the sampling points at the moment is taken as the recovery voltage baseline time; the temperature characteristic value and the humidity characteristic value are as follows: and carrying out three-point averaging on the temperature and humidity sensor numerical value extracted at the peak value of the sensor and the temperature and humidity sensor numerical values of two adjacent sampling points before and after the peak value to obtain a final temperature characteristic value and a final humidity characteristic value.
The sampled pre-sampling baseline value and the voltage response peak value are combined to obtain the ratio of the resistance of the ith sensor at the response peak value moment to the resistance of the sensor at the pre-sampling baseline, namely the resistance ratio, which is recorded as RiThe formula is as follows:
after the characteristic value is processed, the resistance ratio, the maximum positive slope of the voltage, the voltage peak area, the recovery voltage baseline time, the temperature and the humidity and the like are left, the single chip microcomputer minimum system module 6 is loaded with the BP artificial neural network, the processed sensor characteristic value is used as a training sample, and the BP artificial neural network is trained, specifically: 26 neurons of an input layer, 9 neurons of a middle hidden layer, 3 neurons of an output layer and 3 neurons of the output layer of the BP artificial neural network respectively represent gas concentration values of carbon monoxide, methane and formaldehyde, initial weights and threshold values of each neuron of the input layer and the middle layer are randomly generated, a loss function adopts a mean square error, and the sum of the mean square errors of calculated values and actual values of the concentrations of three detection gases is adopted. The method for determining the number of the neurons in the middle hidden layer comprises the following steps: determined from empirical formulas, the formulas are as follows:
wherein n is the number of input neurons, m is the number of output neurons, l is the number of intermediate layer neurons, and a is a constant in the range of [1,10 ]. Selecting 9 conditions with the number of the intermediate neurons being 6-14, calculating the sum of the mean square errors of the concentration calculation values of the three detection gases and the actual value for each condition for a plurality of times, and averaging the sum of the mean square errors, wherein the number of the calculated mean square errors for each condition is the same, so as to obtain 9 average values, wherein the number of the intermediate neurons corresponding to the minimum value is 9, and the number is the optimal number of the intermediate hidden layer neurons.
The input layer-to-mid layer activation function is tansig, so the output mid _ output of each mid layer neuron can be calculated by the following equation. Wherein n represents the result obtained by accumulating the multiple inputs of a single neuron after calculating the weight and the threshold of each input.
mid_output＝tansig(n)＝2/(1+exp(-2*n))-1
While the mid-layer-to-output-layer activation function is purelin, the output of each output layer neuron can be calculated by the following equation.
output＝purelin(n)＝n
The specific value of the sensor resistance at the 6 voltage response peak moments and the sensor resistance at the baseline before sample introduction, the maximum positive slope of 6 voltages, the peak area of 6 voltages, the baseline time of 6 recovery voltages, 1 temperature characteristic value and 1 humidityThe characteristic values are input into a BP artificial neural network loaded by a minimum system module 6 of the single chip microcomputer as a training set, the BP artificial neural network is trained through a Bayesian regularization algorithm, and a loss function is less than 10-3And stopping training, and finally determining the weight and the threshold of each neuron of the input layer and the middle layer to obtain the trained BP artificial neural network.
26 characteristic values of the gas to be detected, which is collected and processed by the sensor, are input into the BP artificial neural network, the concentration of the gas to be detected is output through calculation of the BP artificial neural network, and the calculation result is sent to an upper computer through the serial port module 7.
The single-chip microcomputer minimum system module 6 can also be loaded with a BP artificial neural network optimized by a classic principal component analysis algorithm, the classic principal component analysis algorithm carries out optimization processing in advance aiming at 26 input characteristic values of the BP artificial neural network, an original input covariance matrix is calculated, a characteristic root is a principal component variance, the characteristic roots are arranged from large to small, characteristic vectors corresponding to the first 4 characteristic roots are taken as coefficients of principal components, the cumulative variance contribution rate reaches more than 85%, and most information of the original input characteristic values can be explained. From the coefficients and the input 26 eigenvalues, 4 principal components F can be obtainedi＝ai0+ai1x1+ai2x2+…+ai26x26I ═ 1,2,3,4 as new characteristic values, where aijFor the calculated coefficients, i is 1,2,3,4, j is 1,2 … 26, x1～x26Is the 26 input eigenvalues. At the moment, the BP artificial neural network comprises 4 neurons of an input layer, 9 neurons of a middle hidden layer and 3 neurons of an output layer, 4 main components are used as a training set, the BP artificial neural network is trained through a Bayesian regularization algorithm, and the weight and the threshold of each neuron of the input layer and the middle layer are determined, so that the trained BP artificial neural network is obtained and used for sample injection and detection.
The minimum system module 6 of the single chip microcomputer can also be loaded with a BP artificial neural network optimized by a classical partial least square algorithm, 26 characteristic values in a training set are input as independent variables, the actual concentration output of 3 gases to be detected is output as dependent variables, and the independent variables are processed byRespectively extracting components t1 and u1 from independent variables and dependent variables by a classical partial least squares algorithm, wherein t1 is independent variable xiLinear combination of, i.e. t1＝a0+a1x1+a2x2+…+a26x26U1 is a linear combination of dependent variables, i.e. u1＝β0+β1x1+β2x2+…+β26x26The covariance of both t1 and u1 is maximized. Let t1 and u1 be the first component. Continuing to extract the second component with the residual error of the first component, repeating the step, and taking the first four components, which have cross validity
Coefficient a between ti, available for each component extraction, and the respective input argumentijI is 1,2,3,4, j is 1,2 … 26. . Four groups of coefficients obtained according to a classical partial least square method are subjected to pre-optimization processing aiming at 26 input characteristic values of the BP artificial neural network, and four components F are calculatedi＝ai0+ai1x1+ai2x2+…+ai26x26I ═ 1,2,3,4 as new characteristic values, where x1～x26Is the 26 input eigenvalues. At the moment, the BP artificial neural network comprises 4 neurons of an input layer, 9 neurons of a middle hidden layer and 3 neurons of an output layer, the 4 components are used as a training set, the BP artificial neural network is trained through a Bayesian regularization algorithm, and the weight and the threshold of each neuron of the input layer and the middle layer are determined, so that the trained BP artificial neural network is obtained and used for sample introduction and detection.
For various lower computer algorithm test examples of the mixed gas sample of 240ppm of carbon monoxide, 8400ppm of methane and 0ppm of formaldehyde, the threshold voltages of 6 sensors in the detection device are respectively as follows: 0.6V, 1.0V, 0.7V, 0.3V, 0.5V, 0.8V.
When the algorithm loaded in the minimum system module 6 of the single chip microcomputer is partial least square regression or principal component regression, the errors of the detected carbon monoxide and methane are very large, only formaldehyde is 0, and the errors are relatively accurate, when the algorithm loaded in the minimum system module 6 of the single chip microcomputer is a BP artificial neural network, the errors of the detected three gas values are reduced to some extent and are relatively close to actual values, and when the algorithm loaded in the minimum system module 6 of the single chip microcomputer is principal component analysis + BP artificial neural network or partial least square + BP artificial neural network, the detected result is relatively accurate, and the errors of the detected result and the actual values are very small.
The gas flow channel for cleaning and sample injection is completely consistent in the gas chamber, and the cleaning time is longer than the sample injection time, and is more than 10 minutes until each sensor recovers below the voltage baseline value. Too long cleaning time can cause waste of time and resources, and too short cleaning time can cause that part of gas to be measured still has residue and the baseline cannot be restored to the initial level in time to form obvious baseline drift. After the cleaning is finished, the detection can be continued until the detection is stopped artificially.
The invention relates to a sensor combination of a multi-component harmful gas detection device in kitchen environment, which is an MP-9, MP-4, MP503, TGS821, TGS816, TGS2602 and SHT20 digital temperature and humidity sensor, wherein the detectable target gas concentration ranges are as follows:
carbon monoxide detection range: 10 to 1000 ppm;
detection range of methane: 500-10000 ppm;
the formaldehyde detection range is as follows: 1 to 50 ppm.
The allowable concentration ranges of the interfering gas under normal detection are respectively as follows: 10-100ppm of hydrogen and 10-50ppm of ethanol.
The above embodiments are only for illustrating the technical idea of the present invention, and the protection scope of the present invention is not limited thereby, and any modifications made on the basis of the technical scheme according to the technical idea of the present invention fall within the protection scope of the present invention.
Claims (9)
1. A multi-component harmful gas detection method in a kitchen environment is characterized in that the method is realized based on a multi-component harmful gas detection device in the kitchen environment, and the device comprises a sample injection module (1), a power supply module (2), a gas chamber and sensor module (3), a signal processing module (4), an AD conversion module (5), a singlechip minimum system module (6) and a serial port module (7);
the sample injection module (1) comprises an air pump, a drying and filtering pipe and a heating chamber and is used for introducing dried, filtered and temperature-controlled gas into the air chamber and the sensor module (3) at a constant speed; the air pump is used for enabling the flow rate of gas in the sample introduction process and the sensor cleaning process to be consistent and unchanged, and ensuring the relative stability of voltage signals output by the sensor before and after the sensor cleaning process; the interior of the drying filter tube is filled with allochroic silica gel and is used for filtering particulate impurities and water vapor in the sample gas and indicating the performance condition of the silica gel through color change; the air pump, the drying filter pipe and the heating chamber are connected in sequence;
the power supply module (2) is used for providing different required voltages for the sample introduction module (1) and the singlechip minimum system module (6) and is powered by a +12V switching power supply;
the air chamber and sensor module (3) comprises an air chamber and a sensor array, the main body of the air chamber is of a cuboid structure, a gas detection chamber is arranged above the inner part of the air chamber, the side surface of the gas detection chamber is provided with an air inlet hole (9), the air inlet (9) is connected with an air inlet pipe (8), the bottom in the air chamber is provided with 6 MOS sensor circular grooves (10) for fixing and replacing the sensors, a bottom empty groove (12) with the height of 3mm is arranged at the bottom of the air chamber, an air outlet groove (13) is arranged at the bottom of the air chamber and at the side opposite to the air inlet hole (9), the bottom empty groove (12) is communicated with an air outlet groove (13), the bottom of the air outlet groove (13) is provided with a temperature and humidity sensor site (11), the side surface of the air outlet groove (13) is provided with an air outlet hole (14), the air outlet hole (14) is connected with an air outlet pipe (15), gas to be detected is converged from a bottom empty groove (12) at the bottom of the gas chamber and then is discharged through a gas outlet pipe (15); the sensor array comprises 6 MOS sensors and 1 temperature and humidity sensor; the MOS sensor is arranged in the MOS sensor circular groove, and the head of the MOS sensor is positioned in the gas detection chamber and is consistent with the height of the gas inlet hole (9); the temperature and humidity sensor is arranged at a temperature and humidity sensor position point (11) at the bottom of the gas outlet groove and used for detecting the actual temperature and humidity of the heated gas flowing through the sensor array;
the signal processing module (4) adopts a differential amplification circuit and RC filtering and is used for carrying out differential amplification and filtering processing on output signals of the sensor array and adjusting the variation range of the output signals;
the AD conversion module (5) is used for rapidly completing AD conversion of the voltage signal of the sensor after differential amplification and filtering processing, and inputting the voltage signal into the singlechip minimum system module (6);
the single chip microcomputer minimum system module (6) is provided with an algorithm model for detecting the concentration of the gas to be detected, and is used for directly carrying out analysis and operation on the received signal on a lower computer to obtain a final result and sending the final result to the upper computer through a serial port module (7);
the serial port module (7) adopts a USART serial port and sends a detection result to an upper computer through a serial port-USB signal line;
the detection of the multi-component harmful gas in the kitchen environment comprises the following steps:
(1) sensor preheating
The sensor of the detection device needs to be preheated for at least half an hour under the condition of no air intake, so that the sensor is fully preheated;
(2) sensor cleaning
The single chip microcomputer minimum system module (6) is connected with an air pump in the sample injection module (1) through a relay, standard air is led into an air chamber at a constant flow rate of 1000 ml/min, the sensor array is cleaned, impurity interference is removed, the operation lasts for at least 30 minutes until the output voltage of the sensor array is stable, the voltage of each sensor is reduced to be below a preset voltage baseline value, the state of sample injection and detection is achieved, and the step (3) is executed;
(3) sample introduction and detection, specifically comprising the following steps:
(3.1) sampling and collecting a sensor characteristic value, specifically: introducing gas to be detected which is dried, filtered and temperature-controlled in a sample introduction module (1) into an air chamber for detection, wherein the gas to be detected lasts for 5 minutes, during the period, a singlechip minimum system module (6) samples a voltage signal output by a sensor array, the voltage signal is sampled at equal time intervals, the voltage signal is converted into a digital signal through a signal processing module (4) and an AD conversion module (5), and then characteristic value data of the sensor is obtained and stored, the characteristic value data of the sensor comprises a sample introduction front baseline value, a voltage maximum positive slope, a voltage response peak value, a voltage peak area, a voltage recovery baseline time, a temperature characteristic value and a humidity characteristic value, and the singlechip minimum system module (6) is loaded with a BP artificial neural network;
the pre-sampling baseline value is a sensor output voltage value subjected to differential amplification processing before a single detection process and is marked as Bi, i is 1,2,3,4,5 and 6; the voltage response peak value is the maximum voltage value output by the sensor after differential amplification processing in a single detection process and is marked as Pi, i is 1,2,3,4,5 and 6; the voltage peak area is the accumulated sum of voltage difference values of voltage response rise and output voltage of the sensor relative to a baseline value before sample injection in the single detection sample injection process; the maximum positive slope of the voltage is the maximum slope of a response curve when the voltage response rises in the sample introduction process of single detection; the recovery voltage baseline time is counted from sampling points in the cleaning process of single detection, when the voltage is reduced to the voltage of the baseline before detection, the cleaning is considered to be complete, and the number of the sampling points at the moment is taken as the recovery voltage baseline time; the temperature characteristic value and the humidity characteristic value are as follows: carrying out three-point averaging on the temperature and humidity sensor numerical value extracted at the peak value of the sensor and the temperature and humidity sensor numerical values of two adjacent sampling points before and after the peak value to obtain a final temperature characteristic value and a final humidity characteristic value;
processing the sampled pre-sampling baseline value and voltage response peak value to obtain the ratio of the i-th sensor resistance at the response peak value moment to the sensor resistance at the pre-sampling baseline, namely the resistance ratio, which is recorded as RiThe formula is as follows:
(3.2) training the BP artificial neural network by taking the characteristic values collected and processed by the sensor as a training set, specifically comprising the following steps: 26 neurons of an input layer, 9 neurons of a middle hidden layer, 3 neurons of an output layer and 3 neurons of the output layer of the BP artificial neural network respectively represent gas concentration values of carbon monoxide, methane and formaldehyde, initial weights and threshold values of each neuron of the input layer and the middle layer are randomly generated, a loss function adopts a mean square error, and the sum of the mean square errors of calculated values and actual values of the concentrations of three gases to be detected is the sum of the mean square errors of the calculated values and the actual values;
the input layer-to-mid layer activation function is tansig, so the output mid _ output of each mid layer neuron is calculated by the following equation; wherein n represents the result obtained by accumulating a plurality of inputs of a single neuron and the weight and threshold of each input;
mid_output＝tansig(n)＝2/(1+exp(-2*n))-1
while the mid-layer-to-output-layer activation function is purelin, so the output of each output layer neuron is calculated by the following equation;
output＝purelin(n)＝n
inputting the ratio of the sensor resistance at the 6 voltage response peak moments to the sensor resistance at the baseline before sample injection, the 6 voltage maximum positive slopes, the 6 voltage peak areas, the 6 voltage baseline recovery times, the 1 temperature characteristic value and the 1 humidity characteristic value as training sets into a BP artificial neural network loaded by a singlechip minimum system module (6), training the BP artificial neural network through a Bayesian regularization algorithm, and finally determining the weight and the threshold of each neuron of an input layer and a middle layer to obtain the trained BP artificial neural network;
(3.3) inputting 26 characteristic values of the gas to be detected, which is acquired by the sensor, into a BP artificial neural network, calculating by the BP artificial neural network, outputting the concentrations of 3 gases to be detected, and sending the calculation result to an upper computer through a serial port module (7);
(4) cleaning state
And (3) after the detection is finished, introducing standard air into the detection air chamber at a constant flow rate of 1000 ml/min by using the sample injection air pump, cleaning the sensor array for more than 10 minutes until each sensor recovers below a voltage baseline value, and repeating the steps (3.3) and (4) to continue sample injection and detection until the detection is stopped artificially.
2. The method for detecting the multi-component harmful gases in the kitchen environment according to claim 1, wherein the air pump is a laboratory micro vacuum pump, flow rate adjustment is realized through PWM (pulse-width modulation) waves, and the air pump is operated all the time in the whole detection process.
3. The method for detecting the multi-component harmful gases in the kitchen environment according to claim 1, wherein the heating chamber is composed of a copper pipe, a heating sheet, heat-conducting glue and a fixing frame; the heating chamber is formed by 3D printing of high-performance heat-resistant nylon, and 4 copper pipes with the outer diameter of 6mm and the inner diameter of 4mm are filled in the heating chamber.
4. The method for detecting the multi-component harmful gases in the kitchen environment according to claim 1, wherein the power supply module (2) adopts voltage stabilizing chips UA7805 and REG1117-3.3 to generate +5V voltage and +3.3V voltage respectively, and two chip LED lamps are connected in series with a resistor and then connected between the output of the voltage stabilizing chip and the ground to indicate whether the voltage conversion is normal or not.
5. The method for detecting the multi-component harmful gases in the kitchen environment according to claim 1, wherein the gas chamber is made of resin, future 8000 of the resin is used in the gas chamber to fill most of useless space, so that the gas volume in the gas chamber is further reduced, and the time for cleaning the sensor is shortened; intake pipe (8) and outlet duct chi (15) cun of air chamber are external diameter 6mm, internal diameter 4mm, and the silicone tube size of connecting as the gas circuit is external diameter 8mm, internal diameter 4mm to guarantee good gas tightness.
6. The method of claim 1, wherein a combination of the sensor arrays is: gas sensors MP-9, MP-4, MP503, TGS821, TGS816, TGS2602 and a digital temperature and humidity sensor SHT 20.
7. The method for detecting the multi-component harmful gases in the kitchen environment according to claim 1, wherein the harmful gases detected by the sensor array are carbon monoxide, methane and formaldehyde in the kitchen environment, and the interference gases detected are hydrogen and ethanol.
8. The method according to claim 1, wherein the monolithic minimum system module (6) is loaded into a BP artificial neural network optimized by a classical principal component analysis algorithm, the classical principal component analysis algorithm performs pre-optimization processing on 26 input eigenvalues of the BP artificial neural network, an original input covariance matrix is calculated, the eigenvalues are principal component variances, the eigenvalues are arranged from large to small, eigenvectors corresponding to the first 4 eigenvalues are taken as coefficients of principal components, and 4 principal components F are obtained according to the coefficients and the input 26 eigenvaluesi＝ai0+ai1x1+ai2x2+…+ai26x26I ═ 1,2,3,4 as new characteristic values, where aijFor the calculated coefficients, i is 1,2,3,4, j is 1,2 … 26, x1～x2626 input characteristic values; at the moment, the BP artificial neural network comprises 4 neurons of an input layer, 9 neurons of a middle hidden layer and 3 neurons of an output layer, 4 main components are used as a training set, the BP artificial neural network is trained through a Bayesian regularization algorithm, and the weight and the threshold of each neuron of the input layer and the middle layer are determined, so that the trained BP artificial neural network is obtained and used for sample injection and detection.
9. The method according to claim 1, characterized in that the monolithic minimum system module (6) loads a BP artificial neural network optimized by a classical partial least squares algorithm, 26 eigenvalues in a training set are input as independent variables, 3 actual concentrations of the gas to be detected are output as dependent variables, the classical partial least squares algorithm extracts components t1 and u1 from the independent variables and the dependent variables respectively, t1 is a linear combination of the independent variables, u1 is a linear combination of the dependent variables, and the covariance of both t1 and u1 is maximized; let t1 and u1 be the first component;continuing to extract the second component with the residual error of the first component, repeating the steps, and taking the first four components to obtain t of the four componentsiAnd the coefficient a between each input argumentij1,2,3,4, j 1,2 … 26; four groups of coefficients obtained according to a classical partial least square method are subjected to pre-optimization processing aiming at 26 input characteristic values of the BP artificial neural network, and four components F are calculatedi＝ai0+ai1x1+ai2x2+…+ai26x26I ═ 1,2,3,4 as new characteristic values, where x1～x2626 input characteristic values; at the moment, the BP artificial neural network comprises 4 neurons of an input layer, 9 neurons of a middle hidden layer and 3 neurons of an output layer, the 4 components are used as a training set, the BP artificial neural network is trained through a Bayesian regularization algorithm, the weight and the threshold of each neuron of the input layer and the middle layer are determined, and the trained BP artificial neural network is obtained and used for sample introduction and detection.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910848479.7A CN110426421B (en) | 2019-09-09 | 2019-09-09 | Device and method for detecting multi-component harmful gas in kitchen environment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910848479.7A CN110426421B (en) | 2019-09-09 | 2019-09-09 | Device and method for detecting multi-component harmful gas in kitchen environment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110426421A CN110426421A (en) | 2019-11-08 |
| CN110426421B true CN110426421B (en) | 2020-10-16 |
Family
ID=68418842
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910848479.7A Active CN110426421B (en) | 2019-09-09 | 2019-09-09 | Device and method for detecting multi-component harmful gas in kitchen environment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110426421B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112819158B (en) * | 2021-02-05 | 2024-02-27 | 凌坤(南通)智能科技有限公司 | Gas identification method based on optimized BP neural network |
| CN113537334A (en) * | 2021-07-09 | 2021-10-22 | 长安大学 | Gas-liquid multi-component real-time intelligent detection method |
| CN114002303B (en) * | 2021-12-31 | 2022-04-05 | 中国农业科学院农业资源与农业区划研究所 | Calibration method for gas sensing in cold-chain logistics and multi-source sensing device |
| CN114965872B (en) * | 2022-04-27 | 2023-10-13 | 重庆科技学院 | Electronic nose and method for multi-sensor data fusion |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101000318A (en) * | 2006-12-28 | 2007-07-18 | 中国科学院合肥物质科学研究院 | Sensor array based on temp control and gas investigating method |
| CN202854118U (en) * | 2012-10-27 | 2013-04-03 | 西安科技大学 | Embedded networking intelligent multi-parameter gas detection system |
| CN205562572U (en) * | 2016-04-14 | 2016-09-07 | 怀化学院 | Portable sample quickly pretreatment device for monitoring environment and food |
| CN106289394A (en) * | 2016-08-04 | 2017-01-04 | 苏州云白环境设备股份有限公司 | A kind of Wearable real time environment gas controlling device and monitoring method thereof |
| CN106770738A (en) * | 2016-12-03 | 2017-05-31 | 浙江大学 | The expiratory air multi-analyte immunoassay instrument and detection method of a kind of gas concentration lwevel amendment |
-
2019
- 2019-09-09 CN CN201910848479.7A patent/CN110426421B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101000318A (en) * | 2006-12-28 | 2007-07-18 | 中国科学院合肥物质科学研究院 | Sensor array based on temp control and gas investigating method |
| CN202854118U (en) * | 2012-10-27 | 2013-04-03 | 西安科技大学 | Embedded networking intelligent multi-parameter gas detection system |
| CN205562572U (en) * | 2016-04-14 | 2016-09-07 | 怀化学院 | Portable sample quickly pretreatment device for monitoring environment and food |
| CN106289394A (en) * | 2016-08-04 | 2017-01-04 | 苏州云白环境设备股份有限公司 | A kind of Wearable real time environment gas controlling device and monitoring method thereof |
| CN106770738A (en) * | 2016-12-03 | 2017-05-31 | 浙江大学 | The expiratory air multi-analyte immunoassay instrument and detection method of a kind of gas concentration lwevel amendment |
Non-Patent Citations (4)
| Title |
|---|
| An optimized multi-classifiers ensemble learning for identification ofginsengs based on electronic nose;Xiyang Sun等;《Sensors and Actuators A: Physical》;20170909(第266期);654-661 * |
| Discrimination of Different Species of Dendrobium with an Electronic Nose Using Aggregated Conformal Predictor;You Wang等;《Sensors》;20190225(第19期);964-979 * |
| 基于智能电子鼻的冰箱冷藏食品新鲜度原位检测技术;王敏等;《传感技术学报》;20190228;第32卷(第2期);161-166 * |
| 穿戴式实时环境气体监测系统;高凡等;《中国生物医学工程学报》;20151231;第34卷(第6期);135-144 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110426421A (en) | 2019-11-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110426421B (en) | Device and method for detecting multi-component harmful gas in kitchen environment | |
| CN109444232B (en) | Multichannel intelligent polluted gas monitoring device and diffusion tracing method | |
| US9121837B2 (en) | Method and device for environmental monitoring | |
| CN102590283B (en) | Method for detecting freshness of grass carp by using electronic nose | |
| CN1194227C (en) | Fast non-destructive detection method and device of food smell based on gas sensor array technology | |
| JP2002022692A (en) | Odor measuring apparatus | |
| CN103389323B (en) | Method for evaluating ages of precious medicinal materials quickly and losslessly | |
| CN105445158A (en) | High-accuracy real-time online detecting instrument for atmospheric pollution | |
| CN108535201B (en) | House refuse component real-time detection apparatus and method in a kind of incinerator | |
| CN106290532A (en) | The intelligent water quality trace heavy metal on-line monitoring of a kind of internet of things oriented and early warning system | |
| EP2873971A1 (en) | An artificial olfactory system and an application thereof | |
| US20160054297A1 (en) | Method and apparatus for detecting breath alcohol concentration based on acoustic breath sampler | |
| CN105866190A (en) | Electronic nose device for food detection and detection method | |
| CN101059525B (en) | Self-adaptive monitoring method for gas dissolved in oil of traction transformer, and the device thereof | |
| CN105866050A (en) | Low-cost lossless and fast detecting equipment for apple moldy core | |
| CN109406719A (en) | A kind of odor detection system of carrier gas sample introduction | |
| CN104089656B (en) | A kind of stockyard spontaneous combustionof coal detection method and device | |
| CN206074625U (en) | Bionic olfactory detection and analysis device based on dynamic air-distributing | |
| CN106443031B (en) | Bionic olfactory detection and analysis device and its determination method based on dynamic air-distributing | |
| CN102778445B (en) | Intelligent analyzer and detection method for standard state dry basis | |
| CN103983667B (en) | A kind of free fatty acid rapid determination device and detection method | |
| CN106979998B (en) | Bionic smell rapid nondestructive detection device and detection method for apple freshness | |
| CN100561195C (en) | The non-disperse infrared spectrum determination method and the device that are suitable for wide environmental temperature range | |
| CN110243877A (en) | A kind of fast gas quantitative testing device suitable for multiple gases | |
| CN206710409U (en) | The bionic olfactory Rapid non-destructive testing device of apple freshness |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |