CN109060892B - SF based on graphene composite material sensor array6Method for detecting decomposition product - Google Patents

SF based on graphene composite material sensor array6Method for detecting decomposition product Download PDF

Info

Publication number
CN109060892B
CN109060892B CN201810666669.2A CN201810666669A CN109060892B CN 109060892 B CN109060892 B CN 109060892B CN 201810666669 A CN201810666669 A CN 201810666669A CN 109060892 B CN109060892 B CN 109060892B
Authority
CN
China
Prior art keywords
gas
sensor array
neuron
output
sensitive
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
Application number
CN201810666669.2A
Other languages
Chinese (zh)
Other versions
CN109060892A (en
Inventor
杨爱军
褚继峰
王小华
骆挺
荣命哲
刘定新
李育灵
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian Jiaotong University
Changzhi Power Supply Co of State Grid Shanxi Electric Power Co Ltd
Original Assignee
Xian Jiaotong University
Changzhi Power Supply Co of State Grid Shanxi Electric Power Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Xian Jiaotong University, Changzhi Power Supply Co of State Grid Shanxi Electric Power Co Ltd filed Critical Xian Jiaotong University
Priority to CN201810666669.2A priority Critical patent/CN109060892B/en
Publication of CN109060892A publication Critical patent/CN109060892A/en
Application granted granted Critical
Publication of CN109060892B publication Critical patent/CN109060892B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • G01N27/125—Composition of the body, e.g. the composition of its sensitive layer
    • G01N27/127—Composition of the body, e.g. the composition of its sensitive layer comprising nanoparticles

Abstract

The invention relates to SF based on a graphene composite material sensor array6A method for detecting a decomposition product, comprising: processing and preparing a sensor array; arranging the prepared sensor array in a closed air chamber for signal acquisition, and electrifying for initialization; performing gas-sensitive test on the sensor array and storing the test result as a sample; and constructing a gas identification network model and identifying the atmosphere. According to the invention, the graphene-metal oxide nano gas-sensitive material is added into the sensor array, so that the energy consumption can be effectively reduced; by constructing a gas identification network model in combination with a sensor array, SF can be identified6And effectively identifying the gas decomposition products.

Description

SF based on graphene composite material sensor array6Method for detecting decomposition product
Technical Field
The invention belongs to the field of gas decomposition product detection, and particularly relates to SF (sulfur hexafluoride) based on a graphene composite material sensor array6A method for detecting a decomposition product.
Background
Sulfur hexafluoride (SF)6) The high-voltage power device is widely applied to high-voltage power devices such as gas insulated switchgear, circuit breakers, gas insulated buses and the like. It was found that SF6The gas will decompose under the action of high-temperature discharge to generate thionyl fluoride (SOF)2) Sulfuryl fluoride (SO)2F2) Sulfur dioxide (SO)2) And hydrogen sulfide (H)2S) and the like, and the component content of the decomposition products is related to the discharge intensity in the equipment, so that the method can be used for equipment fault diagnosis. The traditional gas chromatography-mass spectrometry and infrared absorption spectrometry have high precision, but SF cannot be realized6On-line monitoring of the decomposition products.
Gas sensors based on nanomaterials are available in SF for their advantages of small size, low cost, good portability, large potential for on-line monitoring, and the like6The detection of gas decomposition products is gradually applied. However, such sensors have the problem of high operating temperature and are suitable for SF6The cross sensitivity problem of various gases generated after decomposition is not well solved.
The graphene has the characteristics of high specific surface area, good electron mobility and low noise, and can be compounded with metal oxide to obtain a gas-sensitive material with high sensitivity at low temperature, so that the problem of high working temperature of the traditional device can be effectively solved.
Disclosure of Invention
Against the currently existing SF6The invention provides a solution based on a graphene nanocomposite sensor array, and aims to solve the problem of detecting the content of a decomposition product component. For the problem that the traditional gas sensor has high working temperature (generally higher than 200 ℃), the composite material is constructed by introducing graphene, so that the power consumption of the sensor is reduced. Meanwhile, for the problem of insufficient output information of a single sensor, the invention can increase the output information by designing a novel sensor array mode. In addition, the problem of cross sensitivity of a single sensor is solved by utilizing the constructed sensor array and combining a pattern recognition algorithm.
SF based on graphene composite material sensor array6A method for detecting a decomposition product, comprising:
step S100: processing and preparing a sensor array;
step S200: arranging the prepared sensor array in a closed air chamber for signal acquisition, and electrifying for initialization;
step S300: performing gas-sensitive test on the sensor array and storing the test result as a sample;
step S400: constructing a gas identification network model and identifying atmosphere;
wherein, step S400 includes:
step S401: collecting characteristic parameters of the sensor array in different atmosphere environments;
step S402: establishing a gas recognition network model based on an error inverse propagation algorithm and training;
(a) normalization processing of characteristic parameters:
extracting the input characteristic parameters of each sample:
Figure GDA0002680332920000021
in the formula (I), the compound is shown in the specification,
Figure GDA0002680332920000022
the ith output result of the sensor array of the kth sample, d represents the dimension of the input characteristic parameter;
v. establishing output characteristic parameters according to sample types:
Figure GDA0002680332920000031
in the formula (I), the compound is shown in the specification,
Figure GDA0002680332920000032
representing the jth atmosphere of the kth sample, m representing the dimension of the output characteristic parameter,
Figure GDA0002680332920000033
taking 0 or 1, if 1 is yes, and if not, 0 is not, and distinguishing each sample type according to the output characteristic parameters;
obtaining a training set D:
Figure GDA0002680332920000034
in the formula, 1 represents the total number of samples;
and vi, inputting and outputting characteristic parameters for normalization treatment:
Figure GDA0002680332920000035
Figure GDA0002680332920000036
in the formula (I), the compound is shown in the specification,
Figure GDA0002680332920000037
is the characteristic parameter before the normalization,
Figure GDA0002680332920000038
in order to obtain the normalized characteristic parameters,
Figure GDA0002680332920000039
is the minimum value of the characteristic parameter in the kth sample,
Figure GDA00026803329200000310
is the maximum value of the characteristic parameter in the kth sample,
Figure GDA00026803329200000311
is the output parameter before the normalization,
Figure GDA00026803329200000312
outputting parameters after normalization;
obtaining a normalized training set D:
Figure GDA00026803329200000313
(b) establishing a gas recognition network model based on an error inverse propagation algorithm, wherein the specific training process is as follows:
defining a learning rate eta epsilon (0, 1), and randomly initializing all connection weights and threshold values in the neural network;
calculating neural network output
Figure GDA0002680332920000041
Figure GDA0002680332920000042
Figure GDA0002680332920000043
Figure GDA0002680332920000044
Figure GDA0002680332920000045
In the formula, alphahInput to the first neuron of the hidden layer, bhIs the output of the h-th neuron of the hidden layer, betajIs the input of the jth output neuron, gammahThreshold for the h-th neuron of the hidden layer, θjThreshold value for the j-th neuron of the output layer, vihThe connection weight, omega, between the ith neuron of the input layer and the h-th neuron of the hidden layerhjThe connection weight between the h-th neuron of the hidden layer and the j-th neuron of the output layer is shown, d represents the number of neurons of the input layer, and q represents the number of neurons of the hidden layer; calculating gradient term g for output layer neuronsj：
Figure GDA0002680332920000046
Computing gradient term e for hidden neuronsh：
Figure GDA0002680332920000047
Wherein m represents the number of neurons in the output layer;
combining the calculated output layer neuron gradient terms and the calculated hidden layer neuron gradient terms to train a neural network;
updating the connection right of the output layer neuron and the hidden layer neuron:
ω′hj＝ωhj+Δωhj＝ωhj+ηgjbh
Figure GDA0002680332920000051
output layer, hidden layer neuron threshold update:
θ′j＝θj+Δθj＝θj-ηgj
γ′h＝γh+Δγh＝γh-ηeh
calculating the accumulated error in the training set D:
Figure GDA0002680332920000052
Figure GDA0002680332920000053
step S403: identifying the actual atmosphere environment by using the trained gas identification network model;
(a) according to different sample types, a plurality of groups of samples are collected by using the sensor array, and a test set T is constructed:
Figure GDA0002680332920000054
in the formula, n represents the number of test set samples;
(b) and classifying the test set samples by using the trained gas recognition network model, and checking the reliability of the model.
Preferably, the step S100 includes:
step S101: processing a sensor substrate: processing a designed electrode pattern on the sensor substrate by an electron beam evaporation coating process and a photoetching process, and leading out corresponding electrode leads, wherein each electrode pair is crossed in a brush shape and does not intersect;
step S102: preparing a nano gas-sensitive film: respectively compounding graphene with tin oxide (SnO2), indium oxide (In2O3), cerium oxide (CeO2) and tungsten oxide (WO3) by adopting a hydrothermal chemical synthesis method to form different nano gas-sensitive materials, mixing the obtained nano gas-sensitive materials with ethanol, and dropwise coating the mixture on the surfaces of all interdigital electrode groups to form a nano gas-sensitive film;
step S103: assembling a sensor: different nanometer gas-sensitive materials are attached to the surfaces of different interdigital electrode groups, and the same nanometer gas-sensitive material is attached to the surfaces of interdigital electrodes in the same interdigital electrode group.
Preferably, the sensor array comprises a sensor substrate and a nano gas-sensitive film; the sensor base comprises a sensor substrate and an interdigital electrode group; the interdigital electrode group comprises interdigital electrodes; the nano gas-sensitive film is attached to the surface of the interdigital electrode group.
Preferably, the preparation material of the sensor substrate comprises any one of monocrystalline silicon, glass and ceramic.
The method according to claim 3, wherein the interdigital electrode is made of a material comprising any one of gold, platinum and silver-palladium, and the thickness of the interdigital electrode is 50-300 nm.
Preferably, the width of the interdigital electrode is 20-200 um, and the distance between the interdigital electrodes is 20-200 um.
Preferably, the preparation material of the nano gas-sensitive film comprises reduced graphene oxide, tin oxide, indium oxide, cerium oxide and tungsten oxide.
Preferably, the thickness of the nano gas-sensitive film is 100 nm-1 um.
Preferably, the characteristic parameter is a response signal of the sensor array under each atmosphere, and the atmosphere refers to a mixed gas obtained by mixing a plurality of gases according to different proportions.
Compared with the prior art, the invention has the following beneficial effects:
1. the sensor array has the advantages of simple structure, small volume, low cost and simple processing steps;
2. the sensor array provided by the invention is loaded with the graphene-metal oxide nano gas-sensitive material, can work at a low temperature, can effectively reduce energy consumption, and has the advantages of high sensitivity, short response recovery time and good repeatability;
3. the sensor array is composed of a plurality of interdigital electrode groups, and the interdigital electrode groups can obtain a plurality of output results of the same gas sensitive material on the premise of not increasing the number of sensors, thereby providing more input quantity for a pattern recognition algorithm and bringing possibility for improving the accuracy of the sensors;
4. the sensor array combines the gas identification network algorithm of error quasi-propagation, and can realize multi-component SF6The identification of the types and the contents of the gas decomposition products effectively solves the defect that a single sensor is sensitive to the cross of various gases.
Drawings
Fig. 1 shows an SF based on a graphene composite sensor array according to an embodiment of the present invention6A flow chart of a method for detecting a decomposition product;
fig. 2 is a flow chart illustrating a processing and preparation method of a graphene composite material based sensor array according to an embodiment of the present invention;
fig. 3 is a schematic structural diagram of a graphene composite material based sensor array according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a sensor array signal acquisition shown in an embodiment of the present invention;
FIG. 5 shows a sensor pair with different SF concentrations of 20ppm based on reduced graphene oxide/tin dioxide nano gas sensitive materials according to an embodiment of the present invention6A dynamic response curve diagram of the gas decomposition product;
FIG. 6 shows an embodiment of the present invention illustrating a sensor array for different SF concentrations of 20ppm6A response result diagram of the decomposition product;
FIG. 7 is a flow chart illustrating gas identification network training based on an error back propagation algorithm according to an embodiment of the present invention;
fig. 8 is a schematic diagram of a gas identification network model based on an error back propagation algorithm according to an embodiment of the present invention.
Detailed Description
The embodiments of the present invention will be described in detail below with reference to the accompanying drawings, which are only a part of the embodiments of the present invention and are not intended to limit the present invention.
SF based on graphene composite material sensor array6The analyte detection method, as shown in fig. 1, includes:
step S100: processing and preparing a sensor array;
step S200: arranging the prepared sensor array in a closed air chamber, and performing power-on initialization;
step S300: performing gas-sensitive test on the sensor array and storing the test result as a sample;
step S400: and constructing a gas identification network model and identifying the atmosphere, wherein the atmosphere is mixed gas obtained by mixing a plurality of gases according to different proportions.
In the embodiment of step S100, a specific manufacturing process of the sensor array is shown in fig. 2, and specifically includes:
step S101: processing a sensor substrate: processing a designed electrode pattern on the sensor substrate by an electron beam evaporation coating process and a photoetching process, and leading out corresponding electrode leads, wherein each electrode pair is crossed in a brush shape but not intersected;
step S102: preparing a nano gas-sensitive film: respectively compounding graphene with tin oxide (SnO2), indium oxide (In2O3), cerium oxide (CeO2) and tungsten oxide (WO3) by adopting a hydrothermal chemical synthesis method to form different nano gas-sensitive materials, mixing the obtained nano gas-sensitive materials with ethanol, and dropwise coating the mixture on the surfaces of all interdigital electrode groups to form a nano gas-sensitive film;
step S103: assembling a sensor: different nanometer gas-sensitive materials are attached to the surfaces of different interdigital electrode groups, and the same nanometer gas-sensitive material is attached to the surfaces of interdigital electrodes in the same interdigital electrode group.
The sensor array prepared by the present embodiment, as shown in fig. 3, includes: sensor substrate, nanometer gas sensitive film. The sensor base comprises a sensor substrate and 4 interdigital electrode groups, wherein each interdigital electrode group comprises 4 interdigital electrodes, the sensor substrate can adopt monocrystalline silicon, glass or ceramic as a preparation material, and preferably adopts monocrystalline silicon; the interdigital electrode can adopt gold, platinum or silver-palladium as a preparation material, and the thickness of the interdigital electrode is 50-300 nm, preferably 100 nm; the width of the interdigital electrodes is 20-200 um, preferably 100um, the distance between the interdigital electrodes is 20-200 um, preferably 100um, and the resistance between the interdigital electrodes is 1k omega-1G omega; the nano gas-sensitive film is prepared by respectively reducing graphene oxide (rGO) and tin oxide (SnO)2) Indium oxide (In)2O3) Cerium oxide (CeO)2) And tungsten oxide (WO)3) The prepared interdigital electrode is attached to the surface of an interdigital electrode group, has the thickness of 100 nm-1 um, preferably 300nm, and is used for realizing the electric connection between the interdigital electrodes;
the 4 prepared nanometer gas-sensitive films are respectively attached to the surfaces of 4 interdigital electrode groups, and each interdigital electrode group can obtain 4 similar output results at one time aiming at a certain single atmosphere, so that the scale of the sensor array is reduced, and the test precision is also ensured. The 4 interdigital electrode groups in the embodiment can generate 16 groups of signals, and the identification of the concentration of the multi-component gas can be realized.
In the specific implementation mode of the step S200, the sensor array is arranged in a sealed gas chamber with a volume of 800mL, and the atmosphere environment in the gas chamber is switched by means of dynamic gas distribution. The sensor array response signal acquisition circuit is shown in fig. 4, and corresponding connecting terminals are led out from the gas chamber, so that the gas-sensitive resistor R of the sensorciRespectively connected with a standard resistor RbiIn series, the sensor response signal is tested by applying a voltage of 5V across the terminalsNumber ViThen the gas-sensitive resistor R can be calculatedciThe variation of (2). The sensor response value S is defined as the relative variation of the nano gas-sensitive film resistance:
Figure GDA0002680332920000101
in the formula, RaRepresenting the resistance, R, of the sensor in the background atmospheregRepresenting the resistance of the sensor in the target atmosphere.
After the sensor array is arranged, power-on initialization is carried out to test whether the NI-USB-6218 data acquisition card can transmit the signal ViAnd transmitting to an upper computer.
In the specific implementation process of step S300, under the condition that the sensor array requires on-line monitoring, the test gas chamber is filled with background gas (SF)6) And standard gas (balance gas is SF)6SOF of2、SO2F2、SO2、H2S) and controlling the concentration of the target gas by changing the flow rate ratio of the mass flow controller, and recording the resistance change condition of the sensor under the target gas concentration.
In this embodiment, SOF is used2And H2S two typical SF6The decomposition product gas mixture is exemplified, and classification of species and identification of concentration are carried out. The gas concentration points were 0, 10, and 20ppm, respectively, for a total of 9 combinations of atmospheres, as shown in table 1:
TABLE 1
Figure GDA0002680332920000111
According to 9 atmosphere environments shown in the table 1, the output conditions of the sensor arrays are respectively tested, and the array output results are uploaded to an upper computer and stored as a sample set. Wherein, fig. 5 shows that the reduced graphene oxide/tin dioxide nano gas-sensitive material is used for different SF6Dynamic response curves of the gas decomposition products, FIG. 6 shows the response results of the sensor array, wherein the response results areAnd the error of the mean value of the response values of 4 interdigital electrode pairs in each interdigital electrode group is obtained by calculating the standard deviation of the response values.
In the specific implementation process of step S400, a gas identification network model based on an error back propagation algorithm is constructed for SF6The decomposition product of (a) is identified, as shown in fig. 7, the specific construction steps are as follows:
step S401: collecting characteristic parameters of the sensor array in different atmosphere environments;
the characteristic parameters are response signals of the sensor array under each atmosphere, and a sensor response value S is extracted as the characteristic parameters, so that the following requirements are met:
S＝Sij (2)
step S402: establishing and training a gas recognition network model based on an error inverse propagation algorithm, wherein the specific steps comprise;
(a) normalization processing of characteristic parameters:
extracting input characteristic parameters of each sample:
Figure GDA0002680332920000121
in the formula (I), the compound is shown in the specification,
Figure GDA0002680332920000122
the ith output of the sensor array for the kth sample, d represents the dimension of the input characteristic parameter.
Establishing output characteristic parameters according to sample types (different atmosphere environments):
Figure GDA0002680332920000123
in the formula (I), the compound is shown in the specification,
Figure GDA0002680332920000124
denotes the jth atmosphere of the kth sample, and m denotes the dimension of the output characteristic parameter.
Figure GDA0002680332920000125
And taking 0 or 1, wherein 1 is yes, and 0 is not, and distinguishing each sample type according to the output characteristic parameters.
In summary, a training set D can be obtained:
Figure GDA0002680332920000126
in the formula, l represents the total number of samples.
And ix, normalizing the input and output characteristic parameters:
Figure GDA0002680332920000127
Figure GDA0002680332920000128
in the formula (I), the compound is shown in the specification,
Figure GDA0002680332920000129
the characteristic parameters before the normalization are carried out,
Figure GDA00026803329200001210
is a normalized characteristic parameter.
Figure GDA00026803329200001211
Is the minimum value of the characteristic parameter in the kth sample,
Figure GDA0002680332920000131
is the maximum value of the characteristic parameter in the kth sample.
Figure GDA0002680332920000132
Is the output parameter before the normalization,
Figure GDA0002680332920000133
and outputting parameters after normalization.
Obtaining a normalized training set D:
Figure GDA0002680332920000134
(b) a gas recognition network model based on an error inverse propagation algorithm is established, as shown in fig. 8, and includes 16 input layer neurons, 20 hidden layer neurons, and 9 output layer neurons, and the specific training process is as follows:
defining a learning rate eta epsilon (0, 1), and randomly initializing all connection weights and threshold values in the neural network.
Calculating neural network output
Figure GDA0002680332920000135
Figure GDA0002680332920000136
Figure GDA0002680332920000137
Figure GDA0002680332920000138
Figure GDA0002680332920000139
In the formula, alphahInput to the first neuron of the hidden layer, bhIs the output of the h-th neuron of the hidden layer. Beta is ajIs the input of the jth output neuron; gamma rayhThreshold for the h-th neuron of the hidden layer, θjA threshold for the jth neuron of the output layer; v. ofihThe connection weight, omega, between the ith neuron of the input layer and the h-th neuron of the hidden layerhjThe connection weight between the h-th neuron of the hidden layer and the j-th neuron of the output layer is set; d represents the input layer neuron number and q represents the hidden layer neuron number.
Calculating gradient term g for output layer neuronsj：
Figure GDA0002680332920000141
Calculating gradient term e for cryptic neuronsh：
Figure GDA0002680332920000142
In the formula, m represents the number of neurons in the output layer.
xv. combining the calculated output layer, hidden layer neuron gradient terms, training the neural network.
Updating the connection right of the output layer neuron and the hidden layer neuron:
ω′hj＝ωhj+Δωhj＝ωhj+ηgjbh (15)
Figure GDA0002680332920000143
output layer, hidden layer neuron threshold update:
θ′j＝θj+Δθj＝θj-ηgj (17)
γ′h＝γh+Δγh＝γh-ηeh (18)
computing the cumulative error in the training set D:
Figure GDA0002680332920000144
Figure GDA0002680332920000145
the final goal of the algorithm is to have a training setThe cumulative error E over D is minimized. Computing output by comparing networks
Figure GDA0002680332920000146
And actual output
Figure GDA0002680332920000147
And (4) stopping training the neural network and entering a testing stage (step (3)) when the accumulated error meets the precision requirement, and otherwise, continuously repeating the step (b).
Step S403: the method comprises the following steps of recognizing the actual atmosphere environment by using a trained gas recognition network model, wherein the specific steps are as follows:
(a) according to different sample types, a plurality of groups of samples are collected by using the sensor array, and a test set T is constructed:
Figure GDA0002680332920000151
in the formula, n represents the number of test set samples.
In this embodiment, in order to identify 9 possible atmosphere environments, 16 sets of response signals are acquired as one sample through 4 interdigital electrode sets, and 180 sets of samples in total of 20 sets of response results are acquired in each atmosphere environment to train the gas identification network model, and if the error reaches 10-4I.e. the training is considered to be complete.
(b) And classifying the test set samples by using the trained gas recognition network model, and checking the reliability of the model.
In this embodiment, the test set T is classified by using the gas recognition network model obtained by the training, and the recognition result is shown in table 2:
TABLE 2
Figure GDA0002680332920000152
According to the identification result, the sensor array can accurately identify the SOF2And H2And (4) component content of S mixed gas.
In the following, the present invention provides the prior art method for SF6The results of the decomposition are shown in Table 3:
TABLE 3
Figure GDA0002680332920000161
Compared with the existing method, the graphene-metal oxide nanocomposite sensor array disclosed by the invention is combined with a gas identification network model based on an error inverse propagation network algorithm, so that the identification of the component content of the mixed gas can be realized, the online monitoring potential is also realized, and a solid foundation is laid for realizing the fault diagnosis of high-voltage power equipment.
The principle and the implementation of the present invention are explained by applying the specific embodiments of the present invention, and the above description of the embodiments is only used to help understand the using method and the core idea of the present invention. The present invention may be modified in the specific embodiments and applications according to the actual situations, and the above embodiments do not limit the application scope of the present invention. The addition, modification and replacement of technical features with those of the prior art are all within the scope of the present invention without departing from the technical features of the present invention.

Claims (9)

1. SF based on graphene composite material sensor array6A method for detecting a decomposition product, comprising:
step S100: processing and preparing a sensor array;
step S200: arranging the prepared sensor array in a closed air chamber for signal acquisition, and electrifying for initialization;
step S300: performing gas-sensitive test on the sensor array and storing the test result as a sample;
step S400: constructing a gas identification network model and identifying atmosphere;
wherein, step S400 includes:
step S401: collecting characteristic parameters of the sensor array in different atmosphere environments;
step S402: establishing a gas recognition network model based on an error inverse propagation algorithm and training;
(a) normalization processing of characteristic parameters:
i. extracting input characteristic parameters of each sample:
Figure FDA0002680332910000011
in the formula (I), the compound is shown in the specification,
Figure FDA0002680332910000012
the ith output result of the sensor array of the kth sample, d represents the dimension of the input characteristic parameter;
establishing an output characteristic parameter according to the sample type:
Figure FDA0002680332910000013
in the formula (I), the compound is shown in the specification,
Figure FDA0002680332910000014
representing the jth atmosphere of the kth sample, m representing the dimension of the output characteristic parameter,
Figure FDA0002680332910000021
taking 0 or 1, if 1 is yes, and if not, 0 is not, and distinguishing each sample type according to the output characteristic parameters;
obtaining a training set D:
Figure FDA0002680332910000022
wherein l represents the total number of samples;
and iii, carrying out normalization processing on the input and output characteristic parameters:
Figure FDA0002680332910000023
Figure FDA0002680332910000024
in the formula (I), the compound is shown in the specification,
Figure FDA0002680332910000025
is the characteristic parameter before the normalization,
Figure FDA0002680332910000026
in order to obtain the normalized characteristic parameters,
Figure FDA0002680332910000027
is the minimum value of the characteristic parameter in the kth sample,
Figure FDA0002680332910000028
is the maximum value of the characteristic parameter in the kth sample,
Figure FDA0002680332910000029
is the output parameter before the normalization,
Figure FDA00026803329100000210
outputting parameters after normalization;
obtaining a normalized training set D:
Figure FDA00026803329100000211
(b) establishing a gas recognition network model based on an error inverse propagation algorithm, wherein the specific training process is as follows:
i. defining a learning rate eta epsilon (0, 1), and randomly initializing all connection weights and threshold values in the neural network;
calculating neural network output
Figure FDA00026803329100000212
Figure FDA0002680332910000031
Figure FDA0002680332910000032
Figure FDA0002680332910000033
Figure FDA0002680332910000034
In the formula, alphahInput to the h-th neuron of the hidden layer, bhIs the output of the h-th neuron of the hidden layer, betajIs the input of the jth output neuron, gammahThreshold for the h-th neuron of the hidden layer, θjIs the threshold of the jth neuron of the output layer, uihIs the connection weight between the ith neuron of the input layer and the h neuron of the hidden layer, omegahjThe connection weight between the h-th neuron of the hidden layer and the j-th neuron of the output layer is shown, d represents the number of neurons of the input layer, and q represents the number of neurons of the hidden layer;
computing gradient term g for output layer neuronsj：
Figure FDA0002680332910000035
Computing gradient term e for hidden neuronsh：
Figure FDA0002680332910000036
Wherein m represents the number of neurons in the output layer;
training a neural network by combining the calculated output layer and hidden layer neuron gradient terms;
updating the connection right of the output layer neuron and the hidden layer neuron:
ω′hj＝ωhj+Δωhj＝ωhj+ηgjbh
Figure FDA0002680332910000041
output layer, hidden layer neuron threshold update:
θ′j＝θj+Δθj＝θj-ηgj
γ′h＝γh+Δγh＝γh-ηeh
v. calculate the cumulative error in training set D:
Figure FDA0002680332910000042
Figure FDA0002680332910000043
step S403: identifying the actual atmosphere environment by using the trained gas identification network model;
(a) according to different sample types, a plurality of groups of samples are collected by using the sensor array, and a test set T is constructed:
Figure FDA0002680332910000044
in the formula, n represents the number of test set samples;
(b) and classifying the test set samples by using the trained gas recognition network model, and checking the reliability of the model.
2. The method according to claim 1, wherein the step S100 preferably comprises:
step S101: processing a sensor substrate: processing a designed electrode pattern on the sensor substrate by an electron beam evaporation coating process and a photoetching process, and leading out corresponding electrode leads, wherein each electrode pair is crossed in a brush shape and does not intersect;
step S102: preparing a nano gas-sensitive film: respectively mixing graphene with tin oxide (SnO) by hydrothermal chemical synthesis method2) Indium oxide (In)2O3) Cerium oxide (CeO)2) Tungsten oxide (WO)3) Compounding to form different nano gas-sensitive materials, mixing and melting the obtained nano gas-sensitive materials with ethanol, and dripping the mixture on the surface of each interdigital electrode group to form a nano gas-sensitive film;
step S103: assembling a sensor: different nanometer gas-sensitive materials are attached to the surfaces of different interdigital electrode groups, and the same nanometer gas-sensitive material is attached to the surfaces of interdigital electrodes in the same interdigital electrode group.
3. The method of claim 1, wherein the sensor array comprises a sensor substrate, a nano gas-sensitive film; the sensor base comprises a sensor substrate and an interdigital electrode group; the interdigital electrode group comprises interdigital electrodes; the nano gas-sensitive film is attached to the surface of the interdigital electrode group.
4. The method of claim 3, wherein the sensor substrate is made of a material including any one of single crystal silicon, glass, and ceramic.
5. The method according to claim 3, wherein the interdigital electrode is made of a material comprising any one of gold, platinum and silver-palladium, and the thickness of the interdigital electrode is 50-300 nm.
6. The method according to claim 3, wherein the width of the interdigital electrodes is 20-200 μm, and the pitch of the interdigital electrodes is 20-200 μm.
7. The method of claim 3, wherein the nano gas-sensitive thin film is prepared from materials including reduced graphene oxide, tin oxide, indium oxide, cerium oxide, and tungsten oxide.
8. The method of claim 3, wherein the nano gas-sensitive film has a thickness of 100nm to 1 μm.
9. The method of claim 1, wherein the characteristic parameter is a response signal of the sensor array under each atmosphere, and the atmosphere is a mixed gas obtained by mixing a plurality of gases in different proportions.
CN201810666669.2A 2018-06-26 2018-06-26 SF based on graphene composite material sensor array6Method for detecting decomposition product Active CN109060892B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810666669.2A CN109060892B (en) 2018-06-26 2018-06-26 SF based on graphene composite material sensor array6Method for detecting decomposition product

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810666669.2A CN109060892B (en) 2018-06-26 2018-06-26 SF based on graphene composite material sensor array6Method for detecting decomposition product

Publications (2)

Publication Number Publication Date
CN109060892A CN109060892A (en) 2018-12-21
CN109060892B true CN109060892B (en) 2020-12-25

Family

ID=64821536

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810666669.2A Active CN109060892B (en) 2018-06-26 2018-06-26 SF based on graphene composite material sensor array6Method for detecting decomposition product

Country Status (1)

Country Link
CN (1) CN109060892B (en)

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110038515A (en) * 2019-04-28 2019-07-23 三峡大学 GIS characteristic gas adsorbed film
CN111912877A (en) * 2019-05-09 2020-11-10 天津大学 Sensor array-based organic gas detection and identification chip
CN110530935B (en) * 2019-08-31 2020-12-29 中国石油大学(华东) Construction method of molybdenum disulfide-based gas-sensitive sensing array and SF (sulfur hexafluoride) sensor6Application in gas decomposition component detection
CN110647989B (en) * 2019-09-16 2023-04-18 长春师范大学 Graphene defect modification prediction method based on neural network
CN110688751A (en) * 2019-09-24 2020-01-14 西南大学 Simulation method for detecting SF6 by using platinum-doped modified graphite alkyne sensor
CN112505097A (en) * 2020-08-20 2021-03-16 江门市阳邦智能科技有限公司 Cerium oxide sensor and preparation method thereof
CN113848238A (en) * 2021-09-24 2021-12-28 广东电网有限责任公司 Composite material based on cerium oxide/graphene, preparation method and application thereof, and sulfuryl fluoride gas-sensitive sensor
CN114018989A (en) * 2021-11-05 2022-02-08 广东电网有限责任公司 Miniature array type gas sensor for detecting sulfur hexafluoride decomposition products

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103346799A (en) * 2013-06-05 2013-10-09 中国科学院微电子研究所 Method for identifying gas based on compressed sensing theory
CN105241497A (en) * 2015-09-23 2016-01-13 国网山东省电力公司日照供电公司 Transformer monitoring system and fault diagnosis method
CN106443259A (en) * 2016-09-29 2017-02-22 国网山东省电力公司电力科学研究院 Transformer fault diagnosis new method based on Euclidean clustering and SPO-SVM
CN106896219A (en) * 2017-03-28 2017-06-27 浙江大学 The identification of transformer sub-health state and average remaining lifetime method of estimation based on Gases Dissolved in Transformer Oil data
CN107180983A (en) * 2017-05-16 2017-09-19 华中科技大学 A kind of SOFC pile method for diagnosing faults and system
CN107256393A (en) * 2017-06-05 2017-10-17 四川大学 The feature extraction and state recognition of one-dimensional physiological signal based on deep learning

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103149246A (en) * 2012-09-27 2013-06-12 中国石油大学(华东) Graphene film humidity sensor
US10001516B2 (en) * 2016-02-03 2018-06-19 International Business Machines Corporation Reducing noise and enhancing readout throughput in sensor array
CN105891271B (en) * 2016-03-31 2018-08-07 吉林大学 It is a kind of based on graphene/resistor-type gas sensor of stannic oxide/zinc oxide composite, preparation method and applications
CN106093135B (en) * 2016-06-02 2020-11-10 中国石油大学(华东) Power transformer fault intelligent diagnosis device based on graphene gas sensor array
CN106248776B (en) * 2016-08-03 2019-03-01 西安交通大学 A kind of sensor array of detectable multiple groups part mixed gas
CN106290488B (en) * 2016-09-18 2020-01-10 江南大学 Amino-functionalized carbon nanotube resistance type formaldehyde gas sensor and preparation method thereof
CN106680328A (en) * 2017-01-04 2017-05-17 清华大学深圳研究生院 Gas sensor array and manufacturing method thereof

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103346799A (en) * 2013-06-05 2013-10-09 中国科学院微电子研究所 Method for identifying gas based on compressed sensing theory
CN105241497A (en) * 2015-09-23 2016-01-13 国网山东省电力公司日照供电公司 Transformer monitoring system and fault diagnosis method
CN106443259A (en) * 2016-09-29 2017-02-22 国网山东省电力公司电力科学研究院 Transformer fault diagnosis new method based on Euclidean clustering and SPO-SVM
CN106896219A (en) * 2017-03-28 2017-06-27 浙江大学 The identification of transformer sub-health state and average remaining lifetime method of estimation based on Gases Dissolved in Transformer Oil data
CN107180983A (en) * 2017-05-16 2017-09-19 华中科技大学 A kind of SOFC pile method for diagnosing faults and system
CN107256393A (en) * 2017-06-05 2017-10-17 四川大学 The feature extraction and state recognition of one-dimensional physiological signal based on deep learning

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
基于LE-ELM的热力参数传感器故障诊断;吕广红等;《仪表技术与传感器》;20171231(第10期);第93-97页 *
基于嗅觉网络传输的重症疾病诊断机制与算法研究;李联宁;《计算机科学》;20150630;第42卷(第6期);第128-133页 *

Also Published As

Publication number Publication date
CN109060892A (en) 2018-12-21

Similar Documents

Publication Publication Date Title
CN109060892B (en) SF based on graphene composite material sensor array6Method for detecting decomposition product
CN204666549U (en) Based on the mixed gas detection system of BP neural network
Netto et al. Biopolymer based ionogels as active layers in low-cost gas sensors for electronic noses
Zhang et al. An electronic pollen detection method using Coulter counting principle
US11525797B2 (en) Method for detecting an air discharge decomposed product based on a virtual sensor array
CN107944213A (en) The online Source Apportionments of PMF, system, terminal device and computer-readable recording medium
Bellando et al. Sensitivity, noise and resolution in a BEOL-modified foundry-made ISFET with miniaturized reference electrode for wearable point-of-care applications
AU2020102518A4 (en) A method of gas sensor calibration based on linear optimization
CN103364703B (en) The quick evaluating method of reliability under the many stress of LED
Chen et al. A quick classifying method for tracking and erosion resistance of htv silicone rubber material via laser-induced breakdown spectroscopy
CN112834572B (en) Gas sensor testing method applied to detection of sulfur hexafluoride and decomposition product gas sensor thereof
CN105354347A (en) Magnesite ore grade classification modeling method
CN102095782B (en) Gas on-line detection device based on micro-nano carbon nano tube film three-electrode
Bošković et al. Full-self-powered humidity sensor based on electrochemical aluminum–water reaction
CN110763830B (en) Method for predicting content of free calcium oxide in cement clinker
CN104849322A (en) Impedance biosensor and bio-impedance detection analysis method
CN115809427B (en) Mixed gas identification method based on mixed strategy optimization BP neural network
CN106706719A (en) Tin sulfide nano flower film sensitive to ammonia gas at low temperature
CN103336026A (en) Polymer piezoelectric gas sensor system for detecting gases
CN107703215A (en) A kind of GIS device discharge fault chemical diagnosis method
CN102095792B (en) Online gas detection device based on carbon-nano-tube membrane micron-nano electrode
Ni et al. Classification and Concentration Predictions of Volatile Organic Compounds Using an Electronic Nose Based on XGBoost-Random Forest Algorithms
Casans et al. Novel resistance measurement method: Analysis of accuracy and thermal dependence with applications in fiber materials
Kulmala et al. The legacy of Finnish–Estonian air ion and aerosol workshops
Kerkhoff et al. A dependable microelectronic peptide synthesizer using electrode data

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