CN116626238A - Dual-channel detection compass system for sensing air flow and air combination, air leakage detection method, data fusion and tracking method - Google Patents
Dual-channel detection compass system for sensing air flow and air combination, air leakage detection method, data fusion and tracking method Download PDFInfo
- Publication number
- CN116626238A CN116626238A CN202310596318.XA CN202310596318A CN116626238A CN 116626238 A CN116626238 A CN 116626238A CN 202310596318 A CN202310596318 A CN 202310596318A CN 116626238 A CN116626238 A CN 116626238A
- Authority
- CN
- China
- Prior art keywords
- gas
- data
- concentration
- network
- sensor
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 78
- 238000000034 method Methods 0.000 title claims abstract description 65
- 230000004927 fusion Effects 0.000 title claims abstract description 24
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 30
- 238000012545 processing Methods 0.000 claims abstract description 30
- 239000007789 gas Substances 0.000 claims description 252
- 239000010410 layer Substances 0.000 claims description 118
- 210000002569 neuron Anatomy 0.000 claims description 62
- 238000009792 diffusion process Methods 0.000 claims description 60
- 238000012549 training Methods 0.000 claims description 57
- 230000006870 function Effects 0.000 claims description 49
- 238000009826 distribution Methods 0.000 claims description 34
- 238000013528 artificial neural network Methods 0.000 claims description 25
- 238000010606 normalization Methods 0.000 claims description 22
- 230000008569 process Effects 0.000 claims description 19
- 230000008859 change Effects 0.000 claims description 18
- 238000007781 pre-processing Methods 0.000 claims description 16
- 239000000779 smoke Substances 0.000 claims description 14
- 230000005540 biological transmission Effects 0.000 claims description 12
- 238000010586 diagram Methods 0.000 claims description 10
- 238000012360 testing method Methods 0.000 claims description 10
- 241000287127 Passeridae Species 0.000 claims description 9
- 238000012546 transfer Methods 0.000 claims description 9
- 230000009977 dual effect Effects 0.000 claims description 6
- 238000012544 monitoring process Methods 0.000 claims description 6
- 238000004445 quantitative analysis Methods 0.000 claims description 6
- 230000004044 response Effects 0.000 claims description 6
- 238000001914 filtration Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000012937 correction Methods 0.000 claims description 4
- 238000010845 search algorithm Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 230000001186 cumulative effect Effects 0.000 claims description 3
- 230000001419 dependent effect Effects 0.000 claims description 3
- 238000007599 discharging Methods 0.000 claims description 3
- 230000005284 excitation Effects 0.000 claims description 3
- 210000005036 nerve Anatomy 0.000 claims description 3
- 239000002574 poison Substances 0.000 claims description 3
- 231100000614 poison Toxicity 0.000 claims description 3
- 238000003825 pressing Methods 0.000 claims description 3
- 230000000087 stabilizing effect Effects 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 239000011229 interlayer Substances 0.000 claims description 2
- 239000002131 composite material Substances 0.000 claims 1
- 239000000203 mixture Substances 0.000 claims 1
- 230000035945 sensitivity Effects 0.000 abstract description 3
- 238000005259 measurement Methods 0.000 description 10
- 238000013461 design Methods 0.000 description 4
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 description 4
- LFQSCWFLJHTTHZ-UHFFFAOYSA-N Ethanol Chemical compound CCO LFQSCWFLJHTTHZ-UHFFFAOYSA-N 0.000 description 3
- 230000002431 foraging effect Effects 0.000 description 3
- QGZKDVFQNNGYKY-UHFFFAOYSA-N Ammonia Chemical compound N QGZKDVFQNNGYKY-UHFFFAOYSA-N 0.000 description 2
- ATUOYWHBWRKTHZ-UHFFFAOYSA-N Propane Chemical compound CCC ATUOYWHBWRKTHZ-UHFFFAOYSA-N 0.000 description 2
- 230000006399 behavior Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000012790 confirmation Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000007667 floating Methods 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- OTMSDBZUPAUEDD-UHFFFAOYSA-N Ethane Chemical compound CC OTMSDBZUPAUEDD-UHFFFAOYSA-N 0.000 description 1
- 229910021529 ammonia Inorganic materials 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000003912 environmental pollution Methods 0.000 description 1
- 231100001261 hazardous Toxicity 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000003345 natural gas Substances 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 244000062645 predators Species 0.000 description 1
- 239000001294 propane Substances 0.000 description 1
- 238000005067 remediation Methods 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 238000009423 ventilation Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/0004—Gaseous mixtures, e.g. polluted air
- G01N33/0009—General constructional details of gas analysers, e.g. portable test equipment
-
- 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
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/0004—Gaseous mixtures, e.g. polluted air
- G01N33/0009—General constructional details of gas analysers, e.g. portable test equipment
- G01N33/0062—General constructional details of gas analysers, e.g. portable test equipment concerning the measuring method or the display, e.g. intermittent measurement or digital display
Landscapes
- Chemical & Material Sciences (AREA)
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Medicinal Chemistry (AREA)
- Food Science & Technology (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- Combustion & Propulsion (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Investigating Or Analyzing Materials By The Use Of Electric Means (AREA)
Abstract
A dual-channel detection compass system for sensing air flow and air combination, a gas leakage detection method and a data fusion and tracking method belong to the field of gas detection and positioning systems and methods. Existing dangerous gas detection has accidental errors and is affected by cross sensitivity. The invention comprises the following steps: the method comprises the steps that gas to be detected enters a multi-sensor array, the sensor array outputs a voltage signal according to the content of dangerous gas in the gas to be detected, the SHT30 sends the temperature and humidity result of the measured exhaled gas to a data acquisition module, data are sent through a serial port, after nRF52832 receives the data, a plurality of data are fused to obtain a final concentration value, and finally the temperature and humidity, the residual electric quantity and the concentration value after data fusion are sent; and displaying on an interface of the data processing, transmitting and displaying module. According to the method, the multi-sensor array and the data information fusion algorithm are combined, so that the content of the dangerous gas in the environment is detected more accurately, and the concentration and the position of the dangerous gas are judged more accurately.
Description
Technical Field
The invention relates to a multi-data fusion method, in particular to a dual-channel detection compass system for sensing air flow and air combination, an air leakage detection method and a data fusion and tracking method.
Background
Methane (natural gas), ethane, propane, ammonia and ethanol are several gases which are commonly contacted in our lives, and after reaching a certain concentration, the gases have certain dangers and have potential safety hazards. In order to avoid the potential safety hazards caused by these gases, how to rapidly and accurately detect the gases is a problem to be solved. For a long time, in order to avoid the hidden trouble caused by the gas, people strive to keep good ventilation indoors, and no related household electronic nose detection products exist temporarily. The electronic nose is only used in the aspects of industrial production, malodor detection, food detection and the like.
Dangerous gas non-human contact detection can be realized by a high-precision gas sensor, but single gas sensor measurement has a plurality of defects, such as accidental error of sensor measurement, cross sensitivity influence and the like. The combination of the multi-gas sensor array and the data fusion algorithm for measurement is clearly the most suitable scheme, so that on one hand, measurement errors caused by a single sensor can be reduced, on the other hand, the selectivity of the system to gas is improved, and the interference of other gases to the sensors is reduced.
The method for data fusion is widely used in the fields of pattern recognition, gas quantitative analysis and the like, and the BP neural network is most widely applied, so that the BP neural network is adopted to realize fusion of gas sensor data and compensation and correction of temperature and humidity, the selectivity of a system can be improved, the interference of external temperature and humidity can be reduced, the measurement error of the system can be reduced, and the measurement accuracy can be improved. The display of the detection result is generally traditional liquid crystal display, the display effect is single, and the current user requirement cannot be met, so that the APP display software capable of being displayed on the smart phone is planned to be designed, and better detection experience can be brought to a user.
Disclosure of Invention
The invention aims to solve the problems that the existing dangerous gas detection has accidental errors, is influenced by cross sensitivity and the like, and provides a dual-channel detection compass system for combined sensing of gas flow and gas, a gas leakage detection method and a data fusion and tracking method.
The above object is achieved by the following technical scheme:
a dual channel detection compass system with combined sensing of air flow and gas, comprising: the device comprises a charging and discharging module, a data acquisition module, a data processing, transmitting and displaying module; wherein,,
The charge-discharge module provides power support for the whole system and provides proper voltage for the system through the step-down voltage stabilizing circuit; comprising the following steps: a through hole conductive slip ring, a stepping motor and an angle correction rod which are connected with the sensor bracket and the base circuit component;
the data acquisition module is used for storing acquired data in the SD in a file form through a FATFS file system and then sending the data through a serial port; comprising the following steps: the device comprises a sensor bracket, a gas flow sensor, a wind direction sensing cabin and a via hole conductive slip ring;
the data acquisition module selects an STM32F103C8T6 microprocessor chip as the core of the singlechip module, adopts 3.3V voltage to supply power, has a maximum clock frequency of 72MHz, has 64KB FLASH and 20KB SRAM in the system, integrates 10 12-bit ADC analog-to-digital converters, USART, SPI, I C and other digital interfaces in the system, and supports serial port debugging and JTAC debugging modes;
the data processing, transmitting and displaying module is used for receiving the data sent by the data acquisition module, performing digital filtering, temperature and humidity compensation and data preprocessing, then fusing the data of the plurality of sensors to obtain a final concentration value, and then transmitting and displaying the temperature and humidity, the residual electric quantity and the concentration value after data fusion in a BLE wireless mode;
The data processing, transmitting and displaying module adopts an nRF52832 microprocessor MCU.
Further, the data acquisition module comprises a top plate component, a gas concentration detection cabin array, a wind speed and direction detection area array and a bottom plate component, and a data line connecting the top plate component and the bottom plate component is arranged in a hollow cylindrical cavity in the center of the device; the top plate component mainly comprises a microprocessor chip, a compass sensor, a gas sensor and a PCB (printed circuit board), wherein data acquired by the compass sensor and the temperature and humidity sensor are transmitted to the microprocessor chip, and the wind speed and direction detection area array comprises 4 wind speed and direction detection areas; the gas concentration detection cabin array comprises 4 gas concentration detection cabins, and 4 gas sensors with the same model are respectively arranged in the centers of the 4 gas concentration detection cabins in an inverted mode;
the sensor is an MEMS gas sensor for identifying the smell of harmful gas, and when the change of the concentration of the gas to be detected causes the change of the resistance value in the sensitive resistor, the back-end circuit realizes the detection of the concentration of the gas by processing the resistance signal.
A gas leakage detection method comprising:
the method comprises the steps of opening a power switch of a data processing, transmitting and displaying module and a data acquisition module, connecting a BLE signal transmitted by a detection system on an APP arranged on the data processing, transmitting and displaying module, beginning to measure by pressing, enabling gas to be detected to enter a multi-sensor array, enabling the sensor array to output a certain voltage signal according to dangerous gas content in the gas to be detected, enabling SHT30 to measure temperature and humidity of exhaled gas, transmitting a measured result to the data acquisition module STM32F103C8T6MCU, firstly storing acquired data in a SD in a file form through a FATFS file system, then transmitting the data to nRF52832 BLSoc through a serial port, carrying out digital filtering and data preprocessing on the data after the nRF52832 receives the data, then fusing the data of a plurality of sensors to obtain a final concentration value, and finally transmitting a temperature and humidity, residual electric quantity and a concentration value after data fusion through a transmission function of a transmission service;
And the data processing, transmitting and displaying module receives the data transmitted by the system and displays the data on an interface of the data processing, transmitting and displaying module.
A data fusion and tracking method for a dual channel detection compass system adapted for combined sensing of air flow and gas, said method comprising:
the basic structure of the BP neural network consists of three layers, namely an input layer, a hidden layer and an output layer, wherein a plurality of hidden layers are possible, neurons of adjacent layers of the network are connected with each other, neurons in each layer are not connected with each other, and the output of the neuron of the previous layer is used as the input of the neuron of the next layer; the characteristics of the whole network are determined by the connection weight of adjacent interlayer neurons and the threshold value on the neurons; the quantitative analysis and the temperature and humidity compensation of dangerous gas are realized through the sensor array and the BP network algorithm,
step one, designing a BP neural network algorithm;
firstly, initializing a network weight, and then inputting a training learning sample;
secondly, calculating the values of hidden layer and output layer neurons in the nerve network according to the set network weight and threshold value, and comparing the actual output value with a corresponding target value to obtain a network error;
Thirdly, judging whether training is finished;
there are three situations that can lead to the end of training: the first time for training and learning is far longer than the set training time; the second is that the cycle number in the training and learning process is larger than the initially set step number; thirdly, the error generated by the system after the sample training is finished is smaller than the initially set error;
the network is not converged after the first and second conditions are over, and the network is converged after the third condition is over;
judging that the training is finished when the third condition is met, and storing the network;
step two, designing the structure of the BP neural network; the method comprises the steps of setting network layers, input layer node numbers, hidden layer node numbers, output layer node numbers, transmission functions, training methods and training parameters;
step three, introducing a sparrow search algorithm to optimize the initial weight and the threshold value of the BP neural network;
designing a method for searching and positioning the leakage source position;
based on a Gaussian model, a gas leakage diffusion concentration gradient distribution map is obtained, so that a gas leakage accident poison area is divided, and a path backtracking and tracking algorithm is constructed; based on a Bayesian decision theory, a positioning algorithm of a gas leakage source is obtained, and an optimal solution of the position information of the gas leakage source is obtained;
Fifthly, constructing a gas leakage diffusion concentration distribution map;
step six, determining the type and concentration of the leaked gas at the same time;
the output values of different sensors are collected in the actual monitoring process, the change rate of the output values is calculated, and the output values are compared with a pre-stored characteristic response database in real time, so that the rapid identification and concentration prediction of the target gas type can be realized; the method comprises the following specific steps: (1) setting the output values of the 3 sensors as S1 (t), S2 (t) and S3 (t), and setting the change rates of the output values as R1 (t) and R2 (t);
(2) substituting Si (t) and Ri (t) into cij=fij [ Si (t), ri (t) ], and calculating to obtain Cij, wherein i=1, 2, 3, j=1, 2, 3; the relation Cjk between the gas concentration Cjk and the sensor output value and the output value change rate is that sijk=fij [ Eijk (t), fijk (t) ], wherein Sijk (t) is the sensor output value, rijk (t) is the sensor output value change rate, cjk is the target gas concentration, i=1, 2, 3 is the sensor number, j=1, 2, 3 is the target gas type number, and k=1, 2, 3, 4 is the target gas concentration number;
(3) analyzing the value of Cij, comparing the calculated concentration C1j, C2j and C3j of different gases by the same sensor in groups, and calculating the concentration error of the target gas Taking the minimum value of Dj as DJ, judging that DJ is smaller than phi, if the minimum value of DJ is satisfied, J is the number of the currently detected target gas, and the concentration of the target gas is
Further, the structure of the BP neural network is designed in the second step; the method comprises the steps of setting network layers, input layer node numbers, hidden layer node numbers, output layer node numbers, transmission functions, training methods and training parameters, and specifically comprises the following steps:
a) Network layer number
The BP network may contain one to more hidden layers;
b) Input layer node number
The number of input layer nodes depends on the dimension of the input vector; the node number of the input layer is determined according to the number of sensors in the sensor array;
c) Hidden layer node number
Obtaining an estimated value of the number of hidden layer nodes:
wherein k is the number of samples, m is the number of neurons of the hidden layer, and n is the number of neurons of the input layer; if i > m, provision is made for
Wherein m and n are the number of neurons of the output layer and the input layer, respectively, and a is a constant between [0, 10 ];
M＝log 2 n (3)
wherein n is the number of neurons of the input layer;
through the three calculation formulas, the range of the number of neurons of the hidden layer can be preliminarily determined to be 3-13, and then the optimal value is determined by continuously adjusting the accuracy requirement to be achieved by training and the self-property of the selection function; or the number of neurons in the hidden layer is changed to be adjustable, or the number of neurons is ensured to be enough when the number of neurons is initially set, and redundant neurons are removed after continuous learning until the optimal value is obtained;
d) Number of neurons in output layer
The number of the neurons of the output layer is determined according to an abstract model obtained in the actual problem;
e) Selection of transfer functions
The implicit layer transfer function is 'log sig', and the output layer transfer function is 'purlin'; the value range of x in the function curve is between 0 and 1, and the input data is preprocessed to be converted into the range of 0 and 1 interval;
f) Selection of training method
Default "tranlm" function is training function in BP network, there is possibility of sinking into local minimum point when the network is trained, thus reduce the modeling accuracy of the system;
using the ReLU as an activation function, outputting 0 when the input of the ReLU is smaller than 0, and linearly increasing when the input of the ReLU is larger than 0;
g) Determination of initial weights
The BP network adopts an iterative updating mode to determine the weight, firstly, the initial weight is defined as a smaller non-zero random value, and the empirical value is (-2.4/F, 2.4/F) or (-3/F, 3/F), wherein F is the number of neurons connected with the weight input end;
h) Data preprocessing
The data preprocessing comprises normalization of gas sensor signals and normalization of gas concentration signals;
the gas sensor signal normalization algorithm is as follows:
where i=1, 2,3,4,5,6max [ i ] represents the maximum response value of the sensor i to all test samples;
The normalization algorithm of the gas concentration signal is as follows:
c′ j ＝c j /c max (5)
wherein, c max Maximum expected result for test set samples 100, j=1, 2, …,20
After the input samples of the test set are subjected to sensor signal normalization processing, each element of the input vector of the array is in the same order of magnitude, the gas concentration normalization can meet the requirement of the output amplitude of the neuron S-shaped excitation function, and when the network predicts, the output of the network obtains a gas prediction concentration value through the inverse transformation of a gas concentration signal normalization algorithm;
after the parameters are determined, the training data is subjected to data preprocessing and is input into a network for learning, and if the network is successfully converged, the required neural network can be obtained.
Further, in the fourth step, a process of searching and positioning the leakage source position is designed, specifically:
selecting a grid set with the highest probability from a possible source grid set of the gas diffusion plumes through reverse estimation as a source path of the gas diffusion plumes; assuming that a certain gas diffuses smoke plume at t 1 At the moment at the position L 1 At t 2 Time of arrival at position L 2 ，t 1＜ t 2 The flow process of the gas diffusion plume can be described as:
where U (L (t)) is the average flow velocity vector of the gas diffusion plume at the position L (t) at the time t, expressed as
[u x (L(t)),u y (L(t)),u z (L(t))] T N (t) is Gaussian noise; if the mobile robot is at t j At the moment at the position L R (t j ) Detecting the gas diffusion smoke plume at t j Some t before the moment l The position at the moment can be written as formula (7):
when the gas sensor detects the gas concentration, firstly, a suspected path of the gas diffusion smoke plume is obtained through reverse pushing, then, all possible gas source grid sets are searched nearby the suspected path according to a historical gas diffusion distribution likelihood diagram and historical wind speed and wind direction information, and finally, the grid set with the highest probability is selected as the source path of the gas diffusion smoke plume; and calculating all risks possibly brought by different decisions as conditional risks of each decision, and selecting the decision with smaller conditional risks as a result of gas leakage source position judgment.
Further, in the fifth step, the process of constructing the gas leakage diffusion concentration distribution map specifically includes:
expanding the discrete and sparse smell information into a gas leakage source diffusion concentration distribution likelihood map of the whole search space by adopting a Gaussian kernel function; dividing the three-dimensional space into three-dimensional grids with uniform density according to a proper scale, taking the three-dimensional space as a minimum space unit constructed by a gas distribution likelihood map, and giving each grid an accumulated weight; setting the cumulative weight of each grid to zero before the gas diffusion concentration is captured; when the gas diffusion concentration is captured, calculating the current weight of the grids in a certain range around by adopting a Gaussian kernel function, and accumulating the current weight of the grids; the independent variable of the Gaussian kernel function is the distance between a certain grid and a gas detection position, and the dependent variable is the current weight of the grid; and normalizing the accumulated weight of all grids to obtain the probability distribution of gas concentration diffusion, namely a gas leakage source diffusion concentration distribution likelihood map, and providing the probability distribution likelihood map as an initial condition for a gas distribution likelihood map updating process in subsequent gas diffusion plume source path backtracking and tracing.
The beneficial effects of the invention are as follows:
the method for detecting the content of the environmental dangerous gas by adopting the multi-sensor array is used for detecting the dangerous gas, and has the characteristics of simple detection operation, high detection speed, continuous detection and the like. The specific points are as follows:
(1) Compared with the traditional single-sensor system measurement method, the detection system based on the gas compass has the advantages of wide detection range, high identification accuracy and high detection speed; the detection flow is simple, and the real-time monitoring and the repair by the maintenance personnel are facilitated.
(2) Through optimizing the inside detection cabin structure of gas compass, enriched through the gas and the air current that detect the cabin, promoted the accuracy that detects greatly, promoted the reliability that the compass was used for detecting and gas tracking to a greater extent.
(3) The sensor array formed by a plurality of different gas sensors is adopted to realize the quantitative detection of dangerous gas, and the selectivity of the gas sensor to the gas is improved, the interference of the gas in other air is eliminated, the error is reduced, and the accuracy of quantitative detection is improved through a corresponding data information fusion algorithm, namely an artificial neural network algorithm.
(4) By adopting BLE wireless communication technology, the intelligent mobile phone can be connected with the detection device, the detection result can be seen through APP detection software on the intelligent mobile phone, and the detection data can be stored.
The combination of the multi-sensor array and the data information fusion algorithm can more accurately detect the content of dangerous gas in the environment, so that the concentration and the approximate position of the dangerous gas can be more accurately judged, and the method has important significance for improving the life quality of people, guaranteeing the life health of staff and reducing environmental pollution and social burden.
Drawings
FIG. 1 is a block diagram of the overall design of a dual channel detection compass system with combined sensing of air flow and air related to the present invention;
FIG. 2 is a diagram of the pin distribution of an STM32F103C8T6 chip according to the present invention;
FIG. 3 is a diagram of the internal structure of an STM32F103C8T6 chip according to the present invention;
fig. 4 is an nRF52832BLESoc and minimum system module according to the present invention;
FIG. 5 is a block diagram of the hardware circuitry of the scent compass in accordance with the present invention;
fig. 6 is a structure of a BP network according to the present invention;
FIG. 7 is a schematic diagram of the quantitative analysis of hazardous gases according to the present invention;
FIG. 8 is a flow chart of the BP neural network algorithm according to the present invention;
FIG. 9 is a flowchart of SSA optimized BPNN prediction in accordance with the present invention;
FIG. 10 is a model of a BP network prediction constructed in accordance with the present invention;
FIG. 11 is a flow chart of a capture strategy for "find-track-locate" in accordance with the present invention;
FIG. 12 is a graph of a gas leakage source diffusion concentration profile according to the present invention;
FIG. 13 is a graph of a gas leakage source diffusion concentration profile according to the present invention;
FIG. 14 is a graph of a gas leakage source diffusion concentration profile according to the present invention;
FIG. 15 is a graph of a gas leakage source diffusion concentration profile according to the present invention;
fig. 16 is an assembled view of the dual channel detection compass according to the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments, based on the embodiments of the invention, which are apparent to those of ordinary skill in the art without inventive faculty, are intended to be within the scope of the invention.
Preferred embodiments of the invention:
referring to fig. 1-15, the present invention provides a dual channel detection compass system for sensing air flow and air combination, as shown in fig. 1 and 16, comprising: the device comprises a charging and discharging module, a data acquisition module, a data processing, transmitting and displaying module; wherein,,
the charge-discharge module provides power support for the whole system and provides proper voltage for the system through the step-down voltage stabilizing circuit; comprising the following steps: a through hole conductive slip ring 5, a stepping motor 6 and an angle correction rod 7 which are connected with the sensor bracket and the base circuit component;
the data acquisition module is used for storing acquired data in the SD in a file form through a FATFS file system and then sending the data through a serial port; comprising the following steps: a sensor bracket 1, a gas flow sensor 2, a wind direction sensing cabin 3 and a through hole conductive slip ring 4;
the data acquisition module 8 selects an STM32F103C8T6 microprocessor chip as the core of the singlechip module, adopts 3.3V voltage to supply power, has a maximum clock frequency of 72MHz, has 64KB FLASH and 20KB SRAM in the system, integrates 10 12-bit ADC analog-to-digital converters, USART, SPI, I C and other digital interfaces in the system, and supports serial port debugging and JTAC debugging modes;
STM32F103C8T6 chip pin distribution is shown in FIG. 2;
the data processing, transmitting and displaying module is used for receiving the data sent by the data acquisition module, performing digital filtering, temperature and humidity compensation and data preprocessing, then fusing the data of the plurality of sensors to obtain a final concentration value, and then transmitting and displaying the temperature and humidity, the residual electric quantity and the concentration value after data fusion in a BLE wireless mode;
the data processing, transmitting and displaying module adopts an nRF52832 microprocessor MCU; the nRF52832 microprocessor MCU is a BLE-based 32-bit mixed signal processor, is internally embedded with a core x-M4F, has a floating point operation unit, has strong floating point operation capability, can be internally provided with a more complex algorithm according to the requirement, supports multiple protocols of a low-power Bluetooth (BLE), ANT and 2.4GHz private protocol stacks, has abundant peripheral resources, and is specially designed for meeting the ultra-low power consumption (ULP) requirement BLE equipment. nRF52832BLESoc and minimum system modules are shown in fig. 4.
The data acquisition module comprises a top plate component, a gas concentration detection cabin array, a wind speed and direction detection area array and a bottom plate component, and a data line connecting the top plate component and the bottom plate component is arranged in a hollow cylindrical cavity in the center of the device; the top plate component mainly comprises a microprocessor chip, a compass sensor, a gas sensor and a PCB (printed circuit board), wherein data acquired by the compass sensor and the temperature and humidity sensor are transmitted to the microprocessor chip, and the wind speed and direction detection area array comprises 4 wind speed and direction detection areas; the gas concentration detection cabin array comprises 4 gas concentration detection cabins, and 4 gas sensors with the same model are respectively arranged in the centers of the 4 gas concentration detection cabins in an inverted mode;
The gas enrichment can effectively improve the detection capability of the monitoring system and increase the concentration of the gas to be detected, so that partial semicircular bulges are added to the internal structure of the detection area, the gas enrichment function is achieved, and the detection capability of the device is enhanced. The scent compass hardware block diagram is shown in fig. 5.
The sensor is a MEMS gas sensor for identifying the smell of harmful gas. The working principle is that under a certain temperature condition, when the change of the concentration of the gas to be detected causes the change of the resistance value in the sensitive resistor, the back-end circuit realizes the detection of the concentration of the gas through the processing of the resistance signal.
A gas leakage detection method comprising:
the method comprises the steps of opening a power switch of a data processing, transmitting and displaying module and a data acquisition module, connecting a BLE signal transmitted by a detection system on an APP arranged on the data processing, transmitting and displaying module, beginning to measure after pressing, continuing to measure for a period of time, enabling gas to be detected to enter a multi-sensor array, enabling the sensor array to output a certain voltage signal according to the dangerous gas content in the gas to be detected, enabling SHT30 to measure the temperature and humidity of the exhaled gas, transmitting a measured result to a data acquisition module STM32F103C8T6MCU, firstly storing acquired data in a SD in a file form through a FATFS file system, transmitting the data to nRF52832BLESoc through a serial port, preprocessing the nRF52832 after receiving the data, fusing the data of a plurality of sensors to obtain a final concentration value, and finally transmitting a temperature and humidity, a residual electric quantity and a concentration value after data fusion through a transmission function of a transmission service;
And the data processing, transmitting and displaying module receives the data transmitted by the system and displays the data on an interface of the data processing, transmitting and displaying module. And drawing the received concentration information into a graph, displaying the maximum concentration value, and judging the approximate direction of dangerous gas leakage to obtain a final detection result. Clicking for storage, and storing the measurement data in a file under an installation catalog, so that the real-time monitoring of workers and the remediation of maintenance workers are facilitated, and the physical health of workers in special environments can be ensured.
A method of data fusion and tracking, the method comprising:
the basic structure of the BP neural network consists of three layers, namely an input layer, a hidden layer and an output layer, wherein a plurality of hidden layers are possible, neurons of adjacent layers of the network are connected with each other, neurons in each layer are not connected with each other, and the output of the neuron of the previous layer is used as the input of the neuron of the next layer. The characteristics of the entire network are determined by the connectivity of the neurons between adjacent layers and the threshold on the neurons. The structure of the BP network is shown in fig. 6. The quantitative analysis and temperature and humidity compensation of dangerous gas are realized through a sensor array and a BP network algorithm, and the schematic diagram is shown in the following figure 7.
A gas quantitative analysis method based on a sensor array and a BP network. The defect of narrow measurement range of a single gas sensor can be overcome, the influence of accidental errors of the single sensor on the measurement accuracy of the system can be reduced by integrating multiple sensor signals, the temperature and humidity compensation of output signals can be performed, and the influence of temperature and humidity on the measurement result can be reduced.
Step one, designing a BP neural network algorithm;
firstly, initializing a network weight, and then inputting a training learning sample, wherein the training learning sample can be input into a network after normalization processing is carried out on the sample, and because the value ranges of all elements in the sample are possibly different and even have great differences, difficulties are brought to computer processing and operation, the network convergence speed of training is greatly slowed down, even training is unsuccessful, and the performance can be improved by preprocessing sample data.
Secondly, calculating the values of hidden layer and output layer neurons in the nerve network according to the set network weight and threshold value, and comparing the actual output value with a corresponding target value to obtain a network error; and (3) carrying out adjustment on the weight and the threshold according to the obtained error, and recalculating to minimize the error.
Thirdly, judging whether training is finished;
there are three situations that can lead to the end of training: the first time for training and learning is far longer than the set training time; the second is that the cycle number in the training and learning process is larger than the initially set step number; thirdly, the error generated by the system after the sample training is finished is smaller than the initially set error;
the network is not converged after the first condition and the second condition are finished, and the network is converged after the third condition is finished, so that the accuracy is higher;
and judging that the third condition is satisfied, namely that the training is finished, and storing the network.
The algorithm flow of the BP neural network is shown in the following figure 8.
Step two, designing the structure of the BP neural network; the method comprises the steps of setting network layers, input layer node numbers, hidden layer node numbers, output layer node numbers, transmission functions, training methods and training parameters;
step three, introducing a sparrow search algorithm (SparrowSearchAlgorithm, SSA) to optimize the initial weight and the threshold value of the BP neural network, so that the prediction accuracy of the model is improved, and the purpose of optimizing the BP neural network is achieved;
the BP network has strong local optimizing capability, but is easily influenced by initial weight to enable the model to sink into local optimizing, so that the model has poor prediction precision and stability.
The sparrow search algorithm mainly simulates the process of foraging of sparrows. During the process of sparrow foraging, the process is divided into discoverers and additioners, wherein the discoverers are responsible for searching food in the population and providing foraging areas and directions for the whole sparrow population, and the additioners use the discoverers to acquire the food. To obtain food, sparrows may generally be foraged using both discoverers and joiners. Individuals in a population will monitor the behavior of other individuals in the population and an attacker in the population will compete with high-intake peers for food resources to increase their predation rate. In addition, anti-predation behavior is made when sparrow populations are attacked by predators.
And selecting the mean square error of the training set and the whole test set as the fitness value. The smaller the fitness function value is, the more accurate the training is, and the prediction accuracy of the model is better. The SSA optimized BPNN prediction flow chart is shown in fig. 9 below.
Designing a method for searching and positioning the leakage source position;
the leakage source capture strategy is designed to discover, track and locate three phases:
finding out identification and characterization of gas diffusion information in an immediate variable flow field environment; tracking prediction, perception and search of a gas cloud cluster diffusion track in an immediate variable flow field environment; positioning confirmation of a gas leakage source in an immediate variable flow field environment;
Based on a Gaussian model, a gas leakage diffusion concentration gradient distribution map is obtained, so that a gas leakage accident poison area is divided, and a path backtracking and tracking algorithm is constructed; based on a Bayesian decision theory, a positioning algorithm of a gas leakage source is obtained, and an optimal solution of the position information of the gas leakage source is obtained;
by constructing a capture strategy of discovery-tracking-positioning, the key problems of rapid gas leakage source capture route design, a gas diffusion concentration model prediction method, a route backtracking and tracking algorithm, a gas leakage source position information confirmation mechanism and the like are solved. The capture strategy design for "find-track-locate" is shown in fig. 10.
Fifthly, constructing a gas leakage diffusion concentration distribution map;
step six, determining the type and concentration of the leaked gas at the same time;
the output values of different sensors are collected in the actual monitoring process, the change rate of the output values is calculated, and the output values are compared with a pre-stored characteristic response database in real time, so that the rapid identification and concentration prediction of the target gas type can be realized; the method comprises the following specific steps: (1) setting the output values of the 3 sensors as S1 (t), S2 (t) and S3 (t), and setting the change rates of the output values as R1 (t) and R2 (t);
(2) Substituting Si (t) and Ri (t) into cij=fij [ Si (t), ri (t) ], and calculating to obtain Cij, wherein i=1, 2, 3, j=1, 2, 3; the relation Cjk between the gas concentration Cjk and the sensor output value and the output value change rate is that sijk=fij [ Eijk (t), fijk (t) ], wherein Sijk (t) is the sensor output value, rijk (t) is the sensor output value change rate, cjk is the target gas concentration, i=1, 2, 3 is the sensor number, j=1, 2, 3 is the target gas type number, and k=1, 2, 3, 4 is the target gas concentration number;
(3) analyzing the value of Cij, comparing the calculated concentration C1j, C2j and C3j of different gases by the same sensor in groups, and calculating the concentration error of the target gasTaking the minimum value of Dj as DJ, judging that DJ is smaller than phi, if the minimum value of DJ is satisfied, J is the number of the currently detected target gas, and the concentration of the target gas is
Step two, designing the structure of the BP neural network; the method comprises the steps of setting network layers, input layer node numbers, hidden layer node numbers, output layer node numbers, transmission functions, training methods and training parameters, and specifically comprises the following steps:
a) Network layer number
The BP network may contain one to more hidden layers; the network of the single hidden layer can realize any nonlinear mapping by properly increasing the number of the neuron nodes; thus, for most applications, a single hidden layer may suffice; when the number of samples is large, adding an hidden layer can obviously reduce the network scale;
b) Input layer node number
The number of input layer nodes depends on the dimension of the input vector; the node number of the input layer is determined according to the number of sensors in the sensor array; thus the number of neurons in the input layer is 8;
c) Hidden layer node number
The number of hidden layer nodes has great influence on the performance of the BP network; generally, a larger hidden layer node number can bring better performance, but the training time is excessively long; obtaining an estimated value of the number of hidden layer nodes:
wherein k is the number of samples, m is the number of neurons of the hidden layer, and n is the number of neurons of the input layer; if i > m, provision is made for
Where m and n are the number of neurons in the output layer and input layer, respectively, and a is a constant between [0, 10 ].
M＝log 2 n (3)
Wherein n is the number of neurons of the input layer;
through the three calculation formulas, the range of the number of neurons of the hidden layer can be preliminarily determined to be 3-13, and then the optimal value is determined by continuously adjusting the accuracy requirement to be achieved by training and the self-property of the selection function; or the number of neurons in the hidden layer is changed to be adjustable, or the number of neurons is ensured to be enough when the number of neurons is initially set, and redundant neurons are removed after continuous learning until the optimal value is obtained;
d) Number of neurons in output layer
The number of the neurons of the output layer is determined according to an abstract model obtained in the actual problem;
e) Selection of transfer functions
The implicit layer transfer function is 'log sig', and the output layer transfer function is 'purlin'; the value range of x in the function curve is between 0 and 1, and the input data is preprocessed to be converted into the range of 0 and 1 interval;
f) Selection of training method
Default "tranlm" function is training function in BP network, there is possibility of sinking into local minimum point when the network is trained, thus reduce the modeling accuracy of the system;
using the ReLU as an activation function, outputting 0 when the input of the ReLU is smaller than 0, and linearly increasing when the input of the ReLU is larger than 0; therefore, the ReLU function cannot have gradient vanishing condition in back propagation, and the learning effect is better; meanwhile, the ReLU function is simple to calculate, exponential operation is not involved, and training time of a network can be shortened to a certain extent;
g) Determination of initial weights
The BP network adopts an iterative updating mode to determine the weight, firstly, the initial weight is defined as a smaller non-zero random value, and the empirical value is (-2.4/F, 2.4/F) or (-3/F, 3/F), wherein F is the number of neurons connected with the weight input end;
h) Data preprocessing
The data preprocessing comprises normalization of gas sensor signals and normalization of gas concentration signals;
the gas sensor signal normalization algorithm is as follows:
where i=1, 2,3,4,5,6max [ i ] represents the maximum response value of the sensor i to all test samples;
the normalization algorithm of the gas concentration signal is as follows:
c′ j ＝c j /c max (5)
wherein, c max Maximum expected result for test set samples 100, j=1, 2, …,20
After the input samples of the test set are subjected to sensor signal normalization processing, each element of the input vector of the array is in the same order of magnitude, so that the input samples can be used as proper input data of a neural network, and calculation errors in chemometry recognition can be reduced; the gas concentration normalization can meet the requirement of the output amplitude of the neuron S-shaped excitation function, and when the network predicts, the output of the network obtains a gas predicted concentration value through the inverse transformation of a gas concentration signal normalization algorithm;
after the parameters are determined, carrying out data preprocessing on training data, inputting the training data into a network for learning, and obtaining a required neural network if the network is successfully converged; the constructed BP network prediction model is shown in the following figure 11.
In the fourth step, a process of searching and positioning the leakage source position is designed, specifically:
Selecting a grid set with the highest probability from a possible source grid set of the gas diffusion plumes through reverse estimation as a source path of the gas diffusion plumes; assuming that a certain gas diffuses smoke plume at t 1 At the moment at the position L 1 At t 2 Time of arrival at position L 2 ，t 1＜ t 2 The flow process of the gas diffusion plume can be described as:
where U (L (t)) is the average flow velocity vector of the gas diffusion plume at the position L (t) at the time t, expressed as
[u x (L(t)),u y (L(t)),u z (L(t))] T N (t) is Gaussian noise; if the mobile robot is at t j At the moment at the position L R (t j ) Detecting the gas diffusion smoke plume at t j Some t before the moment l The position at the moment can be written as formula (7):
when the gas sensor detects the gas concentration, firstly, a suspected path of the gas diffusion smoke plume is obtained through reverse pushing, then, all possible gas source grid sets are searched nearby the suspected path according to a historical gas diffusion distribution likelihood diagram and historical wind speed and wind direction information, and finally, the grid set with the highest probability is selected as the source path of the gas diffusion smoke plume; fig. 10 is a capture strategy design for "find-track-locate".
The minimum risk Bayesian decision is an optimal decision taking into account different losses caused by different errors; for the suspected gas leakage source position to be determined, there may be two states: the presence and absence of a gas leakage source; there are two decisions on this: the presence of a gas leakage source and the absence of a gas leakage source are determined to have four combinations, which produce three decision results in total: correct (risk factor zero), false alarm, missed alarm. The loss caused by false alarm and missed alarm is different, so that different risk coefficients are given to the false alarm and missed alarm (the false alarm, namely the smell Luo Pancuo, judges that the position without the gas leakage source is the gas leakage source, the gas leakage source positioning process ends in failure, the loss is large, namely the risk coefficient is large, the missed alarm, namely the smell compass judges that the position with the gas leakage source is not the gas leakage source, and although the false decision is also made, the gas leakage source positioning is not finished, the mobile robot is carried with the smell compass device to perform searching judgment again, so that the loss is small, namely the risk coefficient is small). And calculating all risks possibly brought by different decisions as conditional risks of each decision, and selecting the decision with smaller conditional risks as a result of gas leakage source position judgment.
In the fifth step, the process of constructing the gas leakage diffusion concentration distribution map specifically comprises the following steps:
expanding the discrete and sparse smell information into a gas leakage source diffusion concentration distribution likelihood map (gas existence probability distribution) of the whole search space by adopting a Gaussian kernel function; dividing the three-dimensional space into three-dimensional grids with uniform density according to a proper scale, taking the three-dimensional space as a minimum space unit constructed by a gas distribution likelihood map, and giving each grid an accumulated weight; setting the cumulative weight of each grid to zero before the gas diffusion concentration is captured; when the gas diffusion concentration is captured, calculating the current weight of the grids in a certain range around by adopting a Gaussian kernel function, and accumulating the current weight of the grids; the independent variable of the Gaussian kernel function is the distance between a certain grid and a gas detection position, and the dependent variable is the current weight of the grid; the current weight of the grid represents the contribution of the grid to the current gas detection result, and the contribution of the grid which is farther from the gas detection concentration position is smaller; normalizing the accumulated weight of all grids to obtain probability distribution of gas concentration diffusion, namely a gas leakage source diffusion concentration distribution likelihood map, and providing the probability distribution likelihood map as an initial condition for a gas distribution likelihood map updating process in subsequent gas diffusion smoke plume source path backtracking and tracing;
And (3) simulating and constructing a gas leakage source diffusion concentration distribution map in an outdoor real gas flow environment, as shown in figures 12-15.
FIG. 12 is a graph illustrating the detection result of scanning in a two-dimensional plane in space; FIG. 13 is a graph result of scanning in three dimensions; FIG. 14 is a cross-sectional view of a plane in which the diffusion concentration of a gas leakage source lies; FIG. 15 is a three-dimensional display of a gas leakage source diffusion concentration profile construction. By means of simulation analysis, a theoretical basis can be provided for the development of related research works of the subject.
The embodiments of the present invention are disclosed as preferred embodiments, but not limited thereto, and those skilled in the art will readily appreciate from the foregoing description that various extensions and modifications can be made without departing from the spirit of the present invention.
Claims (7)
1. A dual-channel detection compass system with air flow and air combination sensing is characterized in that: the composition of the composite material comprises: the device comprises a charging and discharging module, a data acquisition module, a data processing, transmitting and displaying module; wherein,,
the charge-discharge module provides power support for the whole system and provides proper voltage for the system through the step-down voltage stabilizing circuit; comprising the following steps: a through hole conductive slip ring (5), a stepping motor (6) and an angle correction rod (7) which are connected with the sensor bracket and the base circuit component;
The data acquisition module is used for storing acquired data in the SD in a file form through a FATFS file system and then sending the data through a serial port; comprising the following steps: the device comprises a sensor bracket (1), a gas flow sensor (2), a wind direction sensing cabin (3) and a through hole conductive slip ring (4);
the data acquisition module (8) selects an STM32F103C8T6 microprocessor chip as the core of the singlechip module, is powered by 3.3V voltage, has a maximum clock frequency of 72MHz, has 64KB FLASH and 20KB SRAM in the system, integrates 10 12-bit ADC analog-to-digital converters, USART, SPI, I C and other digital interfaces in the system, and supports serial port debugging and JTAC debugging modes;
the data processing, transmitting and displaying module is used for receiving the data sent by the data acquisition module, performing digital filtering, temperature and humidity compensation and data preprocessing, then fusing the data of the plurality of sensors to obtain a final concentration value, and then transmitting and displaying the temperature and humidity, the residual electric quantity and the concentration value after data fusion in a BLE wireless mode;
the data processing, transmitting and displaying module adopts an nRF52832 microprocessor MCU.
2. The dual channel sensing compass system of claim 1, wherein: the data acquisition module comprises a top plate component, a gas concentration detection cabin array, a wind speed and direction detection area array and a bottom plate component, and a data line connecting the top plate component and the bottom plate component is arranged in a hollow cylindrical cavity in the center of the device; the top plate component mainly comprises a microprocessor chip, a compass sensor, a gas sensor and a PCB (printed circuit board), wherein data acquired by the compass sensor and the temperature and humidity sensor are transmitted to the microprocessor chip, and the wind speed and direction detection area array comprises 4 wind speed and direction detection areas; the gas concentration detection cabin array comprises 4 gas concentration detection cabins, and 4 gas sensors with the same model are respectively arranged in the centers of the 4 gas concentration detection cabins in an inverted mode;
The sensor is an MEMS gas sensor for identifying the smell of harmful gas, and when the change of the concentration of the gas to be detected causes the change of the resistance value in the sensitive resistor, the back-end circuit realizes the detection of the concentration of the gas by processing the resistance signal.
3. A gas leakage detection method using the system according to any one of claims 1 or 2, characterized in that:
the method comprises the steps of opening a power switch of a data processing, transmitting and displaying module and a data acquisition module, connecting a BLE signal transmitted by a detection system on an APP arranged on the data processing, transmitting and displaying module, beginning to measure by pressing, enabling gas to be detected to enter a multi-sensor array, enabling the sensor array to output a certain voltage signal according to dangerous gas content in the gas to be detected, enabling SHT30 to measure temperature and humidity of exhaled gas, transmitting a measured result to the data acquisition module STM32F103C8T6MCU, firstly storing acquired data in a SD in a file form through a FATFS file system, then transmitting the data to nRF52832 BLSoc through a serial port, carrying out digital filtering and data preprocessing on the data after the nRF52832 receives the data, then fusing the data of a plurality of sensors to obtain a final concentration value, and finally transmitting a temperature and humidity, residual electric quantity and a concentration value after data fusion through a transmission function of a transmission service;
And the data processing, transmitting and displaying module receives the data transmitted by the system and displays the data on an interface of the data processing, transmitting and displaying module.
4. A data fusion and tracking method suitable for use in one of claims 1-3, characterized in that: the method comprises the following steps:
the basic structure of the BP neural network consists of three layers, namely an input layer, a hidden layer and an output layer, wherein a plurality of hidden layers are possible, neurons of adjacent layers of the network are connected with each other, neurons in each layer are not connected with each other, and the output of the neuron of the previous layer is used as the input of the neuron of the next layer; the characteristics of the whole network are determined by the connection weight of adjacent interlayer neurons and the threshold value on the neurons; the quantitative analysis and the temperature and humidity compensation of dangerous gas are realized through the sensor array and the BP network algorithm,
step one, designing a BP neural network algorithm;
firstly, initializing a network weight, and then inputting a training learning sample;
secondly, calculating the values of hidden layer and output layer neurons in the nerve network according to the set network weight and threshold value, and comparing the actual output value with a corresponding target value to obtain a network error;
Thirdly, judging whether training is finished;
there are three situations that can lead to the end of training: the first time for training and learning is far longer than the set training time; the second is that the cycle number in the training and learning process is larger than the initially set step number; thirdly, the error generated by the system after the sample training is finished is smaller than the initially set error;
the network is not converged after the first and second conditions are over, and the network is converged after the third condition is over;
judging that the training is finished when the third condition is met, and storing the network;
step two, designing the structure of the BP neural network; the method comprises the steps of setting network layers, input layer node numbers, hidden layer node numbers, output layer node numbers, transmission functions, training methods and training parameters;
step three, introducing a sparrow search algorithm to optimize the initial weight and the threshold value of the BP neural network;
designing a method for searching and positioning the leakage source position;
based on a Gaussian model, a gas leakage diffusion concentration gradient distribution map is obtained, so that a gas leakage accident poison area is divided, and a path backtracking and tracking algorithm is constructed; based on a Bayesian decision theory, a positioning algorithm of a gas leakage source is obtained, and an optimal solution of the position information of the gas leakage source is obtained;
Fifthly, constructing a gas leakage diffusion concentration distribution map;
step six, determining the type and concentration of the leaked gas at the same time;
the output values of different sensors are collected in the actual monitoring process, the change rate of the output values is calculated, and the output values are compared with a pre-stored characteristic response database in real time, so that the rapid identification and concentration prediction of the target gas type can be realized; the method comprises the following specific steps:
(1) setting the output values of the 3 sensors as S1 (t), S2 (t) and S3 (t), and setting the change rates of the output values as R1 (t) and R2 (t);
(2) substituting Si (t) and Ri (t) into cij=fij [ Si (t), ri (t) ], and calculating to obtain Cij, wherein i=1, 2, 3, j=1, 2, 3; the relation Cjk between the gas concentration Cjk and the sensor output value and the output value change rate is that sijk=fij [ Eijk (t), fijk (t) ], wherein Sijk (t) is the sensor output value, rijk (t) is the sensor output value change rate, cjk is the target gas concentration, i=1, 2, 3 is the sensor number, j=1, 2, 3 is the target gas type number, and k=1, 2, 3, 4 is the target gas concentration number;
(3) analyzing the value of Cij, comparing the calculated concentration C1j, C2j and C3j of different gases by the same sensor in groups, and calculating the concentration error of the target gas j=1, 2 and 3, taking the minimum value of Dj as DJ, judging that DJ is smaller than phi, if the minimum value of DJ is satisfied, J is the number of the currently detected target gas, and the concentration of the target gas is +.>
5. The method for data fusion and tracking according to claim 4, wherein: step two, designing the structure of the BP neural network; the method comprises the steps of setting network layers, input layer node numbers, hidden layer node numbers, output layer node numbers, transmission functions, training methods and training parameters, and specifically comprises the following steps:
a) Network layer number
The BP network may contain one to more hidden layers;
b) Input layer node number
The number of input layer nodes depends on the dimension of the input vector; the node number of the input layer is determined according to the number of sensors in the sensor array;
c) Hidden layer node number
Obtaining an estimated value of the number of hidden layer nodes:
wherein k is the number of samples, m is the number of neurons of the hidden layer, and n is the number of neurons of the input layer; if i > m, provision is made for
Wherein m and n are the number of neurons of the output layer and the input layer, respectively, and a is a constant between [0, 10 ];
M＝log 2 n(3)
wherein n is the number of neurons of the input layer;
through the three calculation formulas, the range of the number of neurons of the hidden layer can be preliminarily determined to be 3-13, and then the optimal value is determined by continuously adjusting the accuracy requirement to be achieved by training and the self-property of the selection function; or the number of neurons in the hidden layer is changed to be adjustable, or the number of neurons is ensured to be enough when the number of neurons is initially set, and redundant neurons are removed after continuous learning until the optimal value is obtained;
d) Number of neurons in output layer
The number of the neurons of the output layer is determined according to an abstract model obtained in the actual problem;
e) Selection of transfer functions
The implicit layer transfer function is 'log sig', and the output layer transfer function is 'purlin'; the value range of x in the function curve is between 0 and 1, and the input data is preprocessed to be converted into the range of 0 and 1 interval;
f) Selection of training method
A tranlm function is selected as a training function in the BP network, and the probability of sinking into local minimum points exists in the network during training, so that the modeling accuracy of the system is reduced;
using the ReLU as an activation function, outputting 0 when the input of the ReLU is smaller than 0, and linearly increasing when the input of the ReLU is larger than 0;
g) Determination of initial weights
The BP network adopts an iterative updating mode to determine the weight, firstly, the initial weight is defined as a smaller non-zero random value, and the empirical value is (-2.4/F, 2.4/F) or (-3/F, 3/F), wherein F is the number of neurons connected with the weight input end;
h) Data preprocessing
The data preprocessing comprises normalization of gas sensor signals and normalization of gas concentration signals;
the gas sensor signal normalization algorithm is as follows:
where i=1, 2,3,4,5,6max [ i ] represents the maximum response value of the sensor i to all test samples;
The normalization algorithm of the gas concentration signal is as follows:
c′ j ＝c j /c max (5)
wherein, c max Maximum expected result for test set samples 100, j=1, 2, …,20
After the input samples of the test set are subjected to sensor signal normalization processing, each element of the input vector of the array is in the same order of magnitude, the gas concentration normalization can meet the requirement of the output amplitude of the neuron S-shaped excitation function, and when the network predicts, the output of the network obtains a gas prediction concentration value through the inverse transformation of a gas concentration signal normalization algorithm;
after the parameters are determined, the training data is subjected to data preprocessing and is input into a network for learning, and if the network is successfully converged, the required neural network can be obtained.
6. The method for data fusion and tracking according to claim 5, wherein: in the fourth step, a process of searching and positioning the leakage source position is designed, specifically:
by reversely estimating the grid set of possible sources of the gas diffusion smoke plume, selecting the grid set with the highest probability as the gas diffusionA source path of the plume; assuming that a certain gas diffuses smoke plume at t 1 At the moment at the position L 1 At t 2 Time of arrival at position L 2 ，t 1＜ t 2 The flow process of the gas diffusion plume can be described as:
Where U (L (t)) is the average flow velocity vector of the gas diffusion plume at the position L (t) at time t, denoted as [ U ] x (L(t)),u y (L(t)),u z (L(t))] T N (t) is Gaussian noise; if the mobile robot is at t j At the moment at the position L R (t j ) Detecting the gas diffusion smoke plume at t j Some t before the moment l The position at the moment can be written as formula (7):
when the gas sensor detects the gas concentration, firstly, a suspected path of the gas diffusion smoke plume is obtained through reverse pushing, then, all possible gas source grid sets are searched nearby the suspected path according to a historical gas diffusion distribution likelihood diagram and historical wind speed and wind direction information, and finally, the grid set with the highest probability is selected as the source path of the gas diffusion smoke plume; and calculating all risks possibly brought by different decisions as conditional risks of each decision, and selecting the decision with smaller conditional risks as a result of gas leakage source position judgment.
7. The method for data fusion and tracking according to claim 6, wherein: in the fifth step, the process of constructing the gas leakage diffusion concentration distribution map specifically comprises the following steps:
expanding the discrete and sparse smell information into a gas leakage source diffusion concentration distribution likelihood map of the whole search space by adopting a Gaussian kernel function; dividing the three-dimensional space into three-dimensional grids with uniform density according to a proper scale, taking the three-dimensional space as a minimum space unit constructed by a gas distribution likelihood map, and giving each grid an accumulated weight; setting the cumulative weight of each grid to zero before the gas diffusion concentration is captured; when the gas diffusion concentration is captured, calculating the current weight of the grids in a certain range around by adopting a Gaussian kernel function, and accumulating the current weight of the grids; the independent variable of the Gaussian kernel function is the distance between a certain grid and a gas detection position, and the dependent variable is the current weight of the grid; and normalizing the accumulated weight of all grids to obtain the probability distribution of gas concentration diffusion, namely a gas leakage source diffusion concentration distribution likelihood map, and providing the probability distribution likelihood map as an initial condition for a gas distribution likelihood map updating process in subsequent gas diffusion plume source path backtracking and tracing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310596318.XA CN116626238A (en) | 2023-05-25 | 2023-05-25 | Dual-channel detection compass system for sensing air flow and air combination, air leakage detection method, data fusion and tracking method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310596318.XA CN116626238A (en) | 2023-05-25 | 2023-05-25 | Dual-channel detection compass system for sensing air flow and air combination, air leakage detection method, data fusion and tracking method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116626238A true CN116626238A (en) | 2023-08-22 |
Family
ID=87620864
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310596318.XA Pending CN116626238A (en) | 2023-05-25 | 2023-05-25 | Dual-channel detection compass system for sensing air flow and air combination, air leakage detection method, data fusion and tracking method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116626238A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117420180A (en) * | 2023-10-20 | 2024-01-19 | 中国矿业大学 | VOC (volatile organic compound) detection early warning system with decision |
-
2023
- 2023-05-25 CN CN202310596318.XA patent/CN116626238A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117420180A (en) * | 2023-10-20 | 2024-01-19 | 中国矿业大学 | VOC (volatile organic compound) detection early warning system with decision |
| CN117420180B (en) * | 2023-10-20 | 2024-03-19 | 中国矿业大学 | VOC (volatile organic compound) detection early warning system with decision |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108989976B (en) | Fingerprint positioning method and system in intelligent classroom | |
| CN105206057B (en) | Detection method and system based on Floating Car resident trip hot spot region | |
| CN106289691B (en) | A kind of bridge block impact vibration detection method and detection device based on microwave radar device | |
| CN110346517B (en) | Smart city industrial atmosphere pollution visual early warning method and system | |
| CN110320495A (en) | A kind of indoor orientation method based on Wi-Fi, bluetooth and PDR fusion positioning | |
| CN109146144A (en) | Based on the landslide disaster prediction technique for improving Bayesian network | |
| CN110398320A (en) | A kind of gas leakage detection localization method and system being easy to Continuous optimization | |
| CN110046834A (en) | Workplace pernicious gas health risk quantitative evaluating system and its appraisal procedure | |
| CN110401978B (en) | Indoor positioning method based on neural network and particle filter multi-source fusion | |
| CN106991491B (en) | RBF neural network's environmental pollution monitoring system based on genetic algorithm optimization | |
| CN108828643B (en) | Indoor and outdoor seamless positioning system and method based on grey prediction model | |
| CN116626238A (en) | Dual-channel detection compass system for sensing air flow and air combination, air leakage detection method, data fusion and tracking method | |
| AU2020102518A4 (en) | A method of gas sensor calibration based on linear optimization | |
| CN108614071A (en) | Distributed outside atmosphere quality-monitoring accuracy correction system and parameter updating method | |
| CN116308958A (en) | Carbon emission online detection and early warning system and method based on mobile terminal | |
| CN103415027B (en) | WIFI indoor signal distribution model automatically selects and localization method | |
| CN113155196A (en) | Bridge operation real-time monitoring system based on AIoT and monitoring method thereof | |
| CN108562698A (en) | A kind of big data air quality indexes information processing system | |
| CN113936413A (en) | Early fire monitoring and early warning method and device | |
| CN110401977A (en) | A kind of more floor indoor orientation methods returning more Classification and Identification devices based on Softmax | |
| CN112560981A (en) | Training method, apparatus, device, program and storage medium for generating countermeasure model | |
| CN110503348B (en) | Individual air pollution exposure simulation measurement method based on position matching | |
| CN116665001A (en) | Intelligent park multisource data dynamic monitoring and real-time analysis system and method | |
| CN115342814A (en) | Unmanned ship positioning method based on multi-sensor data fusion | |
| CN112085148A (en) | Closed container gas leakage detection method based on ant colony algorithm |
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 |