CN117907174A - Dust concentration detection system and method based on fusion of light absorption and charge induction method - Google Patents
Dust concentration detection system and method based on fusion of light absorption and charge induction method Download PDFInfo
- Publication number
- CN117907174A CN117907174A CN202410033474.XA CN202410033474A CN117907174A CN 117907174 A CN117907174 A CN 117907174A CN 202410033474 A CN202410033474 A CN 202410033474A CN 117907174 A CN117907174 A CN 117907174A
- Authority
- CN
- China
- Prior art keywords
- concentration
- current
- light intensity
- dust
- output
- 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
- 239000000428 dust Substances 0.000 title claims abstract description 115
- 230000031700 light absorption Effects 0.000 title claims abstract description 43
- 230000004927 fusion Effects 0.000 title claims abstract description 41
- 230000006698 induction Effects 0.000 title claims description 55
- 238000001514 detection method Methods 0.000 title claims description 35
- 238000000034 method Methods 0.000 title claims description 23
- 238000012545 processing Methods 0.000 claims abstract description 28
- 230000007613 environmental effect Effects 0.000 claims abstract description 19
- 230000006870 function Effects 0.000 claims description 20
- 108010076504 Protein Sorting Signals Proteins 0.000 claims description 14
- 239000011159 matrix material Substances 0.000 claims description 14
- 230000002159 abnormal effect Effects 0.000 claims description 12
- 239000013307 optical fiber Substances 0.000 claims description 10
- 239000000523 sample Substances 0.000 claims description 10
- 238000010606 normalization Methods 0.000 claims description 9
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000005516 engineering process Methods 0.000 claims description 5
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 238000012937 correction Methods 0.000 claims description 4
- 230000003213 activating effect Effects 0.000 claims description 3
- 230000004913 activation Effects 0.000 claims description 3
- 230000002238 attenuated effect Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000002474 experimental method Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 239000012530 fluid Substances 0.000 claims description 2
- 238000012544 monitoring process Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 2
- 239000012528 membrane Substances 0.000 description 2
- 238000005303 weighing Methods 0.000 description 2
- 239000003245 coal Substances 0.000 description 1
- 239000013618 particulate matter Substances 0.000 description 1
- 230000029058 respiratory gaseous exchange Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000009423 ventilation Methods 0.000 description 1
Landscapes
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The environment sensing system comprises an air inlet pipeline, a temperature and humidity sensor, a light absorption subunit, a charge sensing subunit and a split body, wherein the temperature and humidity sensor, the light absorption subunit and the charge sensing subunit are integrated in the air inlet pipeline in a serial mode, the temperature and humidity sensor is arranged at the front end of the air inlet pipeline, and a split body is arranged between the light absorption subunit and the charge sensing subunit; the data processing system consists of a singlechip and an external circuit, and is internally provided with a humidity-error model, a concentration-current model, a concentration-light intensity model and an information fusion model, so as to correct and fuse the data detected by the environment sensing system; the alarm display system is connected with the data processing system, receives signal information fed back by the data processing system, and outputs a corresponding alarm mode; the invention can monitor and early warn environmental dust with high and low concentration at the same time.
Description
Technical Field
The invention relates to a dust concentration detection system, in particular to a dust concentration detection system and method based on fusion of light absorption and a charge induction method, which belong to the technical field of air particulate matter concentration monitoring and early warning and are suitable for high-concentration and low-concentration environment dust concentration monitoring and early warning.
Background
In recent years, with the importance of national occupational health of workers, various large coal mines, machinery, building industry and the like have attracted attention for dust generation in working sites. Dust detection equipment in various places is also continuously applied, however, the existing detection technology and equipment are often difficult to detect environmental dust with high and low concentration. The detection accuracy can be improved by sampling and detecting the field operation environment through a filter membrane weighing method, but the detection result is not timeliness. In a word, the existing method cannot have the characteristics of real-time monitoring and high-precision detection.
Disclosure of Invention
The invention aims to provide a dust concentration detection system and method based on fusion of light absorption and charge induction, which can monitor and early warn high-concentration and low-concentration environmental dust at the same time.
In order to achieve the above purpose, the invention provides a dust concentration detection system based on fusion of light absorption and charge induction method, comprising a power supply system, an environment induction system, a data processing system and an alarm display system, wherein the power supply system supplies power to the environment induction system, the environment induction system comprises an air inlet pipeline, a temperature and humidity sensor, a light absorption subunit, a charge induction subunit and a split body, the temperature and humidity sensor, the light absorption subunit and the charge induction subunit are integrated in the air inlet pipeline in a serial connection mode, the temperature and humidity sensor is arranged at the front end of the air inlet pipeline, and a split fluid is arranged between the light absorption subunit and the charge induction subunit;
The data processing system consists of a singlechip and an external circuit, a humidity-error model, a concentration-current model, a concentration-light intensity model and an information fusion model are built in the data processing system, data detected by the environment sensing system are corrected and fused, and the data processing system is connected with the alarm display system;
The alarm display system is connected with the data processing system, receives signal information fed back by the data processing system and outputs a corresponding alarm mode.
The light absorption subunit comprises a laser probe, a convex lens and an optical fiber spectrometer, is arranged behind the temperature and humidity sensor, has good applicability in a low-concentration dust environment, and has a negative correlation between the light intensity sensed by the optical fiber spectrometer and the dust concentration of the environment; the charge induction subunit comprises an induction electrode and a current amplifier, is arranged at the rear end of the air inlet pipeline, has good applicability in a high-concentration dust environment, and has positive correlation between the internal current and the environment dust concentration; the flow splitting body consists of a triangular splitting column and a deceleration column arranged on an extension line of the splitting column, dust flow passing through the light absorption subunit is split to two ends of a pipeline by the triangular splitting column, is decelerated and split again after flowing to the deceleration column, and the split inner dust flow is in a tiny karman vortex street phenomenon and is in full contact with an electrode, so that the detection precision can be improved.
A dust concentration detection method based on fusion of light absorption and charge induction method comprises the following steps:
S1: starting equipment, and receiving data information from an environment sensing system by a data processing system, wherein the data information comprises temperature T, humidity H, current I and light intensity L, and specifically comprises the following steps:
The dust flow enters an air inlet pipeline, and the temperature T and the humidity H of the dust flow are detected by a temperature and humidity sensor; scattered light emitted by the laser probe becomes parallel light after passing through the convex lens, and the light intensity of the parallel light is attenuated after passing through dust flow and finally sensed by the optical fiber spectrometer, namely light intensity L; the dust generates induction current after contacting with the induction electrode, but the induction current is smaller, and the current I is obtained by amplifying the dust by using an amplifier;
S2: preprocessing the data information received in the step S1 to obtain a processed signal sequence of humidity H 0, temperature T 0, current I 0 and light intensity L 0;
S3: correcting the preprocessed current I 0 and the light intensity L 0 signal sequence by using a humidity-error model to obtain corrected current and light intensity data;
S4: the corrected current and light intensity data are respectively input into a concentration-current model and a concentration-light intensity model, a certain weight is respectively given to the output of the concentration-current model and the output of the concentration-light intensity model according to the situation, the environmental dust concentration is obtained by adding, and the environmental dust concentration data is displayed on an alarm display system;
S5: performing feature fusion on the output of the temperature, concentration-current model and the output of the concentration-light intensity model by using a GRU network;
S6: classifying the feature fusion result by using a Softmax function to obtain a judging result of whether the concentration of environmental dust exceeds a limit value;
s7: if the current environment dust concentration exceeds the limit or the environment dust concentration is about to exceed the limit, the alarm display system is made to alarm and display related information.
The step S2 of preprocessing the data information includes the steps of:
S21: soft threshold denoising is carried out on the received signal sequences of the temperature T, the humidity H, the current I and the light intensity L by using a wavelet basis function of a sym wavelet system;
s22: detecting abnormal values in the data by using a local abnormal factor detection technology, and filling the abnormal values by using an interpolation method;
S23: and carrying out normalization processing on the modified data, wherein the normalization relation is as follows:
Wherein: Mapping at/>, after normalization Results in;
x is the original data;
x min and x max are the minimum and maximum values of the raw data, respectively;
S24: the processed signals of humidity H 0, temperature T 0, current I 0 and light intensity L 0 are obtained.
The step S3 carries out error correction on the preprocessed data by adopting a BP neural network, and specifically comprises the following steps:
S31: network initialization takes a signal sequence H 0 consisting of humidity data as input; error matrix Err of current and light intensity as output
Err＝[err1,err2]T
Wherein: err 1 is the current error sequence;
err 2 is the light intensity error sequence;
So the input layer node l is 1, and the output layer node m is 2; the number of hidden layer nodes n is:
Preliminarily taking a as 3 and hiding layer node
S32: calculating an output H i of the hidden layer node:
Hi＝f(wi-ai)1≤i≤5
Wherein: Activating a function for Sigmoid;
w i is the connection weight between the input layer and the hidden layer;
a i is a threshold in the hidden layer;
i is the node number of the hidden layer;
S33: the output of the output layer node is calculated, namely error err k:
Wherein: w ik is the weight between the hidden layer and the output layer;
b k is the threshold of the hidden layer;
S34: calculating a corrected current sequence I r and a corrected light intensity sequence L r
Ir＝I0+err1
Lr＝L0+err2。
Step S4 comprises the steps of:
S41: respectively inputting the corrected current sequence I r and the corrected light intensity sequence L r into a concentration-current model and a concentration-light intensity model;
S42: respectively outputting a concentration value V i and a concentration value V l predicted by a concentration-current model and a concentration-light intensity model;
S43: if V i＞Vh and V l＞Vh, the output V i of the concentration-current model is given weight W i1, the output V l of the concentration-light intensity model is given weight W l1, and the dust concentration is calculated
Vr＝Vi*wi1+Vl*wl1
Wherein: v h is the concentration high-low judgment value determined by experiments, and W i1＞Wl1;
S44: if V i＜Vh and V l＜Vh, the output V i of the concentration-current model is given a weight W i2, the output V l of the concentration-light intensity model is given a weight W l2, and W i2＜Wl2, the dust concentration is
Vr＝Vi*wi2+Vl*wl2；
S45: if V i＞Vh and V l＜Vh or V i＜Vh and V l＞Vh appear, a fault signal is sent to the alarm system, and a buzzer of the alarm display system sounds intermittently to remind a worker to perform fault detection and replacement on the light absorption subunit and the charge induction subunit respectively.
The step S5 adopts GRU network to perform feature fusion, and specifically comprises the following steps:
S51: inputting the output V i of the temperature T, concentration-current model and the output V l of the concentration-light intensity model into a GRU information fusion model;
s52: respectively calculating sample entropy of the temperature T, the output V i of the concentration-current model and the output V l of the concentration-light intensity model, and forming a feature matrix SE by calculation results;
S53: inputting the feature matrix SE into the GRU network, and calculating the output r t of the reset gate:
rt＝σ(wr*[ht-1,SE])
Wherein: sigma is a Sigmoid activation function;
w r is the weight parameter of the layer;
h t-1 is the output of the hidden layer at the previous time;
S54: the output u t of the update gate is calculated from the feature matrix SE:
ut＝σ(wu*[ht-1,SE])
wherein: w u is the weight parameter of the layer;
S55: calculating an output g t of the memory state:
gt＝tanh(wg*[rt⊙ht-1,SE])
Wherein: w g is the weight parameter of the layer;
The ". Altern represents Hadmard product operation;
S56: and (3) calculating and outputting:
ht＝(1-ut)⊙ht-1+utgt。
step S6 includes the steps of:
s61: using a Softmax function as a classification function, taking a result h t of feature fusion as input, and taking a discrimination value rho of the current dust concentration overrun and a discrimination value tau of the dust concentration about to overrun as output;
s62: if ρ=1 or τ=1, the alarm system is set to alarm;
S63: if ρ=0 and τ=0, the environment continues to be monitored.
The step S7 specifically includes the following steps:
S71: if ρ=1, namely the current dust concentration exceeds the limit, turning off the green LED lamp, intermittently and rapidly sounding the buzzer, and rapidly flashing the red LED lamp;
S72: if τ=1, namely the dust concentration is over-limit, the green LED lamp is always on, the buzzer is suddenly sounded intermittently, and the red LED lamp is kept in a closed state;
S73: if ρ=0 and τ=0, i.e. the current dust concentration is not exceeded and the dust concentration will not be exceeded, the green LED lamp is normally on and the buzzer and the red LED lamp remain off.
Compared with the prior art, the invention senses the related parameters of the dust environment by arranging the environment sensing system, including temperature T, humidity H, current I and light intensity L, the data processing system receives the parameter signals from the environment sensing system, processes, corrects and fuses the related parameter signals detected by the environment sensing system by the built-in humidity-error model, concentration-current model, concentration-light intensity model and information fusion model, thereby obtaining dust concentration data of the dust environment, and displays and feeds back alarm by the alarm display system; the light intensity sensed by the light absorption subunit has a negative correlation with the dust concentration in the environment, the light absorption subunit has better applicability in the environment with low dust concentration, the internal current of the charge induction subunit has a positive correlation with the dust concentration, and the charge induction subunit has better applicability in the environment with high dust concentration; in addition, the invention also introduces the environmental temperature possibly influenced by the dust concentration as input, and adopts an information fusion algorithm to early warn the environmental dust overrun.
Drawings
FIG. 1 is a schematic diagram of the structure of the present invention;
FIG. 2 is a diagram showing the effect of dust diversion after the diversion in FIG. 1;
FIG. 3 is a schematic diagram of a data processing flow of the present invention;
FIG. 4 is a diagram of a humidity-error model network of the present invention;
FIG. 5 is a block diagram of a GRU feature fusion of the invention;
In the figure: 1. the system comprises a power supply system 2, an environment sensing system 21, a ventilation pipeline 22, a laser probe 23, a convex lens 24, an optical fiber spectrometer 25, a three-edged split-flow column 26, a deceleration column 27, a sensing electrode 28, a current amplifier 3, a data processing system 41, a display device 42 and an alarm device.
Detailed Description
The invention is further described below with reference to the accompanying drawings.
As shown in fig. 1 and fig. 2, a dust concentration detection system based on fusion of light absorption and charge induction method comprises a power supply system 1, an environment induction system 2, a data processing system 3 and an alarm display system, wherein the power supply system 1 supplies power to the environment induction system 2, the environment induction system 2 comprises an air inlet pipeline 21, a temperature and humidity sensor, a light absorption subunit, a charge induction subunit and a split body, the temperature and humidity sensor, the light absorption subunit and the charge induction subunit are integrated in the air inlet pipeline 21 in a serial connection mode, the temperature and humidity sensor is arranged at the front end of the air inlet pipeline 21, a split body is arranged between the light absorption subunit and the charge induction subunit, so that dust airflow is split at the front end of the charge induction subunit and generates a tiny karman vortex street phenomenon, and further contacts with electrodes in the charge induction subunit more fully, and the detection precision is improved;
The data processing system 3 consists of a singlechip and an external circuit, wherein a humidity-error model, a concentration-current model, a concentration-light intensity model and an information fusion model are built in the data processing system 3, data detected by the environment sensing system 2 are corrected and fused, and the data processing system 3 is connected with an alarm display system.
The alarm display system comprises a display device 41 and an alarm device 42, wherein the display device 41 comprises an LCD display screen, a red LED lamp and a green LED lamp, the alarm device 42 is a buzzer, and the alarm display system is connected with the data processing system 3, receives signal information fed back by the data processing system 3 and outputs a corresponding alarm mode.
The light absorption subunit comprises a laser probe 22, a convex lens 24 and an optical fiber spectrometer 23, and after the light absorption subunit is arranged on the temperature and humidity sensor, the light absorption subunit has better applicability in a low-concentration dust environment, and the light intensity sensed by the optical fiber spectrometer 23 and the environment dust concentration are in a negative correlation; the charge induction subunit comprises an induction electrode 28 and a current amplifier 29, the charge induction subunit is arranged at the rear end of the air inlet pipeline 21, the charge induction subunit has better applicability in a high-concentration dust environment, and the internal current of the charge induction subunit has positive correlation with the concentration of environmental dust; the flow splitting body is composed of a triangular splitting column 26 and a deceleration column 27 arranged on an extension line of the splitting column, dust flow passing through the light absorption subunit is split to two ends of the air inlet pipeline 21 by the triangular splitting column 26, is decelerated and split again after flowing to the deceleration column 27, and the split inner dust flow is in a tiny karman vortex street phenomenon and is in full contact with an electrode, so that the detection precision can be improved.
As shown in fig. 3 to 5, a dust concentration detection method based on fusion of light absorption and charge induction method comprises the following steps:
s1: the device is turned on, and the data processing system 3 receives data information from the environment sensing system 2, wherein the data information comprises temperature T, humidity H, current I and light intensity L, and specifically comprises:
The dust flow enters the air inlet pipeline 21, and the temperature T and the humidity H of the dust flow are detected by a temperature and humidity sensor; scattered light emitted by the laser probe 22 becomes parallel light after passing through the convex lens 24, and the light intensity of the parallel light is attenuated after passing through dust flow and finally sensed by the optical fiber spectrometer 23, namely, light intensity L; the dust generates induced current after contacting with the induction electrode 28, but the induced current is smaller, and the current is amplified by using the current amplifier 29 to obtain current I;
S2: preprocessing the data information received in the step S1 to obtain a processed signal sequence of humidity H 0, temperature T 0, current I 0 and light intensity L 0;
S3: correcting the preprocessed current I 0 and the light intensity L 0 signal sequence by using a humidity-error model to obtain corrected current and light intensity data;
s4: the corrected current and light intensity data are respectively input into a concentration-current model and a concentration-light intensity model, a certain weight is respectively given to the output of the concentration-current model and the output of the concentration-light intensity model according to the situation, the environmental dust concentration is obtained by adding, and the environmental dust concentration data is displayed on an LCD display screen of the display device 41;
S5: performing feature fusion on the output of the temperature, concentration-current model and the output of the concentration-light intensity model by using a GRU network;
S6: classifying the feature fusion result by using a Softmax function to obtain a judging result of whether the concentration of environmental dust exceeds a limit value;
s7: if the current environmental dust concentration exceeds the limit or the environmental dust concentration is about to exceed the limit, the display device 42 and the alarm device 42 of the alarm display system are made to alarm and display related information.
The step S2 of preprocessing the data information includes the steps of:
S21: soft threshold denoising is carried out on the received signal sequences of the temperature T, the humidity H, the current I and the light intensity L by using a wavelet basis function of a sym wavelet system;
s22: detecting abnormal values in the data by using a local abnormal factor detection technology, and filling the abnormal values by using an interpolation method;
S23: and carrying out normalization processing on the modified data, wherein the normalization relation is as follows:
Wherein: Mapping at/>, after normalization Results in;
x is the original data;
x min and x max are the minimum and maximum values of the raw data, respectively;
S24: the processed signals of humidity H 0, temperature T 0, current I 0 and light intensity L 0 are obtained.
The step S3 carries out error correction on the preprocessed data by adopting a BP neural network, and specifically comprises the following steps:
S31: network initialization takes a signal sequence H 0 consisting of humidity data as input; error matrix Err of current and light intensity as output
Err＝[err1,err2]T
Wherein: err 1 is the current error sequence;
err 2 is the light intensity error sequence;
So the input layer node l is 1, and the output layer node m is 2; the number of hidden layer nodes n is:
Preliminarily taking a as 3 and hiding layer node
S32: calculating an output H i of the hidden layer node:
Hi＝f(wi-ai) 1≤i≤5
Wherein: Activating a function for Sigmoid;
w i is the connection weight between the input layer and the hidden layer;
a i is a threshold in the hidden layer;
i is the node number of the hidden layer;
S33: the output of the output layer node is calculated, namely error err k:
Wherein: w ik is the weight between the hidden layer and the output layer;
b k is the threshold of the hidden layer;
S34: calculating a corrected current sequence I r and a corrected light intensity sequence L r
Ir＝I0+err1
Lr＝L0+err2。
Step S4 comprises the steps of:
S41: respectively inputting the corrected current sequence I r and the corrected light intensity sequence L r into a concentration-current model and a concentration-light intensity model;
S42: respectively outputting a concentration value V i and a concentration value V l predicted by a concentration-current model and a concentration-light intensity model;
S43: if V i＞Vh and V l＞Vh, the output V i of the concentration-current model is given weight W i1, the output V l of the concentration-light intensity model is given weight W l1, and the dust concentration is calculated
Vr＝Vi*wi1+Vl*wl1
Wherein: v h is the concentration high-low judgment value determined by experiments, and W i1＞Wl1;
S44: if V i＜Vh and V l＜Vh, the output V i of the concentration-current model is given a weight W i2, the output V l of the concentration-light intensity model is given a weight W l2, and W i2＜Wl2, the dust concentration is
Vr＝Vi*wi2+Vl*wl2；
S45: if V i＞Vh and V l＜Vh or V i＜Vh and V l＞Vh appear, a fault signal is sent to the alarm device 42, and the buzzer of the alarm device 42 sounds intermittently to remind the staff to perform fault detection and replacement on the light absorbing subunit and the charge sensing subunit respectively.
The step S5 adopts GRU network to perform feature fusion, and specifically comprises the following steps:
S51: inputting the output V i of the temperature T, concentration-current model and the output V l of the concentration-light intensity model into a GRU information fusion model;
s52: respectively calculating sample entropy of the temperature T, the output V i of the concentration-current model and the output V l of the concentration-light intensity model, and forming a feature matrix SE by calculation results;
S53: inputting the feature matrix SE into the GRU network, and calculating the output r t of the reset gate:
rt＝σ(wr*[ht-1,SE])
Wherein: sigma is a Sigmoid activation function;
w r is the weight parameter of the layer;
h t-1 is the output of the hidden layer at the previous time;
S54: the output u t of the update gate is calculated from the feature matrix SE:
ut＝σ(wu*[ht-1,SE])
wherein: w u is the weight parameter of the layer;
S55: calculating an output g t of the memory state:
gt＝tanh(wg*[rt⊙ht-1,SE])
Wherein: w g is the weight parameter of the layer;
The ". Altern represents Hadmard product operation;
S56: and (3) calculating and outputting:
ht＝(1-ut)⊙ht-1+ut⊙gt。
step S6 includes the steps of:
s61: using a Softmax function as a classification function, taking a result h t of feature fusion as input, and taking a discrimination value rho of the current dust concentration overrun and a discrimination value tau of the dust concentration about to overrun as output;
s62: if ρ=1 or τ=1, the alarm system is set to alarm;
S63: if ρ=0 and τ=0, the environment continues to be monitored.
The step S7 specifically includes the following steps:
S71: if ρ=1, namely the current dust concentration exceeds the limit, turning off the green LED lamp, intermittently and rapidly sounding the buzzer, and rapidly flashing the red LED lamp;
S72: if τ=1, namely the dust concentration is over-limit, the green LED lamp is always on, the buzzer is suddenly sounded intermittently, and the red LED lamp is kept in a closed state;
S73: if ρ=0 and τ=0, i.e. the current dust concentration is not exceeded and the dust concentration will not be exceeded, the green LED lamp is normally on and the buzzer and the red LED lamp remain off. .
The following is a description of an embodiment of the dust concentration detection system of the present invention for monitoring the dust concentration in a work place in real time
The dust concentration detection system is deployed in a crowd concentration area in a workplace, and the deployment height is 1.5 to 1.6 meters, so that the collection range is ensured to be in a human breathing area. The environment information is acquired through the environment sensing system, and the acquired data information comprises a temperature sequence (T), a humidity sequence (H), a light intensity sequence (L) and a current sequence (I). The monitoring data of the system in a certain period are shown in table 1
Table 1 monitoring data for a period of time in an environmental sensing system
Soft threshold denoising is carried out on the received temperature, humidity, current and light intensity signal sequences by using a wavelet basis function of a sym wavelet system, then abnormal values in data are detected by using a local abnormal factor detection technology, the abnormal values are filled by an interpolation method, and finally, the modified data are normalized, wherein the data after pretreatment in the period are shown in table 2.
Table 2 monitoring data after pretreatment for a certain period of time
I.e. the temperature sequence after pretreatment is T 0 = [0.3 0.3 0.25 0.275 0.25];
The humidity sequence is H 0 = [0.325 0.3 0.325 0.325 0.35];
the light intensity sequence is L 0 = [0.76 0.79 0.79 0.76 0.76];
the current sequence is I 0 = [0.17 0.23 0.19 0.21 0.17];
The humidity-error model established by BP neural network is adopted to carry out error correction on the preprocessed light intensity and current data, a signal sequence H 0 = [0.325 0.3 0.325 0.325 0.35] formed by the humidity data is taken as input, an error matrix Err of the current and the light intensity is taken as output, and the error matrix is calculated by the humidity-error model to be
The corrected current sequence is I r＝I0+err1 = [0.183 0.221 0.203 0.223 0.187] and the light intensity sequence is L r＝L0+err2 = [0.784 0.779 0.814 0.784 0.787].
The corrected current sequence I r＝I0+err1 = [0.183 0.221 0.203 0.223 0.187] and the light intensity sequence L r＝L0+err2 = [0.784 0.779 0.814 0.784 0.787] are input into a concentration-current model and a concentration-light intensity model, respectively, and predicted concentration values V i = [1.71 1.75 1.73 1.75 1.71] and V l = [1.18 1.16 1.24 1.18 1.19] are output, respectively. V h in the application scene takes 6mg/m 3, and as V i＜Vh and V l＜Vh are adopted, w i2＝0.27wl2 =0.73, and the dust concentration Vr＝Vi*wi2+Vl*wl2＝Vi*0.27+Vl*0.73＝[1.323 1.319 1.372 1.334 1.330], is displayed on an LCD screen in real time.
The result of detecting the on-site dust concentration by using the filter membrane weighing method is 1.39mg/m 3, so that the dust concentration error detected by using the system in the periodThe average error is only 3.94%, so the system has a good monitoring effect on the low-concentration dust environment.
And finally, inputting the preprocessed temperature sequence T 0 = [0.3 0.3 0.25 0.275 0.25], the output V i = [1.71 1.75 1.73 1.75 1.71] of the concentration-current model and the output V l = [1.18 1.16 1.24 1.18 1.19] of the concentration-light intensity model into a GRU information fusion model, classifying fusion results by using a softmax function, wherein the classification results ρ=0 and τ=0, namely, the current dust concentration is not overrun and the dust concentration is not overrun, so that the green LED lamp is always on, and the buzzer and the red LED lamp are kept in a closed state.
Claims (9)
1. The dust concentration detection system based on fusion of light absorption and charge induction is characterized by comprising a power supply system, an environment induction system, a data processing system and an alarm display system, wherein the power supply system supplies power to the environment induction system, the environment induction system comprises an air inlet pipeline, a temperature and humidity sensor, a light absorption subunit, a charge induction subunit and a split body, the temperature and humidity sensor, the light absorption subunit and the charge induction subunit are integrated in the air inlet pipeline in a serial connection mode, the temperature and humidity sensor is arranged at the front end of the air inlet pipeline, and a split fluid is arranged between the light absorption subunit and the charge induction subunit;
the data processing system consists of a singlechip and an external circuit, and is internally provided with a humidity-error model, a concentration-current model, a concentration-light intensity model and an information fusion model, so as to correct and fuse the data detected by the environment sensing system;
The alarm display system is connected with the data processing system, receives signal information fed back by the data processing system and outputs a corresponding alarm mode.
2. The dust concentration detection system based on fusion of light absorption and charge induction method as set forth in claim 1, wherein the light absorption subunit comprises a laser probe, a convex lens and an optical fiber spectrometer, and after the light absorption subunit is installed on a temperature and humidity sensor, the light absorption subunit has better applicability in a low-concentration dust environment, and the light intensity sensed by the optical fiber spectrometer has a negative correlation with the environmental dust concentration; the charge induction subunit comprises an induction electrode and a current amplifier, is arranged at the rear end of the air inlet pipeline, has good applicability in a high-concentration dust environment, and has positive correlation between the internal current and the environment dust concentration; the flow splitting body consists of a triangular splitting column and a deceleration column arranged on an extension line of the splitting column, dust flow passing through the light absorption subunit is split to two ends of the pipeline by the triangular splitting column, is decelerated and split again after flowing to the deceleration column, and the split inner dust flow is in a tiny karman vortex street phenomenon and is in full contact with the electrode.
3. A dust concentration detection method based on fusion of light absorption and charge induction according to claim 2, comprising the steps of:
S1: starting equipment, and receiving data information from an environment sensing system by a data processing system, wherein the data information comprises temperature T, humidity H, current I and light intensity L, and specifically comprises the following steps:
The dust flow enters an air inlet pipeline, and the temperature T and the humidity H of the dust flow are detected by a temperature and humidity sensor; scattered light emitted by the laser probe becomes parallel light after passing through the convex lens, and the light intensity of the parallel light is attenuated after passing through dust flow and finally sensed by the optical fiber spectrometer, namely light intensity L; the dust generates induction current after contacting with the induction electrode, but the induction current is smaller, and the current I is obtained by amplifying the dust by using an amplifier;
S2: preprocessing the data information received in the step S1 to obtain a processed signal sequence of humidity H 0, temperature T 0, current I 0 and light intensity L 0;
S3: correcting the preprocessed current I 0 and the light intensity L 0 signal sequence by using a humidity-error model to obtain corrected current and light intensity data;
S4: the corrected current and light intensity data are respectively input into a concentration-current model and a concentration-light intensity model, a certain weight is respectively given to the output of the concentration-current model and the output of the concentration-light intensity model according to the situation, the environmental dust concentration is obtained by adding, and the environmental dust concentration data is displayed on an alarm display system;
S5: performing feature fusion on the output of the temperature, concentration-current model and the output of the concentration-light intensity model by using a GRU network;
S6: classifying the feature fusion result by using a Softmax function to obtain a judging result of whether the concentration of environmental dust exceeds a limit value;
s7: if the current environment dust concentration exceeds the limit or the environment dust concentration is about to exceed the limit, the alarm display system is made to alarm and display related information.
4. A dust concentration detection method based on fusion of light absorption and charge sensing according to claim 3, wherein the preprocessing of the data information in step S2 comprises the steps of:
S21: soft threshold denoising is carried out on the received signal sequences of the temperature T, the humidity H, the current I and the light intensity L by using a wavelet basis function of a sym wavelet system;
s22: detecting abnormal values in the data by using a local abnormal factor detection technology, and filling the abnormal values by using an interpolation method;
S23: and carrying out normalization processing on the modified data, wherein the normalization relation is as follows:
Wherein: Mapping at/>, after normalization Results in;
x is the original data;
x max is the minimum value and the maximum value of the original data respectively;
S24: the processed signals of humidity H 0, temperature T 0, current I 0 and light intensity L 0 are obtained.
5. The dust concentration detection method based on fusion of light absorption and charge induction method according to claim 3, wherein step S3 uses a BP neural network to perform error correction on the preprocessed data, and specifically comprises the following steps:
S31: network initialization takes a signal sequence H 0 consisting of humidity data as input; error matrix Err of current and light intensity as output
Err＝[err1,err2]T
Wherein: err 1 is the current error sequence;
err 2 is the light intensity error sequence;
So the input layer node l is 1, and the output layer node m is 2; the number of hidden layer nodes n is:
1≤a≤10
Preliminarily taking a as 3 and hiding layer node
S32: calculating an output H i of the hidden layer node:
Hi＝f(wi-ai)1≤i≤5
Wherein: Activating a function for Sigmoid;
w i is the connection weight between the input layer and the hidden layer;
a i is a threshold in the hidden layer;
i is the node number of the hidden layer;
S33: the output of the output layer node is calculated, namely error err k:
1≤k≤2
Wherein: w ik is the weight between the hidden layer and the output layer;
b k is the threshold of the hidden layer;
S34: calculating a corrected current sequence I r and a corrected light intensity sequence L r
Ir＝I0+err1
Lr＝L0+err2。
6. A dust concentration detection method based on fusion of light absorption and charge induction according to claim 3, wherein step S4 comprises the steps of:
S41: respectively inputting the corrected current sequence I r and the corrected light intensity sequence L r into a concentration-current model and a concentration-light intensity model;
S42: respectively outputting a concentration value V i and a concentration value V l predicted by a concentration-current model and a concentration-light intensity model;
S43: if V i＞Vh and V l＞Vh, the output V i of the concentration-current model is given weight W i1, the output V l of the concentration-light intensity model is given weight W l1, and the dust concentration is calculated
Vr＝Vi*wi1+Vl*wl1
Wherein: v h is the concentration high-low judgment value determined by experiments, and W i1＞Wl1;
S44: if V i＜Vh and V l＜Vh, the output V i of the concentration-current model is given a weight W i2, the output V l of the concentration-light intensity model is given a weight W l2, and W i2＜Wl2, the dust concentration is
Vr＝Vi*wi2+Vl*wl2；
S45: if V i＞Vh and V l＜Vh or V i＜Vh and V l＞Vh appear, a fault signal is sent to the alarm system, and a buzzer of the alarm display system sounds intermittently to remind a worker to perform fault detection and replacement on the light absorption subunit and the charge induction subunit respectively.
7. The dust concentration detection method based on fusion of light absorption and charge induction method according to claim 3, wherein step S5 adopts a GRU network for feature fusion, and specifically comprises the following steps:
S51: inputting the output V i of the temperature T, concentration-current model and the output V l of the concentration-light intensity model into a GRU information fusion model;
s52: respectively calculating sample entropy of the temperature T, the output V i of the concentration-current model and the output V l of the concentration-light intensity model, and forming a feature matrix SE by calculation results;
S53: inputting the feature matrix SE into the GRU network, and calculating the output r t of the reset gate:
rt＝σ(wr*[ht-1,SE])
Wherein: sigma is a Sigmoid activation function;
w r is the weight parameter of the layer;
h t-1 is the output of the hidden layer at the previous time;
S54: the output u t of the update gate is calculated from the feature matrix SE:
ut＝σ(wu*[ht-1,SE])
wherein: w u is the weight parameter of the layer;
S55: calculating an output g t of the memory state:
gt＝tanh(wg*[rt⊙ht-1,SE])
Wherein: w g is the weight parameter of the layer;
The ". Altern represents Hadmard product operation;
S56: and (3) calculating and outputting:
ht＝(1-ut)⊙ht-1+ut⊙gt。
8. A dust concentration detection method based on fusion of light absorption and charge sensing according to claim 3, wherein step S6 comprises the steps of:
s61: using a Softmax function as a classification function, taking a result h t of feature fusion as input, and taking a discrimination value rho of the current dust concentration overrun and a discrimination value tau of the dust concentration about to overrun as output;
s62: if ρ=1 or τ=1, the alarm system is set to alarm;
S63: if ρ=0 and τ=0, the environment continues to be monitored.
9. A dust concentration detection method based on fusion of light absorption and charge induction method according to claim 3, wherein step S7 specifically comprises the steps of:
S71: if ρ=1, namely the current dust concentration exceeds the limit, turning off the green LED lamp, intermittently and rapidly sounding the buzzer, and rapidly flashing the red LED lamp;
S72: if τ=1, namely the dust concentration is over-limit, the green LED lamp is always on, the buzzer is suddenly sounded intermittently, and the red LED lamp is kept in a closed state;
S73: if ρ=0 and τ=0, i.e. the current dust concentration is not exceeded and the dust concentration will not be exceeded, the green LED lamp is normally on and the buzzer and the red LED lamp remain off.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410033474.XA CN117907174A (en) | 2024-01-10 | 2024-01-10 | Dust concentration detection system and method based on fusion of light absorption and charge induction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410033474.XA CN117907174A (en) | 2024-01-10 | 2024-01-10 | Dust concentration detection system and method based on fusion of light absorption and charge induction method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117907174A true CN117907174A (en) | 2024-04-19 |
Family
ID=90684963
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410033474.XA Pending CN117907174A (en) | 2024-01-10 | 2024-01-10 | Dust concentration detection system and method based on fusion of light absorption and charge induction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117907174A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107402173A (en) * | 2017-08-22 | 2017-11-28 | 苏州同阳科技发展有限公司 | AC coupled formula electric charge induction method dust concentration monitoring device and method |
| CN110779839A (en) * | 2018-07-31 | 2020-02-11 | 深圳市白麓嵩天科技有限责任公司 | Charge induction dust measuring method based on trend fluctuation analysis |
| CN111579446A (en) * | 2020-05-19 | 2020-08-25 | 中煤科工集团重庆研究院有限公司 | Dust concentration detection method based on optimal fusion algorithm |
| CN117233054A (en) * | 2023-11-13 | 2023-12-15 | 中国科学技术大学 | Method for correcting contamination interference of optical fiber probe, correction system and sensor |
| US20230410012A1 (en) * | 2022-06-16 | 2023-12-21 | Qingdao university of technology | Project disaster warning method and system based on collaborative fusion of multi-physics monitoring data |
-
2024
- 2024-01-10 CN CN202410033474.XA patent/CN117907174A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107402173A (en) * | 2017-08-22 | 2017-11-28 | 苏州同阳科技发展有限公司 | AC coupled formula electric charge induction method dust concentration monitoring device and method |
| CN110779839A (en) * | 2018-07-31 | 2020-02-11 | 深圳市白麓嵩天科技有限责任公司 | Charge induction dust measuring method based on trend fluctuation analysis |
| CN111579446A (en) * | 2020-05-19 | 2020-08-25 | 中煤科工集团重庆研究院有限公司 | Dust concentration detection method based on optimal fusion algorithm |
| US20230410012A1 (en) * | 2022-06-16 | 2023-12-21 | Qingdao university of technology | Project disaster warning method and system based on collaborative fusion of multi-physics monitoring data |
| CN117233054A (en) * | 2023-11-13 | 2023-12-15 | 中国科学技术大学 | Method for correcting contamination interference of optical fiber probe, correction system and sensor |
Non-Patent Citations (1)
| Title |
|---|
| 陈建阁;吴付祥;王杰;: "电荷感应法粉尘浓度检测技术", 煤炭学报, no. 03, 15 March 2015 (2015-03-15), pages 231 - 236 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US6967582B2 (en) | Detector with ambient photon sensor and other sensors | |
| CN115077627B (en) | Multi-fusion environmental data supervision method and supervision system | |
| CN108830305A (en) | A kind of real-time fire monitoring method of combination DCLRN network and optical flow method | |
| CN111986436A (en) | Comprehensive flame detection method based on ultraviolet and deep neural networks | |
| CN112002095A (en) | Fire early warning method in mine tunnel | |
| CN113192283B (en) | Wireless fire early warning system with multi-sensor information fusion | |
| JP2007148869A (en) | Abnormality detection device | |
| CN106056831A (en) | Smoke alarm control method based on computer processing and system thereof | |
| CN112198209A (en) | Formaldehyde detection method and formaldehyde detection device | |
| CN114839317B (en) | Comprehensive early warning method and system for atmosphere meshing | |
| CN206773865U (en) | A kind of gas recombination detector for cable fire | |
| CN116972401A (en) | Exhaust-heat boiler energy storage monitoring method based on flue gas fluctuation | |
| CN117907174A (en) | Dust concentration detection system and method based on fusion of light absorption and charge induction method | |
| CN205981287U (en) | Bridge, tunnel, piping lane or healthy management system of big dam structure | |
| CN212569977U (en) | Fire identification alarm device based on grey and RBF double-layer neural networks | |
| CN107831699A (en) | A kind of intelligent data acquisition analysis method and system | |
| CN111798638A (en) | Auxiliary system fire information processing method based on information fusion | |
| CN204791387U (en) | Intelligence smoke detector | |
| Wang et al. | Forest fire detection system based on Fuzzy Kalman filter | |
| CN208170631U (en) | AI intelligent fresh air equipment and AI intelligent environment control system | |
| CN110120142B (en) | Fire smoke video intelligent monitoring early warning system and early warning method | |
| JP2008140222A (en) | Abnormality detection device and abnormality detection method | |
| CN213424156U (en) | Fire prediction device | |
| CN116546043A (en) | Cable tunnel inspection personnel environment safety monitoring device and method | |
| CN114582083B (en) | Tunnel monitoring multi-sensor data fusion fire disaster early warning method |
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 |