CN108415336A - Fluorescence temperature demodulating system based on radial base neural net and method - Google Patents
Fluorescence temperature demodulating system based on radial base neural net and method Download PDFInfo
- Publication number
- CN108415336A CN108415336A CN201810344413.XA CN201810344413A CN108415336A CN 108415336 A CN108415336 A CN 108415336A CN 201810344413 A CN201810344413 A CN 201810344413A CN 108415336 A CN108415336 A CN 108415336A
- Authority
- CN
- China
- Prior art keywords
- temperature
- neural net
- radial base
- fbg
- demodulator
- 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
- 230000001537 neural effect Effects 0.000 title claims abstract description 45
- 238000000034 method Methods 0.000 title claims abstract description 22
- 238000012360 testing method Methods 0.000 claims abstract description 17
- 238000012549 training Methods 0.000 claims description 27
- 230000006870 function Effects 0.000 claims description 16
- 230000003287 optical effect Effects 0.000 claims description 13
- 239000000523 sample Substances 0.000 claims description 13
- 238000009529 body temperature measurement Methods 0.000 claims description 9
- 238000013528 artificial neural network Methods 0.000 claims description 7
- 239000007850 fluorescent dye Substances 0.000 claims description 7
- 238000013480 data collection Methods 0.000 claims description 6
- 230000005284 excitation Effects 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000001186 cumulative effect Effects 0.000 claims description 3
- 230000000977 initiatory effect Effects 0.000 claims description 3
- 238000003064 k means clustering Methods 0.000 claims description 3
- 210000005036 nerve Anatomy 0.000 claims description 3
- 210000002569 neuron Anatomy 0.000 claims description 3
- 239000013598 vector Substances 0.000 claims description 3
- 238000000465 moulding Methods 0.000 claims 1
- 238000004519 manufacturing process Methods 0.000 abstract description 9
- 230000008569 process Effects 0.000 abstract description 5
- 239000000463 material Substances 0.000 abstract description 3
- 238000005516 engineering process Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000013475 authorization Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000009841 combustion method Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 238000010438 heat treatment Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000004861 thermometry Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B19/00—Programme-control systems
- G05B19/02—Programme-control systems electric
- G05B19/04—Programme control other than numerical control, i.e. in sequence controllers or logic controllers
- G05B19/042—Programme control other than numerical control, i.e. in sequence controllers or logic controllers using digital processors
- G05B19/0423—Input/output
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/20—Pc systems
- G05B2219/24—Pc safety
- G05B2219/24215—Scada supervisory control and data acquisition
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Investigating, Analyzing Materials By Fluorescence Or Luminescence (AREA)
Abstract
The invention discloses a kind of fluorescence temperature demodulating system and method based on radial base neural net, the data that fluorescence temperature (FBG) demodulator and insulating box are acquired by master controller are compared, and contrast signal is forwarded to PC machine, data are handled in PC machine, finally obtain the radial base neural net for meeting production requirement, and it is burned onto in fluorescence temperature (FBG) demodulator, this system can realize that datamation acquires, program batch is downloaded, test of product performance, it accomplishes scale production, calibration process is solved to need to expend a large amount of manpower financial capacity's material resources, production efficiency is low, problem with high costs.
Description
Technical field
The invention belongs to fibre optical sensor fields, and in particular to a kind of fluorescence temperature demodulation based on radial base neural net
System and method.
Background technology
Fluorescence thermometry has small high temperature resistant, electromagnetism interference, high-voltage isulation, size, high sensitivity and long lifespan etc.
The incomparable advantage of electric sensing technology obtains extensive concern and research.Nowadays the commercialization application stage is stepped into, is answered
For industries such as high-voltage electrical equipment, Microwave Industry, heat treatment unit, petrochemical industry, aerospaces.However, by more than 30 years
Development, fluorescence thermometric product is still small range, small-scale application, and application is not achieved in mainly measurement accuracy of tracing it to its cause
Demand, calibration process are cumbersome, of high cost and system stability is to be improved.
Fluorescence temperature-measurement principle be based on certain physical attribute of fluorescent material itself and the relationship of temperature, as fluorescence intensity with
Fluorescence lifetime is measured into trip temperature because its apparent advantage has been increasingly becoming major way using fluorescence lifetime.According to glimmering
The generality of flash ranging temperature is theoretical, and phosphorescence afterglow curve is exponential curve, twilight sunset constant and excitation light source intensity, coupling efficiency, light
Other variables of the systems such as fine length, loss are unrelated, but find that its influence factor is various in putting into practice, and such as pop one's head in consistent
Property, channel characteristic, light source drift, environmental change etc. can all influence temperature measurement accuracy and stability.Thus, practical decay curve is presented
Go out non-exponential dynamic trend.
To solve the above-mentioned problems, 2013104125995 (Authorization Notice No. of Patent No.：CN103428982B China)
Patent of invention《A kind of Self-adjustment of luminous intensity circuit for fluorescence optical fiber excitation》It is middle to use feedback control technology, it is shone by controlling
The electric current that body flows through, adjust automatically light intensity export constant light intensity in scheduled range, eliminate light-intensity variation and are done to measurement
It disturbs, but it is one of influence factor to encourage light intensity, fails the influence for eliminating other external factor；Jia Dan equality people exist《More than fluorescence
The non-exponential component of brightness and processing》It is approximately multi index option function superposition to change the non-exponential of decay curve in article, and use is general
Each service life of Luo Nifa split cavity oscillator twilight sunset, but solution procedure is complicated, it is more demanding to hardware device, still remain calibration
Difficult, production capacity deficiency problem.Above-mentioned improvement project improves the precision and stability of fluorescence temperature measurement system to a certain extent, but all
It is not system total solution with one-sidedness.
Invention content
The fluorescence temperature demodulation based on radial base neural net that the purpose of the present invention is to overcome the above shortcomings and to provide a kind of
System and method, solves that calibration process of the existing technology is cumbersome, production capacity is insufficient, of high cost, precision and stability is not high asks
Topic.
In order to achieve the above object, the fluorescence temperature demodulating system based on radial base neural net, including master controller, it is main
Controller connection fluorescence temperature (FBG) demodulator, insulating box and PC machine, fluorescence temperature (FBG) demodulator and PC machine are all connected with CD writers, fluorescence temperature
It spends (FBG) demodulator and connects insulating box, CD writers connect several burning engine bases；
Master controller is used to acquire fluorescence temperature (FBG) demodulator and the data of insulating box are compared, and contrast signal is forwarded
To PC machine；
PC machine is used to screen qualified radial base neural net, and by CD writers burning in fluorescence temperature (FBG) demodulator.
Fluorescence temperature (FBG) demodulator accesses insulating box by fluorescent probe.
Master controller includes control module, has cd-rom recorder interface for connecting cd-rom recorder in control module and for connecting
The Host Controler Interface of PC machine is connect, control module connects optical system by two paths of signals, and control module passes through on signal all the way
Connect optical system after the excitation light drive module of setting, control module passes through at the data that are disposed on another way signal
Optical system is connected after reason module and fluorescence signal detecting module, fluorescence of the optical system connection for being connect with insulating box is visited
Head, control module connect display.
Fluorescence temperature (FBG) demodulator includes CPU, on CPU there is insulating box interface, fluorescence temperature (FBG) demodulator interface, display to connect
Mouth and PC machine interface.
A kind of working method of the fluorescence temperature demodulating system based on radial base neural net, including off-line training step and
Carry out the stage in line selection；
Off-line training step includes the following steps：
Step 1 sets master controller operating mode to off-line training, be arranged in master controller minimum collecting temperature,
Highest collecting temperature and temperature acquisition interval, when insulating box reaches assigned temperature, master controller sends acquisition to fluorescence
Temperature demodulation instrument carries out data acquisition, chooses two sections that time interval in phosphorescence afterglow curve is Δ t, seeks voltage signal values respectively
It is cumulative and, regard the two values as the input of neural network, PC machine, preservation number be sent to reference temperature composition data collection
According to, and the temperature acquisition interval set according to main controller, it is automatic to carry out acquisition next time until terminating；
Collected data set is done normalized by step 2；
Step 3, using a part for the data after normalized as training set and another part as test set；
Training set three layers of radial base neural net of input are trained, and preserve number of results by step 4, initiation parameter
According to；
Step 5 tests the radial base neural net after training using test set；
Step 6 is downloaded to the trained radial base neural net of PC machine by cd-rom recorder if meeting required precision
In CPU used in fluorescence temperature (FBG) demodulator；
If being unsatisfactory for required precision, increase the data of acquisition, repartition training set and test set, repeats step 3 extremely
Step 6 is until meeting the requirements；
Carry out the stage in line selection to include the following steps：
Step 1, the operating mode that master controller is arranged are product test, and temperature measurement accuracy is arranged；
Step 2, fluorescence temperature (FBG) demodulator acquire the data of insulating box；
Step 3 passes through the radial base nerve after training using real-time data collection as the input of radial base neural net
Network obtains real time temperature；
Whether step 4, master controller judge calorstat temperature with fluorescence temperature (FBG) demodulator temperature measurement accuracy in error permission model
In enclosing, service check is completed.
The specific method is as follows for radial base neural net training：
The first step seeks the central value t of radial basis function using k-means clustering algorithmsi；ti(n) nth iteration is indicated
Ith cluster center, wherein i=1,2,3 ..., I, I value are rule of thumb chosen or randomly choosed；
Second step, initialization；Select I mutually different vectors as initial cluster center ti(O) (i=1,2,3 ...,
I)；
Third walks, and calculates each sample point X of the input spacekWith the Euclidean distance ‖ X of cluster centre pointk-ti(n) ‖ (k=1,
2,3 ..., N), N is total sample number；
4th step, Similarity matching；Enable i*The subscript for representing competition triumph hidden node, to each input sample XkAccording to it
Determine that it sorts out i with the minimum euclidean distance of cluster centre*(Xk), that is, work as i*(Xk)=min ‖ Xk-ti(n) when ‖, XkIt is classified as
i*Class, to which whole samples are divided into I subset U1(n),U2(n),…,UI(n), each subset constitutes one in cluster
The heart is the Clustering Domain of Typical Representative；
5th step updates all kinds of cluster centres；It is adjusted using competition learning rule
η is Learning Step, 0<η<1, n values are added 1, go to third step；
It repeats third to walk to the 5th step, until all cluster centres meet formula：|ti(n+1)-ti(n)|<ε, ε are setting
Threshold value, indicate work as cluster centre tiVariation be less than this threshold value when no longer update, at this point, obtained tiAs radial base
The final Basis Function Center of neural network；
6th step solves width csj；Kernel function using Gaussian function as radial base neural net, then width csjIt can be by
Equations：In formula, CmaxMaximum distance between selected center；
7th step, using the connection weight of least square method calculating hidden layer to output layer neuron, calculation formula is as follows：
Gaussian kernel function is expressed as by the 8th step according to the cluster centre and width acquired above：
9th step, it follows that the output of radial base neural net is：
In formula, ωiConnection weight i.e. between hidden layer and output layer, yiIndicate input sample XkThe reality output of network.
Compared with prior art, system of the invention acquires the number of fluorescence temperature (FBG) demodulator and insulating box by master controller
According to being compared, and contrast signal is forwarded to PC machine, data are handled in PC machine, finally obtains and meets production requirement
Radial base neural net, and be burned onto in fluorescence temperature (FBG) demodulator, this system can realize that datamation acquires, program batch
Amount is downloaded, and test of product performance is accomplished scale production, and is solved calibration process and is needed to expend a large amount of manpower financial capacity's material resources, produces
Inefficiency, problem with high costs.
The method of the present invention has very strong non-linear mapping capability by radial base neural net, can map arbitrary multiple
Miscellaneous non-linear relation, solution mathematical models are difficult to the non-exponential variation that accurate description phosphorescence afterglow curve shows and ask
Topic, and the fault-tolerant ability of radial base neural net and self-learning capability can effectively overcome different fluorescent probes repetition calibration to ask
Topic realizes probe non-calibrating, interchangeable；The used radial base neural net of the present invention avoids being absorbed in compared with BP neural network
Local minimum, faster, anti-interference ability is stronger for convergence rate, approximation accuracy higher；Compared with wavelet neural network, it can keep away
Exempt from the number of hidden nodes to be difficult to determine that problem, design are more convenient.
Description of the drawings
Fig. 1 is the system principle diagram of the present invention；
Fig. 2 is the master controller functional block diagram of the present invention；
Fig. 3 is the fluorescence temperature (FBG) demodulator functional block diagram of the present invention；
In figure：1- insulating boxs, 2- fluorescence temperature (FBG) demodulators, 21- fluorescent probes, 22- optical systems, 23- fluorescence signals are visited
Survey module, 24- data processing modules, 25- excitation light drive modules, 26- control modules, 27- cd-rom recorder interfaces, 28- main controls
Device interface, 29- displays, 3- master controllers, 31- insulating box interfaces, 32- fluorescence temperature (FBG) demodulator interfaces, 33-CPU, 34- are aobvious
Show device interface, 35-PC machine interfaces, 4-PC machines, 5- cd-rom recorders, 6- cd-rom recorder seats.
Specific implementation mode
The invention will be further described below in conjunction with the accompanying drawings.
Referring to Fig. 1, the fluorescence temperature demodulating system based on radial base neural net, including master controller 3, master controller 3
Connection fluorescence temperature (FBG) demodulator 2, insulating box 1 and PC machine, fluorescence temperature (FBG) demodulator 2 and PC machine are all connected with CD writers 5, fluorescence temperature
(FBG) demodulator 2 connects insulating box 1, if fluorescence temperature (FBG) demodulator 2 accesses 1 CD writers 5 of insulating box by fluorescent probe 21 connects dry combustion method
Record engine base 6；
Master controller 3 is for acquiring fluorescence temperature (FBG) demodulator 2 and the data of insulating box 1 are compared, and by contrast signal
It is forwarded to PC machine；
PC machine is used to screen qualified radial base neural net, and by CD writers burning in fluorescence temperature (FBG) demodulator 2.
Referring to Fig. 2, master controller 3 includes control module 26, has the burning for connecting cd-rom recorder 5 in control module 26
Device interface 27 and Host Controler Interface 28 for connecting PC machine, control module 26 connect optical system 22 by two paths of signals,
Control module 26 is passed through by connecting optical system 22, control module 26 after the excitation light drive module 25 that is arranged on signal all the way
Optical system 22, light are connected after the data processing module 24 and fluorescence signal detecting module 23 that are disposed on another way signal
System 22 connects the fluorescent probe 21 for being connect with insulating box 1, and control module 26 connects display 29.
Referring to Fig. 3, fluorescence temperature (FBG) demodulator 2 includes CPU 33, has insulating box interface 31, fluorescence temperature solution on CPU 33
Adjust instrument interface 32, display interface device 34 and PC machine interface 35.
Master controller CPU uses STM32F103C8T6.Master controller 3 passes through serial bus interface and RS485 conversion circuits
Realization is connected with insulating box 1, fluorescence temperature (FBG) demodulator 2 and PC machine 4.
The working method of fluorescence temperature demodulating system based on radial base neural net, including off-line training step and online
Choosing carrys out the stage；
Wherein, off-line training step includes the following steps：
Step 1 sets 3 operating mode of master controller to off-line training, the minimum acquisition temperature of setting in master controller 3
Degree, highest collecting temperature and temperature acquisition interval, when insulating box 1 reaches assigned temperature, master controller 3 send acquisition to
Fluorescence temperature (FBG) demodulator 2 carries out data acquisition, chooses two sections that time interval in phosphorescence afterglow curve is Δ t, seeks voltage respectively
Signal value cumulative and, using the two values as the input of neural network, PC machine is sent to reference temperature composition data collection, guarantor
Deposit data, and the temperature acquisition interval set according to main controller, it is automatic to carry out acquisition next time until terminating；
Between be divided into two sections of decay curves of Δ t, first starting point time is pulse termination time t1, it is assumed that afterglow intensity
For I0, then it is I that second starting point, which is afterglow intensity value,0T at the time of correspondence when/e2, Δ t values are 5 to 20 sampled points；
Collected data set is done normalized by step 2；
Step 3 is used as test set by the 80% of data set as training set and 20%；
Training set three layers of radial base neural net of input are trained, and preserve number of results by step 4, initiation parameter
According to；
The training process of radial base neural net includes three aspects：Hidden layer Gaussian kernel is determined according to all input samples
The central value t of functioni, width csj, on the basis of determining implicit layer parameter, hidden layer is sought to defeated using least square method
Go out the connection weight ω between layeri, it is as follows：
4.1, the central value t of radial basis function is sought using k-means clustering algorithmsi；ti(n) nth iteration i-th is indicated
A cluster centre, wherein i=1,2,3 ..., I, I value are rule of thumb chosen or randomly choosed；
4.2, initialization；Select I mutually different vectors as initial cluster center ti(O) (i=1,2,3 ..., I)；
4.3, calculate each sample point X of the input spacekWith the Euclidean distance ‖ X of cluster centre pointk-ti(n) ‖ (k=1,2,
3 ..., N), N is total sample number；
4.4, Similarity matching；Enable i*The subscript for representing competition triumph hidden node, to each input sample XkAccording to its with it is poly-
The minimum euclidean distance at class center determines that it sorts out i*(Xk), that is, work as i*(Xk)=min ‖ Xk-ti(n) when ‖, XkIt is classified as i-th*Class,
To which whole samples are divided into I subset U1(n),U2(n),…,UI(n), each subset constitutes one using cluster centre as allusion quotation
The Clustering Domain that type represents；
4.5, update all kinds of cluster centres；It is adjusted using competition learning rule
η is Learning Step, 0<η<1, n values are added 1, go to third step；
It repeats third to walk to the 5th step, until all cluster centres meet formula：|ti(n+1)-ti(n)|<ε, ε are setting
Threshold value, indicate work as cluster centre tiVariation be less than this threshold value when no longer update, at this point, obtained tiAs radial base
The final Basis Function Center of neural network；
4.6, solve width csj；Kernel function using Gaussian function as radial base neural net, then width csjIt can be by formula
It solves：In formula, CmaxMaximum distance between selected center；
4.7, using the connection weight of least square method calculating hidden layer to output layer neuron, calculation formula is as follows：
4.8, according to the cluster centre and width acquired above, gaussian kernel function is expressed as：
4.9, it follows that the output of radial base neural net is：
In formula, ωiConnection weight i.e. between hidden layer and output layer, yiIndicate input sample XkThe reality output of network.
Step 5 tests the radial base neural net after training using test set；
Step 6 is downloaded the trained radial base neural net of PC machine by cd-rom recorder (5) if meeting required precision
Into CPU used in fluorescence temperature (FBG) demodulator 2；
If being unsatisfactory for required precision, increase the data of acquisition, repartition training set and test set, repeats step 3 extremely
Step 6 is until meeting the requirements；
Carry out the stage in line selection to include the following steps：
The operating mode of step 1, setting master controller 2 is product test, and temperature measurement accuracy is arranged；
Step 2, fluorescence temperature (FBG) demodulator 2 acquire the data of insulating box 1；
Step 3 passes through the radial base nerve after training using real-time data collection as the input of radial base neural net
Network obtains real time temperature；
Step 4, master controller 3 judge whether 1 temperature of insulating box is fair in error with 2 temperature measurement accuracy of fluorescence temperature (FBG) demodulator
Perhaps it in range, meets the requirements, then fluorescence temperature (FBG) demodulator meets the service check before manufacture；It is undesirable, then it needs to product
The links of production are examined again, comprehensive descision.
Claims (6)
1. the fluorescence temperature demodulating system based on radial base neural net, which is characterized in that including master controller (3), main control
Device (3) connection fluorescence temperature (FBG) demodulator (2), insulating box (1) and PC machine (4), fluorescence temperature (FBG) demodulator (2) and PC machine are all connected with burning
Record machine (5), fluorescence temperature (FBG) demodulator (2) connect insulating box (1), and CD writers (5) connect several burning engine bases (6)；
Master controller (3) is used to acquire fluorescence temperature (FBG) demodulator (2) and the data of insulating box (1) are compared, and comparison is believed
Number it is forwarded to PC machine (4)；
PC machine (4) is used to screen qualified radial base neural net, and by CD writers burning in fluorescence temperature (FBG) demodulator (2)
In.
2. a kind of fluorescence temperature demodulating system based on radial base neural net according to claim 1, which is characterized in that
Fluorescence temperature (FBG) demodulator (2) accesses insulating box (1) by fluorescent probe (21).
3. a kind of fluorescence temperature demodulating system based on radial base neural net according to claim 1, which is characterized in that
Master controller (3) includes control module (26), has the cd-rom recorder interface for connecting cd-rom recorder (5) in control module (26)
(27) and the Host Controler Interface (28) for connecting PC machine (4), control module (26) connect optical system by two paths of signals
(22), control module (26) connects optical system (22) afterwards by the excitation light drive module (25) being arranged on signal all the way, control
Molding block (26) passes through the data processing module (24) and fluorescence signal detecting module (23) that are disposed on another way signal
Connection optical system (22) afterwards, optical system (22) connect the fluorescent probe (21) for being connect with insulating box (1), control module
(26) connection display (29).
4. a kind of fluorescence temperature demodulating system based on radial base neural net according to claim 1, which is characterized in that
Fluorescence temperature (FBG) demodulator (2) includes (33) CPU, has insulating box interface (31), fluorescence temperature (FBG) demodulator interface on CPU (33)
(32), display interface device (34) and PC machine interface (35).
5. a kind of working method of fluorescence temperature demodulating system based on radial base neural net described in claim 1, special
Sign is, including off-line training step and carrys out the stage in line selection；
Off-line training step includes the following steps：
Step 1 sets master controller (3) operating mode to off-line training, the minimum acquisition temperature of setting in master controller (3)
Degree, highest collecting temperature and temperature acquisition interval, when insulating box (1) reaches assigned temperature, master controller (3) sends acquisition life
It enables fluorescence temperature (FBG) demodulator (2) carry out data acquisition, chooses two sections that time interval in phosphorescence afterglow curve is Δ t, respectively
It asks the cumulative of voltage signal values and using the two values as the input of neural network, is sent to reference temperature composition data collection
PC machine (4) preserves data, and the temperature acquisition interval set according to main controller, automatic to carry out acquisition next time until terminating；
Collected data set is done normalized by step 2；
Step 3, using a part for the data after normalized as training set and another part as test set；
Training set three layers of radial base neural net of input are trained, and preserve result data by step 4, initiation parameter；
Step 5 tests the radial base neural net after training using test set；
Step 6 is downloaded the trained radial base neural net of PC machine (4) by cd-rom recorder (5) if meeting required precision
Into CPU used in fluorescence temperature (FBG) demodulator (2)；
If being unsatisfactory for required precision, increase the data of acquisition, repartition training set and test set, repeats step 3 to step
Six until meeting the requirements；
Carry out the stage in line selection to include the following steps：
The operating mode of step 1, setting master controller (2) is product test, and temperature measurement accuracy is arranged；
Step 2, fluorescence temperature (FBG) demodulator (2) acquire the data of insulating box (1)；
Step 3 passes through the radial base neural net after training using real-time data collection as the input of radial base neural net
Obtain real time temperature；
Whether step 4, master controller (3) judge insulating box (1) temperature with fluorescence temperature (FBG) demodulator (2) temperature measurement accuracy in error
In allowable range, service check is completed.
6. a kind of working method of fluorescence temperature demodulating system based on radial base neural net according to claim 5,
It is characterized in that, the specific method is as follows for radial base neural net training：
The first step seeks the central value t of radial basis function using k-means clustering algorithmsi；ti(n) nth iteration is indicated i-th
Cluster centre, wherein i=1,2,3 ..., I, I value are rule of thumb chosen or randomly choosed；
Second step, initialization；Select I mutually different vectors as initial cluster center ti(O) (i=1,2,3 ..., I)；
Third walks, and calculates each sample point X of the input spacekWith the Euclidean distance ‖ X of cluster centre pointk-ti(n) ‖ (k=1,2,
3 ..., N), N is total sample number；
4th step, Similarity matching；Enable i*The subscript for representing competition triumph hidden node, to each input sample XkAccording to its with it is poly-
The minimum euclidean distance at class center determines that it sorts out i*(Xk), that is, work as i*(Xk)=min ‖ Xk-ti(n) when ‖, XkIt is classified as i-th*Class,
To which whole samples are divided into I subset U1(n),U2(n),…,UI(n), each subset constitutes one using cluster centre as allusion quotation
The Clustering Domain that type represents；
5th step updates all kinds of cluster centres；It is adjusted using competition learning rule
η is Learning Step, 0<η<1, n values are added 1, go to third step；
It repeats third to walk to the 5th step, until all cluster centres meet formula：|ti(n+1)-ti(n)|<ε, ε are the threshold of setting
Value indicates to work as cluster centre tiVariation be less than this threshold value when no longer update, at this point, obtained tiAs radial base nerve
The final Basis Function Center of network；
6th step solves width csj；Kernel function using Gaussian function as radial base neural net, then width csjIt can be by formula
It solves：In formula, CmaxMaximum distance between selected center；
7th step, using the connection weight of least square method calculating hidden layer to output layer neuron, calculation formula is as follows：
Gaussian kernel function is expressed as by the 8th step according to the cluster centre and width acquired above：
9th step, it follows that the output of radial base neural net is：
In formula, ωiConnection weight i.e. between hidden layer and output layer, yiIndicate input sample XkThe reality output of network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810344413.XA CN108415336A (en) | 2018-04-17 | 2018-04-17 | Fluorescence temperature demodulating system based on radial base neural net and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810344413.XA CN108415336A (en) | 2018-04-17 | 2018-04-17 | Fluorescence temperature demodulating system based on radial base neural net and method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108415336A true CN108415336A (en) | 2018-08-17 |
Family
ID=63134136
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810344413.XA Pending CN108415336A (en) | 2018-04-17 | 2018-04-17 | Fluorescence temperature demodulating system based on radial base neural net and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108415336A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN202329866U (en) * | 2011-09-23 | 2012-07-11 | 苏州光格设备有限公司 | Fluorescent optical fiber temperature sensing demodulation instrument |
| CN103336697A (en) * | 2013-05-30 | 2013-10-02 | 杭州士兰控股有限公司 | Burning system and burning method |
| CN104729965A (en) * | 2015-01-28 | 2015-06-24 | 东北大学 | PM2.5 concentration detection method based on interzone radial basis function nerve network |
| CN208156464U (en) * | 2018-04-17 | 2018-11-27 | 江苏博亨光电科技有限公司 | Fluorescence temperature demodulating system based on radial base neural net |
-
2018
- 2018-04-17 CN CN201810344413.XA patent/CN108415336A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN202329866U (en) * | 2011-09-23 | 2012-07-11 | 苏州光格设备有限公司 | Fluorescent optical fiber temperature sensing demodulation instrument |
| CN103336697A (en) * | 2013-05-30 | 2013-10-02 | 杭州士兰控股有限公司 | Burning system and burning method |
| CN104729965A (en) * | 2015-01-28 | 2015-06-24 | 东北大学 | PM2.5 concentration detection method based on interzone radial basis function nerve network |
| CN208156464U (en) * | 2018-04-17 | 2018-11-27 | 江苏博亨光电科技有限公司 | Fluorescence temperature demodulating system based on radial base neural net |
Non-Patent Citations (3)
| Title |
|---|
| 蔡兵;: "基于RBF神经网络的热电偶建模方法", 微计算机信息, no. 24 * |
| 陈凯;张倩怡;殷志国;: "基于RBF网络的传动滚筒测温系统", 仪表技术与传感器, no. 09 * |
| 高秀娟;: "RBF网络在热敏电阻测温中的应用", 仪器仪表用户, no. 04, pages 136 - 137 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110415215A (en) | Intelligent detecting method based on figure neural network | |
| CN105259215B (en) | The method of testing of semiconductor gas sensor | |
| CN109165504A (en) | A kind of electric system false data attack recognition method generating network based on confrontation | |
| CN111628494B (en) | Low-voltage distribution network topology identification method and system based on logistic regression method | |
| CN105678344A (en) | Intelligent classification method for power instrument equipment | |
| CN107423754B (en) | Automatic radiation source identification system based on parameter multi-attribute autonomous intelligent decision | |
| CN108549875A (en) | A kind of brain electricity epileptic attack detection method based on the perception of depth channel attention | |
| CN110243595B (en) | Long-range gear box fault monitoring system based on LabVIEW | |
| CN108647642A (en) | Multisensor Crack Damage error comprehensive diagnosis method based on fuzzy Fusion | |
| CN109840671A (en) | Operational development effect calculates equipment, operational development effect calculation method and recording medium | |
| Selivanova et al. | Simulation of intelligent information-measuring systems of thermophysical properties of materials and products | |
| CN208156464U (en) | Fluorescence temperature demodulating system based on radial base neural net | |
| CN110874685A (en) | Intelligent electric energy meter running state distinguishing method and system based on neural network | |
| CN115327270A (en) | Transformer small sample fault diagnosis system based on data-model driving | |
| CN109297534A (en) | For evaluating the environmental parameter Weight Determination and system of indoor environmental quality | |
| CN113850154A (en) | Inverter IGBT (insulated Gate Bipolar transistor) micro fault feature extraction method based on multi-modal data | |
| CN113489514B (en) | Power line communication noise identification method and device based on self-organizing mapping neural network | |
| CN110889207A (en) | System combination model credibility intelligent evaluation method based on deep learning | |
| Yan et al. | Online battery health diagnosis for electric vehicles based on DTW-XGBoost | |
| CN108415336A (en) | Fluorescence temperature demodulating system based on radial base neural net and method | |
| CN108563201A (en) | A kind of parts in small batch machining process quality improvement method of DMAIC drivings | |
| CN106228033B (en) | Three-core cable conductor temperature real-time computing technique based on RBF neural | |
| CN113191075B (en) | Photovoltaic array fault diagnosis method based on improved goblet sea squirt group algorithm | |
| CN109656229A (en) | The construction method of robot end's performance prediction model based on GA-RBF network | |
| CN105699043B (en) | A kind of wind tunnel sensors that improve measure stability and the method for precision |
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 |