CN112798142B - Brillouin optical fiber sensor strain and temperature two-stage rapid identification method based on Bayesian updating and random simulation - Google Patents
Brillouin optical fiber sensor strain and temperature two-stage rapid identification method based on Bayesian updating and random simulation Download PDFInfo
- Publication number
- CN112798142B CN112798142B CN202011589241.6A CN202011589241A CN112798142B CN 112798142 B CN112798142 B CN 112798142B CN 202011589241 A CN202011589241 A CN 202011589241A CN 112798142 B CN112798142 B CN 112798142B
- Authority
- CN
- China
- Prior art keywords
- parameter
- optical fiber
- gain spectrum
- theta
- brillouin optical
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01K—MEASURING TEMPERATURE; MEASURING QUANTITY OF HEAT; THERMALLY-SENSITIVE ELEMENTS NOT OTHERWISE PROVIDED FOR
- G01K11/00—Measuring temperature based upon physical or chemical changes not covered by groups G01K3/00, G01K5/00, G01K7/00 or G01K9/00
- G01K11/32—Measuring temperature based upon physical or chemical changes not covered by groups G01K3/00, G01K5/00, G01K7/00 or G01K9/00 using changes in transmittance, scattering or luminescence in optical fibres
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
- G01B11/00—Measuring arrangements characterised by the use of optical techniques
- G01B11/16—Measuring arrangements characterised by the use of optical techniques for measuring the deformation in a solid, e.g. optical strain gauge
- G01B11/18—Measuring arrangements characterised by the use of optical techniques for measuring the deformation in a solid, e.g. optical strain gauge using photoelastic elements
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Optical Transform (AREA)
- Testing Of Optical Devices Or Fibers (AREA)
- Length Measuring Devices By Optical Means (AREA)
Abstract
The invention provides a Brillouin optical fiber sensor strain and temperature two-stage quick identification method based on Bayes updating and random simulation, which adopts a quick frequency sweep mode with large range, wide space and low average times to obtain a rough Brillouin gain spectrum, obtains a large number of samples which are subjected to joint distribution of each parameter in the Brillouin gain spectrum by a parameter identification method based on the Bayes updating and random simulation methods, statistically estimates the upper and lower limits of the frequency sweep range of the next stage, performs intensive frequency sweep by using a smaller range to further obtain an identification result of the parameter of the Brillouin optical fiber gain spectrum with sufficient accuracy, and finally obtains strain or temperature information of a structure according to the function relationship of the frequency shift of the Brillouin optical fiber and the structural strain and temperature. The method can quickly and accurately process the data obtained by the Brillouin optical fiber sensor, the result is the mean value and standard deviation estimation value of the structural strain or temperature of each measuring point of the Brillouin optical fiber, and the method has good reference value for practical engineering application.
Description
Technical Field
The invention belongs to the technical field of Brillouin optical fiber sensing and structural health monitoring, and particularly relates to a two-stage rapid strain and temperature identification method of a Brillouin optical fiber sensor based on Bayesian updating and random simulation.
Background
In the modern times, with the rapid development of China, large-scale infrastructures such as bridges, tunnels, oil and gas pipelines and high-speed rail tracks are rapidly constructed, and the construction of a large number of projects also brings about the problems of structural safety and reliability. For such engineering situations, structural health monitoring becomes very important as structures suffer from different degrees of damage. Structural health monitoring often needs to acquire strain and temperature information of a structure, and a distributed Brillouin optical fiber sensing technology can provide an effective solution for strain and temperature monitoring. Distributed brillouin fiber sensing is also becoming a new direction of research as infrastructure construction becomes larger and more complex. However, the problem of the testing efficiency of the distributed brillouin optical fiber strain and temperature tester is still to be solved, and a new research direction of the distributed brillouin optical fiber sensor in recent years is formed.
In the process of measuring the strain or temperature of the sensing optical fiber by using the traditional distributed optical fiber strain tester, the strain or temperature distribution range of the optical fiber to be tested is uncertain, so that the detection light needs to be subjected to large-range intensive continuous frequency sweeping and multiple averaging. The method obtains a large amount of useless information, consumes a large amount of test time and reduces the test efficiency of the instrument.
Disclosure of Invention
The invention aims to solve the problems that in the process of measuring the strain or the temperature of a sensing optical fiber, the traditional distributed optical fiber strain and temperature tester needs to carry out large-range intensive continuous frequency sweeping and multiple averaging on detection light due to uncertain strain or temperature distribution range of the optical fiber to be measured, so that a large amount of useless information can be obtained in the process, more time can be consumed, and the efficiency is lower. Aiming at the defects, the invention provides a two-stage rapid identification method of strain and temperature of a Brillouin optical fiber sensor based on Bayesian updating and random simulation, and the method is suitable for relevant test objects in the field of structural health monitoring.
The invention is realized by the following technical scheme, and provides a Brillouin optical fiber sensor strain and temperature two-stage rapid identification method based on Bayesian updating and random simulation, which comprises the following steps:
firstly, laying an optical fiber sensor on a test object, and carrying out sparse and uniform frequency sweeping on a test point to obtain corresponding measurement data;
secondly, performing Brillouin optical fiber gain spectrum parameter identification on the data obtained in the first step based on a Bayesian updating and random simulation method to obtain a plurality of samples which obey joint distribution of each parameter of the Brillouin optical fiber gain spectrum, and representing the joint probability distribution condition of each parameter;
step three, based on the multiple samples which are obtained in the step two and obey joint distribution of parameters of the Brillouin optical fiber gain spectrum, counting and estimating the upper limit and the lower limit of the frequency sweep range of the next stage, and respectively calculating vB+αΔvBAnd vB-αΔvBThe median of (2) is taken as the next stage of each pointUpper and lower limits of the sweep range, wherein the center frequency vBFull width at half maximum Δ vBThe coefficient alpha is more than or equal to 0.5 and less than 0.8;
fourthly, carrying out intensive frequency sweeping according to the frequency sweeping range obtained in the third step to obtain corresponding measurement data;
fifthly, carrying out Brillouin optical fiber gain spectrum parameter identification on the data obtained in the fourth step by using Bayes updating and random simulation methods to obtain a plurality of samples which are subjected to joint distribution of parameters of the Brillouin optical fiber gain spectrum, and statistically determining the center frequency v of the parameters of the Brillouin optical fiber gain spectrumBThe result of the recognition of (1);
and step six, calculating strain or temperature information of the test object according to the function relation between the Brillouin optical fiber frequency shift and the strain and temperature of the test object based on the identification result obtained in the step five, and using the strain or temperature information to further analyze the test object.
Further, the second step is specifically:
the second step is a process of identifying the gain spectrum parameters of the Brillouin optical fiber based on Bayesian updating and random simulation methods, wherein the process is a process of simulating a plurality of parameter sample points which obey the posterior probability distribution of the gain spectrum parameters according to the prior probability distribution f (theta) of the gain spectrum parameters obtained by known information and the likelihood function f (Q | theta) obtained by current new measurement data Q based on Bayesian theorem, wherein theta is the gain spectrum parameter; the process specifically comprises the following steps:
step 2.1, setting prior probability distribution f (theta) of gain spectrum parameters: if the adjacent measuring points of the current measuring point are not identified, setting the prior probability distribution of the gain spectrum parameters as uniform distribution in the measuring range of the optical fiber sensor; if the adjacent measuring points of the current measuring point are identified, setting the prior probability distribution of the gain spectrum parameters as truncated Gaussian distribution;
step 2.2, randomly generating N parameter sample points theta obeying the prior probability distribution f (theta) of the gain spectrum parameterj,kK is 1, …, and N, j is the cycle number, and the annealing index P is setj＝0；
Step 2.3, calculating to obtain a sample point theta based on each parameterj,kLikelihood ofFunction f (Q | theta)j,k) Using f (Q | theta)j,k) Determination of the Next annealing index Pj+1Calculating the resampling weight probability
Is calculated by the formula
If Pj+1If not less than 1, entering the step 2.8, otherwise entering the step 2.4;
step 2.4, calculating parameter update covariance ∑jFor updating the parameter and preventing the parameter sample from being depleted in the resampling process, the calculation formula is
Beta is a constant;
step 2.5, based on the result of step 2.3, probability by weight
For parameter sample point thetaj,kResampling to obtain a new group of samples
k is 1, …, N, so that the parameter distribution is continuously close to the posterior probability distribution of the parameter with the increase of the annealing index;
step 2.6, based on the result of step 2.4, performing a Markov chain Monte Carlo parameter updating process to obtain a new set of samples thetaj+1,k，k＝1,…,N；
Step 2.7, making j equal to j +1, and returning to the step 2.3;
step 2.8, let Pj+1And (3) executing the steps 2.4 to 2.6 to obtain N parameter samples obeying the parameter posterior probability distribution, wherein the N parameter samples are used for representing the parameter probability distribution of the Brillouin optical fiber gain spectrum.
Further, the step 2.3 specifically includes:
step 2.3.1, setting initial value Pj+1＝(Pj+1)/2, the calculation is based on different target parameters θj,kWeight w ofj,k；
Step 2.3.2, calculate wj,kK is 1, …, and the coefficient of variation of N is P with the coefficient of variation of 100%j+1Is optimal;
step 2.3.3, calculating by dichotomy to finally obtain Pj+1。
Further, the step 2.6 specifically includes:
step 2.6.1, from
Updating covariance Σ for mean and parametersjExtracting candidate samples theta for Gaussian distribution of covariancer；
Step 2.6.2, calculate based on thetarPrior probability distribution f (θ)r) And likelihood function f (Q | theta)r) And is combined with the original sample
Comparing to obtain the acceptance probability r, wherein the calculation formula is
Step 2.6.3, press [0, 1 ]]Generates a random number u, accepts the candidate sample if r is greater than ur，θj+1,k＝θrOtherwise, reject the candidate sample θr，
The invention has the beneficial effects that:
1. compared with the traditional method, the method greatly reduces the data quantity required to be obtained by the optical fiber sensor, and can effectively improve the identification speed of the Brillouin optical fiber sensor on the structural strain or temperature;
2. the obtained result is a digital characteristic estimation value of probability distribution such as the mean value, the standard deviation and the like of structural strain or temperature at each measuring point of the Brillouin optical fiber, and the digital characteristic estimation value has good reference value for practical engineering application;
3. the Bayesian updating and random simulation method has the advantages that the parameter distribution of the Brillouin optical fiber gain spectrum can be subjected to robust analysis based on a small amount of data, a large number of samples which obey the joint distribution of each parameter of the gain spectrum can be obtained, and the information of the correlation relationship among the parameters can be naturally utilized when the frequency sweeping range of the second stage is determined;
4. the sweep frequency range of the second stage adopts a median value, and the median value is slightly influenced by outlier samples, so that the method has higher robustness.
Drawings
FIG. 1 is a flow chart of a two-stage rapid identification method of strain and temperature of a Brillouin optical fiber sensor based on Bayesian update and random simulation in the invention;
FIG. 2 is a flow chart of a parameter identification process based on Bayesian update and stochastic simulation methods in accordance with the present invention;
fig. 3 is a schematic diagram of sweep data for a specific position on an optical fiber sensor according to the present invention: wherein, (a) is the sweep frequency data of the specific position required by the large-range intensive continuous sweep frequency of the traditional method, (b) is the sweep frequency data of the specific position required by the large-interval uniform sweep frequency of the first stage of the invention, and (c) is the sweep frequency data of the specific position required by the small-range intensive sweep frequency of the second stage of the invention;
fig. 4 is a schematic diagram of the result of fast strain identification of the brillouin optical fiber sensor structure for a section of optical fiber sensor arranged on a stressed member according to the present invention: wherein (a) is frequency sweep data obtained by the traditional method for visually comparing with the obtained result, and (b) is a gain spectrum parameter v determined by the large-space uniform frequency sweep dataBAn estimated value (red curve) and a dense sweep frequency range (blue curve), (c) small-range dense sweep frequency data obtained according to the dense sweep frequency range, and (d) a gain spectrum parameter v determined according to the small-range dense sweep frequency dataBThe estimated value (red curve) and its three sigma confidence interval (blue curve), (e) the spectral parameter v according to the inventionBDerived from the result of the recognitionThe structural strain estimate (red curve) and its three sigma confidence interval (blue curve) are plotted in meters of length, megahertz of frequency, and microstrain of strain.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In order to improve the testing efficiency of the distributed optical fiber strain and temperature tester, the invention adopts a two-stage detection light frequency sweeping strategy. In the first stage, the detection light is controlled to adopt a fast frequency sweeping mode with large range, wide spacing and low average times to obtain a rough Brillouin gain spectrum, and the reasonable range of frequency sweeping in the next stage is determined according to a fitting result by fitting the probability distribution of gain spectrum parameters through a Lorentz line type. In the second stage, according to the test result in the first stage, the detection light is controlled to be locked in the reasonable range near the Brillouin frequency shift of the tested optical fiber, the accurate Brillouin gain spectrum of the tested optical fiber is obtained by adopting a frequency sweeping mode with a small range and a narrow space, and the central frequency v of the Brillouin gain spectrum is extracted through Lorentz linear fittingBAnd further calculating strain or temperature information of the structure according to the function relation between the Brillouin optical fiber frequency shift and the strain and temperature of the structure.
For such a two-stage probe light frequency sweep strategy, a better identification method is needed to determine the center frequency v of the brillouin gain spectrumBAnd other parameters. Based on the above requirements, the present invention employs bayesian updating and random simulation methods. Bayesian updating is based on Bayesian theorem, existing measurement information and current new measurement data can be fully utilized, and posterior probability distribution of Brillouin gain spectrum parameters, namely relative probability of all possible values of the parameters based on data, is calculated. The stochastic simulation method uses Markov chain Monte Carlo (Markov chain Monte Carlo) methodAccording to the method, a large number of samples which obey the parameter distribution are generated based on the posterior probability distribution of the Brillouin gain spectrum parameters obtained by Bayes updating, and then the digital characteristics such as the mean value, the standard deviation and the like of the parameter distribution are rapidly estimated according to actual needs, so that the requirements of the two-stage detection light frequency sweeping strategy are met.
With reference to fig. 1 to 4, the present invention provides a strain and temperature two-stage fast identification method for a brillouin optical fiber sensor based on bayes update and random simulation, wherein the method includes the following steps:
step one, laying an optical fiber sensor on a test object, and carrying out sparse (sweep interval is dozens of megahertz) uniform sweep frequency on a measuring point to obtain corresponding measurement data;
secondly, performing Brillouin optical fiber gain spectrum parameter identification on the data obtained in the first step based on a Bayesian updating and random simulation method to obtain a plurality of samples which obey joint distribution of each parameter of the Brillouin optical fiber gain spectrum, and representing the joint probability distribution condition of each parameter;
step three, based on the multiple samples which are obtained in the step two and obey joint distribution of parameters of the Brillouin optical fiber gain spectrum, counting and estimating the upper limit and the lower limit of the frequency sweep range of the next stage, and respectively calculating vB+αΔvBAnd vB-αΔvBThe median of (a) is used as the upper and lower limits of the sweep range of the next stage at each point, wherein the central frequency vBFull width at half maximum Δ vBThe coefficient alpha is more than or equal to 0.5 and less than 0.8;
fourthly, carrying out intensive frequency sweeping according to the frequency sweeping range obtained in the third step to obtain corresponding measurement data;
fifthly, carrying out Brillouin optical fiber gain spectrum parameter identification on the data obtained in the fourth step by using Bayes updating and random simulation methods to obtain a plurality of samples which are subjected to joint distribution of parameters of the Brillouin optical fiber gain spectrum, and statistically determining the center frequency v of the parameters of the Brillouin optical fiber gain spectrumBThe result of the recognition of (1);
and step six, calculating strain or temperature information of the test object according to the function relation between the Brillouin optical fiber frequency shift and the strain and temperature of the test object based on the identification result obtained in the step five, and using the strain or temperature information to further analyze the test object.
The second step is specifically as follows:
the second step is a process of identifying the gain spectrum parameters of the Brillouin optical fiber based on Bayesian updating and random simulation methods, wherein the process is a process of simulating a plurality of parameter sample points obeying the posterior probability distribution (target distribution) of the gain spectrum parameters according to the prior probability distribution f (theta) of the gain spectrum parameters obtained by known information and the likelihood function f (Q | theta) obtained by the current new measurement data Q based on Bayesian theorem, wherein theta is the gain spectrum parameters; the process specifically comprises the following steps:
step 2.1, setting prior probability distribution f (theta) of gain spectrum parameters: if the adjacent measuring points of the current measuring point are not identified, setting the prior probability distribution of the gain spectrum parameters as uniform distribution in the measuring range of the optical fiber sensor; if the adjacent measuring points of the current measuring point are identified, setting the prior probability distribution of the gain spectrum parameters as truncated Gaussian distribution; the mean value of Gaussian distribution is the identification result of adjacent measuring points, the variance of Gaussian distribution is taken as a value according to the characteristics of the sensor and is usually set as a larger value, and the boundary of truncated Gaussian distribution is the measurement range of the optical fiber sensor;
step 2.2, randomly generating a large number (N) of parameter sample points theta obeying the prior probability distribution f (theta) of the gain spectrum parameterj,kK is 1, …, and N, j is the cycle number, and the annealing index P is setj＝0；
Step 2.3, calculating to obtain a sample point theta based on each parameterj,kLikelihood function f (Q | theta)j,k) Using f (Q | theta)j,k) Determination of the Next annealing index Pj+1Calculating the resampling weight probability
Is calculated by the formula
If Pj+1Not less than 1, entering the step2.8, otherwise, entering a step 2.4;
step 2.4, calculating parameter update covariance ∑jFor updating the parameter and preventing the parameter sample from being depleted in the resampling process, the calculation formula is
Beta is a constant;
step 2.5, based on the result of step 2.3, probability by weight
For parameter sample point thetaj,kResampling to obtain a new group of samples
k is 1, …, N, so that the parameter distribution is continuously close to the posterior probability distribution of the parameter with the increase of the annealing index;
step 2.6, based on the result of step 2.4, performing a Markov chain Monte Carlo parameter updating process to obtain a new set of samples thetaj+1,k，k＝1,…,N；
Step 2.7, making j equal to j +1, and returning to the step 2.3;
step 2.8, let Pj+1Step 2.4 to step 2.6 are performed to obtain a large number (N) of parameter samples obeying the parameter posterior probability distribution, which is used for characterizing the probability distribution of the parameter of the brillouin optical fiber gain spectrum.
The step 2.3 is specifically as follows:
step 2.3.1, setting initial value Pj+1＝(Pj+1)/2, the calculation is based on different target parameters θj,kWeight w ofj,k；
Step 2.3.2, calculate wj,kK is 1, …, and the coefficient of variation (i.e. the ratio of the standard deviation to the mean) of N is taken as P when the coefficient of variation is 100%j+1Is optimal;
step 2.3.3, calculating by dichotomy to finally obtain Pj+1。
The following describes in detail a specific method of the markov chain monte carlo parameter update process in step 2.6:
step 2.6.1, from
Updating covariance Σ for mean and parametersjExtracting candidate samples theta for Gaussian distribution of covariancer；
Step 2.6.2, calculate based on thetarPrior probability distribution f (θ)r) And likelihood function f (Q | theta)r) And is combined with the original sample
Comparing to obtain the acceptance probability r, wherein the calculation formula is
Step 2.6.3, press [0, 1 ]]Generates a random number u, accepts the candidate sample if r is greater than ur，θj+1,k＝θrOtherwise, reject the candidate sample θr，
According to the Brillouin optical fiber sensor strain and temperature two-stage quick identification method based on Bayes updating and random simulation, the two-stage detection light frequency sweeping strategy is adopted, the robustness of parameter identification under a small amount of data by the Bayes updating method is utilized, the data volume required by the optical fiber sensor is greatly reduced, and the identification speed of the Brillouin optical fiber sensor on the structure strain or temperature can be effectively improved. For the optical fiber sensor in the structural health monitoring, the distance is long and the number of measuring points is large, the time consumption of single measurement can be greatly saved, and therefore the Brillouin optical fiber sensor can better serve the field of structural health monitoring.
Examples
With reference to fig. 4, for a section of optical fiber sensor with a length of 2m arranged on a stressed member in an experiment, the strain and temperature two-stage rapid identification method of the brillouin optical fiber sensor based on bayesian updating and random simulation is utilized to identify the structural strain of each measuring point. There were 197 stations in the range.
The strain of each measuring point is identified by using a Brillouin optical fiber sensor strain and temperature two-stage quick identification method based on Bayesian updating and random simulation in the invention:
the first step is specifically as follows: laying an optical fiber sensor on the concrete stress member, determining that the absolute position coordinate is 55.7m-57.7m, wherein 197 measuring points are in the range, carrying out large-interval uniform frequency sweeping on the measuring points, wherein the lower limit of the frequency sweeping is 10.60GHz, the upper limit of the frequency sweeping is 11.00GHz, the frequency sweeping interval is 80MHz, and obtaining corresponding measurement data, wherein each measuring point is uniformly swept for 11 times;
the second step is specifically as follows: taking the center frequency v of the Brillouin gain spectrumBFull width at half maximum Δ vBPeak g of spectrum0And noise intensity sigma is a parameter to be identified, Lorentz linear fitting is carried out, and the measurement range of the optical fiber sensor is set to be 10.6GHz < vB＜11.0GHz、30MHz＜ΔvB＜150MHz、0μW＜g0The standard deviations of Gaussian distribution in parameter prior distribution are respectively set to be 20MHz, 25MHz, 0.002 muW and 0.0002 muW when adjacent measuring points are identified, the sampling number N is 5000, and the constant beta is 0.5, the data obtained in the step one is subjected to Brillouin optical fiber gain spectrum parameter identification based on Bayesian updating and random simulation methods, a large number of samples which obey the joint distribution of parameters of a gain spectrum are obtained, and the joint probability distribution condition of the parameters is represented;
the third step is specifically as follows: based on a large number of samples which are obtained in the second step and obey joint distribution of parameters of the Brillouin optical fiber gain spectrum, taking alpha as 0.5, and respectively calculating vB+ΔvBV and 2B-ΔvBThe median of/2 is taken as the upper and lower limits of the sweep range of the next stage at each point, as shown in FIG. 4 (b);
the fourth step is specifically as follows: based on the range obtained in the third step, carrying out small-range intensive frequency sweeping with a frequency sweeping interval of 2MHz to obtain corresponding measurement data, and carrying out frequency sweeping for 29 times on average at one measurement point as shown in FIG. 4 (c);
the fifth step is specifically as follows: using the method of the second step to perform Brillouin optical fiber gain spectrum parameter identification on the data obtained in the fourth step to obtain a large number of samples which are subjected to the joint distribution of each parameter of the gain spectrum, and calculating the parameter v of the Brillouin optical fiber gain spectrum of each pointBAs a result of recognition, v is plotted as a mean and a standard deviation (sigma)BThe mean and three sigma confidence intervals along the length of the fiber, as shown in fig. 4 (d);
the sixth step is specifically as follows: the center frequency of the gain spectrum of the Brillouin optical fiber has a functional relation with structural strain and temperature: v. ofB(ε,T)＝vB(εr,Tr)+Cε(ε-εr)+CT(T-Tr) Wherein v isB(ε, T) is the current strain ε, the center frequency of the gain spectrum at the current temperature T, vB(εr,Tr) For reference to strain epsilonrReference temperature TrCenter frequency of gain spectrum at, CεAnd CTRespectively, the strain coefficient and the temperature coefficient of the optical fiber, C in this embodimentε＝0.05MHz/με，C T1 MHz/DEG C, taking the free part of the external optical fiber as reference, epsilonr＝0，TrAnd (4) calculating strain or temperature information of the structure according to the function relationship of the Brillouin optical fiber frequency shift, the structure strain and the temperature based on the identification result obtained in the fifth step, and drawing a variation curve of the structure strain mean value and the three-sigma confidence interval along the length of the optical fiber, wherein the result can be used for further analyzing the structure as shown in figure 4 (e).
In the above embodiment, the total frequency of the frequency sweep of each measuring point is on average 11+29 to 40, while the traditional method needs to sweep hundreds of times, as shown in fig. 3 (a). Compared with the traditional method, the method greatly reduces the data quantity required to be obtained by the optical fiber sensor, and can effectively improve the identification speed of the Brillouin optical fiber sensor on the structural strain or temperature. Meanwhile, the result of the method is a digital characteristic estimation value of probability distribution such as the mean value, the standard deviation and the like of structural strain or temperature at each measuring point of the Brillouin optical fiber, and the method has a good reference value for practical engineering application.
The two-stage rapid identification method of the strain and the temperature of the brillouin optical fiber sensor based on bayesian updating and random simulation, which is provided by the invention, is described in detail, a specific example is applied in the text to explain the principle and the implementation mode of the invention, and the description of the above example is only used for helping to understand the method and the core idea of the invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.
Claims (4)
1. A Brillouin optical fiber sensor strain and temperature two-stage rapid identification method based on Bayesian updating and random simulation is characterized in that: the method comprises the following steps:
firstly, laying an optical fiber sensor on a test object, and carrying out sparse and uniform frequency sweeping on a test point to obtain corresponding measurement data;
secondly, performing Brillouin optical fiber gain spectrum parameter identification on the data obtained in the first step based on a Bayesian updating and random simulation method to obtain a plurality of samples which obey joint distribution of each parameter of the Brillouin optical fiber gain spectrum, and representing the joint probability distribution condition of each parameter;
step three, based on the multiple samples which are obtained in the step two and obey joint distribution of parameters of the Brillouin optical fiber gain spectrum, counting and estimating the upper limit and the lower limit of the frequency sweep range of the next stage, and respectively calculating vB+αΔvBAnd vB-αΔvBThe median of (a) is used as the upper and lower limits of the sweep range of the next stage at each point, wherein the central frequency vBFull width at half maximum Δ vBThe coefficient alpha is more than or equal to 0.5 and less than 0.8;
fourthly, carrying out intensive frequency sweeping according to the frequency sweeping range obtained in the third step to obtain corresponding measurement data;
fifthly, carrying out parameter identification on the Brillouin optical fiber gain spectrum on the data obtained in the fourth step by using Bayes updating and random simulation methods to obtain a plurality of samples which are subjected to joint distribution of parameters of the Brillouin optical fiber gain spectrum, and statistically determining BrillouinFiber gain spectrum parameter center frequency vBThe result of the recognition of (1);
and step six, calculating strain or temperature information of the test object according to the function relation between the Brillouin optical fiber frequency shift and the strain and temperature of the test object based on the identification result obtained in the step five, and using the strain or temperature information to further analyze the test object.
2. The method of claim 1, wherein: the second step is specifically as follows:
the second step is a process of identifying the gain spectrum parameters of the Brillouin optical fiber based on Bayesian updating and random simulation methods, wherein the process is a process of simulating a plurality of parameter sample points which obey the posterior probability distribution of the gain spectrum parameters according to the prior probability distribution f (theta) of the gain spectrum parameters obtained by known information and the likelihood function f (Q | theta) obtained by current new measurement data Q based on Bayesian theorem, wherein theta is the gain spectrum parameter; the process specifically comprises the following steps:
step 2.1, setting prior probability distribution f (theta) of gain spectrum parameters: if the adjacent measuring points of the current measuring point are not identified, setting the prior probability distribution of the gain spectrum parameters as uniform distribution in the measuring range of the optical fiber sensor; if the adjacent measuring points of the current measuring point are identified, setting the prior probability distribution of the gain spectrum parameters as truncated Gaussian distribution;
step 2.2, randomly generating N parameter sample points theta obeying the prior probability distribution f (theta) of the gain spectrum parameterj,kK is 1, …, and N, j is the cycle number, and the annealing index P is setj＝0；
Step 2.3, calculating to obtain a sample point theta based on each parameterj,kLikelihood function f (Q | theta)j,k) Using f (Q | theta)j,k) Determination of the Next annealing index Pj+1Calculating the resampling weight probability
Is calculated by the formula
If Pj+1If not less than 1, entering the step 2.8, otherwise entering the step 2.4;
step 2.4, calculating parameter update covariance ∑jFor updating the parameter and preventing the parameter sample from being depleted in the resampling process, the calculation formula is
Beta is a constant;
step 2.5, based on the result of step 2.3, probability by weight
For parameter sample point thetaj,kResampling to obtain a new group of samples
The parameter distribution is continuously close to the posterior probability distribution of the parameters along with the increase of the annealing index;
step 2.6, based on the result of step 2.4, performing a Markov chain Monte Carlo parameter updating process to obtain a new set of samples thetaj+1,k，k＝1,…,N；
Step 2.7, making j equal to j +1, and returning to the step 2.3;
step 2.8, let Pj+1And (3) executing the steps 2.4 to 2.6 to obtain N parameter samples obeying the parameter posterior probability distribution, wherein the N parameter samples are used for representing the parameter probability distribution of the Brillouin optical fiber gain spectrum.
3. The method of claim 2, wherein: the step 2.3 is specifically as follows:
step 2.3.1, setting initial value Pj+1＝(Pj+1)/2, the calculation is based on different target parameters θj,kWeight w ofj,k；
Step 2.3.2, calculate wj,kK is 1, …, and the coefficient of variation of N is P with the coefficient of variation of 100%j+1Is optimal;
step 2.3.3, calculating by dichotomy to finally obtain Pj+1。
4. The method of claim 3, wherein: the step 2.6 is specifically as follows:
step 2.6.1, from
Updating covariance Σ for mean and parametersjExtracting candidate samples theta for Gaussian distribution of covariancer；
Step 2.6.2, calculate based on thetarPrior probability distribution f (θ)r) And likelihood function f (Q | theta)r) And is combined with the original sample
Comparing to obtain the acceptance probability r, wherein the calculation formula is
Step 2.6.3, press [0, 1 ]]Generates a random number u, accepts the candidate sample if r is greater than ur，θj+1,k＝θrOtherwise, reject the candidate sample θr，
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011589241.6A CN112798142B (en) | 2020-12-28 | 2020-12-28 | Brillouin optical fiber sensor strain and temperature two-stage rapid identification method based on Bayesian updating and random simulation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011589241.6A CN112798142B (en) | 2020-12-28 | 2020-12-28 | Brillouin optical fiber sensor strain and temperature two-stage rapid identification method based on Bayesian updating and random simulation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112798142A CN112798142A (en) | 2021-05-14 |
| CN112798142B true CN112798142B (en) | 2021-09-24 |
Family
ID=75805429
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011589241.6A Active CN112798142B (en) | 2020-12-28 | 2020-12-28 | Brillouin optical fiber sensor strain and temperature two-stage rapid identification method based on Bayesian updating and random simulation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112798142B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113297790B (en) * | 2021-05-19 | 2022-05-10 | 哈尔滨工业大学 | High-speed rail response prediction method based on sparse Bayesian width learning |
| CN113639775B (en) * | 2021-08-11 | 2023-08-29 | 武汉钧恒科技有限公司 | Frequency shift extraction method and device based on Brillouin optical time domain reflectometer |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004109105A (en) * | 2002-07-23 | 2004-04-08 | Jfe Steel Kk | Flaw type classification boundary setting method in surface defect detection, and defect detection method |
| CN103063909A (en) * | 2012-12-18 | 2013-04-24 | 天津理工大学 | Linear frequency modulation signal parameter valuation method based on power spectrum |
| CN107202653A (en) * | 2017-06-27 | 2017-09-26 | 广州市禹成消防科技有限公司 | The noise-reduction method and device of temperature-measuring system of distributed fibers |
| CN109523595A (en) * | 2018-11-21 | 2019-03-26 | 南京链和科技有限公司 | A kind of architectural engineering straight line corner angle spacing vision measuring method |
| CN111199254A (en) * | 2019-12-30 | 2020-05-26 | 郑州科技学院 | High-voltage electrical equipment real-time detection method based on RBF neural network and Bayesian network |
-
2020
- 2020-12-28 CN CN202011589241.6A patent/CN112798142B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004109105A (en) * | 2002-07-23 | 2004-04-08 | Jfe Steel Kk | Flaw type classification boundary setting method in surface defect detection, and defect detection method |
| CN103063909A (en) * | 2012-12-18 | 2013-04-24 | 天津理工大学 | Linear frequency modulation signal parameter valuation method based on power spectrum |
| CN107202653A (en) * | 2017-06-27 | 2017-09-26 | 广州市禹成消防科技有限公司 | The noise-reduction method and device of temperature-measuring system of distributed fibers |
| CN109523595A (en) * | 2018-11-21 | 2019-03-26 | 南京链和科技有限公司 | A kind of architectural engineering straight line corner angle spacing vision measuring method |
| CN111199254A (en) * | 2019-12-30 | 2020-05-26 | 郑州科技学院 | High-voltage electrical equipment real-time detection method based on RBF neural network and Bayesian network |
Non-Patent Citations (1)
| Title |
|---|
| "用于动态应变测量的快速分布式布里渊光纤传感";周登望等;《光学学报》;20180331;第38卷(第3期);第0328005-1至0328005-15页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112798142A (en) | 2021-05-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112798142B (en) | Brillouin optical fiber sensor strain and temperature two-stage rapid identification method based on Bayesian updating and random simulation | |
| CN108460230B (en) | Data fusion-based method for predicting residual life and reliability of corroded oil and gas pipeline | |
| CN111680870B (en) | Comprehensive evaluation method for quality of target motion trail | |
| CN110514967B (en) | Fault positioning method for predicting arrival time of traveling wave head based on NARNN model | |
| CN110081923B (en) | Fault detection method and device for automatic acquisition system of field baseline environmental parameters | |
| CN111122162B (en) | Industrial system fault detection method based on Euclidean distance multi-scale fuzzy sample entropy | |
| CN113484813B (en) | Intelligent ammeter fault rate prediction method and system under multi-environment stress | |
| CN105181152A (en) | Calculation method for frequency shift of distributed Brillouin scattered spectrum | |
| CN110555235A (en) | Structure local defect detection method based on vector autoregressive model | |
| CN114018832B (en) | Evaluation method for protection grade of steel surface coating | |
| CN114861737A (en) | Remote disturbance feature extraction method and system of distributed optical fiber sensing system | |
| CN116522085A (en) | Full-automatic inhaul cable frequency extraction, fixed-order and cable force identification method and application | |
| CN110672058B (en) | On-line calibration data sequence matching method and device of sensor for structure monitoring | |
| CN112949735A (en) | Liquid hazardous chemical substance volatile concentration abnormity discovery method based on outlier data mining | |
| CN116337260B (en) | Railway contact net temperature measurement system and method based on wireless communication | |
| CN110333506B (en) | Method for extracting inhaul cable position parameters of cable force measurement radar | |
| CN113687192B (en) | Method for collecting and positioning discharge signal of power transmission line | |
| CN110907984A (en) | Method for detecting earthquake front infrared long-wave radiation abnormal information based on autoregressive moving average model | |
| CN111307030A (en) | GB-RAR high-speed railway bridge deformation information estimation model considering colored noise influence | |
| CN112649354A (en) | Comprehensive evaluation method for measuring corrosion of metal pipeline by multiple sensors | |
| CN112163259A (en) | Method for determining equivalent wind speed ratio of wind profile of typical railway infrastructure | |
| CN112525073B (en) | Structural crack identification method based on Brillouin gain spectrum characteristic parameters | |
| Chen et al. | Criteria of determining the P/T upper limits of GR&R in MSA | |
| CN111693982A (en) | Radar speed measurement AI algorithm | |
| CN117470189A (en) | High-precision calibration method for leveling points for hydrology and water conservancy |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |