CN103322416A - Pipeline weak leakage detecting device and detecting method based on fuzzy hyperbolic chaos model - Google Patents

Abstract

The invention discloses a pipeline weak leakage detecting device and a detecting method based on a fuzzy hyperbolic chaos model and belongs to the technical field of pipeline detection. Two-stage detection is used, the first-stage detection is abnormal signal detection based on a neural network to mark abnormal signals, and the second-stage detection detects the abnormal signals marked by the first-state detection. By the two-stage detection, the first-stage primary detection is fast, the second-stage detection detects the abnormal signals marked by the first-state detection, and leakage detection capability is increased. The two-stage detection can be implemented on a FPGA (field programmable gate array) hardware platform, and overall system response speed is increased. By the double coupling time-varying delay fuzzy hyperbolic chaos model of the second-stage detection, initial value sensitivity and noise immunity of the chaos system is utilized to increase identification capability of weak leakage signals submerged in noise so as to highlight chaos phenomenon, system requirements for signal to noise ratio is lowered, and weak leakage detection is achieved.

Description

The faint leak detecting device of pipeline reaches based on fuzzy hyperbolic chaotic model detecting method

Technical field

The invention belongs to the pipe detection technical field, be specifically related to the faint leak detecting device of pipeline and reach based on fuzzy hyperbolic chaotic model detecting method.

Background technique

Along with the growth with the pipeline active time of increasing of oil transport pipeline, the safe condition of pipeline transport allows of no optimist.The degree of aging of a lot of pipelines is more serious, and has entered the multiple phase of leakage accident; In addition, the phenomenon that the artificial destruction pipeline is stolen petroleum resources is also quite serious, brings major safety risks.Therefore, pipeline leakage testing becomes the important process content of pipe safety production management.

At present, the large flow Leak testtion of moment such as pipeline generation booster has been had preferably in real time detection technique, but lacked the effective detection to faint leakage signal.The faint leakage here comprises little leakage and two kinds of situations of gradual leakage: little leakage refers to that the constant and leakage rate of unit time internal leakage is less than the leakage of current throughput rate 1%; Gradual leakage refers to the starting leakage amount less than the leakage of current throughput rate 1%, and the leakage rate in the unit time is along with leak time increases increasing.Existing Leak testtion theory is that ducted interference is regarded as random noise, take noise-reduction method that noise is filtered before faut detection, yet faint leakage information has also been filtered when filtering interference, causes detecting the faint leakage failure of pipeline.At present, for the discussion of the faint leakage signal context of detection of pipeline also seldom.

Summary of the invention

For the defective that prior art exists, the purpose of this invention is to provide the faint leak detecting device of a kind of pipeline and reach based on fuzzy hyperbolic chaotic model detecting method, through the hierarchical detection method pipe leakage is detected.

Technological scheme of the present invention is achieved in that the faint leak detecting device of a kind of pipeline, comprising:

Pressure sensor module: be used for to gather the pressure change signal of pipeline head and end, and export to the signal condition module after pressure signal is converted to electrical signal;

Signal condition module: be used for that pressure transducer is passed the electrical signal of coming and carry out filtering, amplification, and the electrical signal that will nurse one's health is exported to the A/D modular converter;

A/D modular converter: be used for electrical signal is carried out analog-to-digital conversion, and the digital signal that produces is exported to Leak testtion module in the FPGA central processing unit module;

FPGA central processing unit module: be used for pipeline leakage signal is detected; FPGA central processing unit module further comprises:

Time-sequence control module: for generation of the work schedule of A/D modular converter;

The Leak testtion module: adopt abnormal signal based on neuron network to detect and the hierarchical detection pattern that becomes fuzzy time-delay hyperbolic chaotic model during based on two coupling detects faint leakage signal, the Leak testtion module further comprises:

Abnormal signal detection sub-module based on neuron network: be used for that real time data is carried out abnormal signal and detect, and with the abnormal signal mark;

Become the faint Leak testtion submodule of fuzzy time-delay hyperbolic chaotic model based on two whens coupling: for detection of the abnormal signal of mark, judge whether leakage signal of these abnormal signals.

Described pressure sensor module connects FPGA through signal condition module, A/D modular converter.

Adopt the faint leak detecting device realization of pipeline based on the detecting method of fuzzy hyperbolic chaotic model, may further comprise the steps:

Step 1: the instantaneous pressure signal that gathers the pressure transducer that is installed in pipe ends;

Step 2: choose certain time period, certain pipeline section, when normally moving, the pipeline pressure data of utilizing the faint leak detecting device of pipeline to collect consist of historical data;

Step 3: choose a part of data in the historical data from step 2, train the neural network model that abnormal signal is detected with these off-line datas;

Step 4: choose another part data from historical data, the best of calculating these data embeds dimension d and delay time T again;

Step 5: the historical data of selecting in the step 4 is carried out phase space reconfiguration, will be converted to matrix form with the data that vector form represents;

Step 6: become fuzzy time-delay hyperbolic chaotic model when utilizing two coupling that the matrix off-line training that obtains in the step 5 detects faint leakage signal;

Step 7: the pipeline pressure real time data is sent into neural network model abnormal signal is detected in real time, the variable weight of neural network model will constantly be adjusted renewal according to the variation of real time data, the predicted value u of neural network model ₁With with u ₁Corresponding actual value x ₁Between difference E ₁With threshold value T ₁Compare, if difference E ₁Threshold value T ₁, then detect abnormal signal, abnormal signal is carried out mark, execution in step 8, otherwise then repeated execution of steps 7;

Step 8: will when the signal of step 7 mark is sent into based on two coupling, become fuzzy time-delay hyperbolic chaotic model faint leakage is detected in real time, the variable weight and the coupling factor that become fuzzy time-delay hyperbolic chaotic model during based on two coupling will constantly be adjusted renewal according to the variation that passes the marking signal of coming in the step 7, become the predicted value u of fuzzy time-delay hyperbolic chaotic model during two coupling ₂With with u ₂Corresponding actual value x ₂Between difference E ₂With threshold value T ₂Compare, if E ₂Threshold value T ₂, then detect faint leakage signal.

The representation that becomes fuzzy time-delay hyperbolic chaotic model during the described two coupling of step 6 into:

$\left\{\begin{array}{c}{X}_n=F\left(X_{n-1}\right)+\mathrm{\ϵ}(Y_n-\mathrm{hF}\left(X_{n-1}\right))\\ Y_n=F\left(Y_{n-1}\right)+\mathrm{\ϵ}(X_n-\mathrm{hF}\left(Y_{n-1}\right))\end{array}\right.$

Wherein, X _n, Y _nBe system variable, h is power gain, and ε is coupling factor, and when ε=0, the coupled relation of two systems disappears; When ε ≠ 0, follow according to Chaotic Synchronization Theory, select to make Y _n→ X _nThe ε that sets up is as coupling factor; When being, top two F () function becomes the function expression of fuzzy time-delay hyperbolic model, two system variable X _n, Y _nCan under the impact of coupling, be tending towards rapidly synchronous.

The expression of F () function is as follows:

$\stackrel{\·}{x}=A\tanh \left(\mathrm{Kx}\right)+\mathrm{Bu}$

Wherein, A, B are normal matrix, and K is diagonal matrix.

Step 8 is described, and become the detecting step of faint Leak testtion module of fuzzy time-delay hyperbolic chaotic model based on two whens coupling as follows:

(1) selected threshold value T ₂: make the leakage signal that comprises in the data of error less than threshold value more few better, the threshold value of this moment is best threshold value, the false alarm reduction rate;

(2) pass the marking signal x that comes in the step 7 _iBecome the input of fuzzy time-delay hyperbolic chaotic model during as two coupling, obtain next step prediction output

(3) utilize in the step 7 and to become fuzzy time-delay hyperbolic chaotic model when passing in real time the in real time two coupling of training of marking signal of coming, the weights that obtain become the weights of fuzzy time-delay hyperbolic chaotic model when being used for two coupling that next step prediction of real-time update uses; Also to adjust coupling factor ε in the time of the weights adjustment, because the change of weights causes X _n, Y _nThe change of function expression can make both no longer synchronous, so will adjust coupling factor, makes X _n→ Y _n

(4) the actual value x in next moment (i+1 constantly) _I+1With its predicted value

Predicated error

As the foundation of faut detection, if E ₂＜T ₂, faint leakage does not then occur; Otherwise, faint leakage then occurs.

(5) repeat (3)-(5), finish the detection to flag data.

Beneficial effect of the present invention: (1) adopts the hierarchical detection method that pipe leakage is detected, and the first order is the abnormal signal detection based on neuron network, and abnormal signal is carried out mark, and the second level is only detected the abnormal signal of first order mark; Detect by above-mentioned two-stage signal, the initial survey of the first order, the speed that realized is fast, and the second level is detected the abnormal signal of the first order again, has improved the detectability to leakage signal.And this hierarchical detection method realizes under the happy hardware platform of FPG, utilizes the happy multibus parallel processing mechanism of FPG, has improved the entire system speed of response.(2) become fuzzy time-delay hyperbolic chaotic model during two coupling of the second level, utilize the chaos system sensitivity to initial and to the immunocompetence of noise, improved being submerged in the recognition capability of the faint leakage signal in the noise.(3) model of the second level adopts two forms that are coupled, and can highlight chaos phenomenon, and the reduction system is to the requirement of signal to noise ratio, thereby realization is to the detection of faint leakage.

Description of drawings

Fig. 1 is the faint leak detecting device structured flowchart of one embodiment of the present invention pipeline;

Fig. 2 is the circuit theory diagrams of one embodiment of the present invention signal condition module;

Fig. 3 is the happy interface circuit figure of one embodiment of the present invention pleasure/D conversion chip and FPG;

Fig. 4 is the middle Leak testtion modular structure block diagram that arranges in the happy central processing unit module of one embodiment of the present invention FPG;

Fig. 5 is the happy interface circuit figure of the happy M of one embodiment of the present invention SDR and FPG;

Fig. 6 is that one embodiment of the present invention is based on fuzzy hyperbolic chaotic model detecting method flow chart;

When being one embodiment of the present invention based on two coupling of fuzzy hyperbolic chaotic model, Fig. 7 becomes fuzzy time-delay hyperbolic chaotic model structured flowchart.

Embodiment

Below in conjunction with accompanying drawing embodiments of the present invention are described in further detail.

Provided the faint leak detecting device structured flowchart of pipeline in the present embodiment, as shown in Figure 1.This device comprises pressure sensor module, signal condition module, A/D modular converter and FPGA central processing unit module, and wherein FPGA central processing unit module further comprises time-sequence control module and Leak testtion module.Wherein, the model of pressure sensor module is PT500-502, and the model of A/D modular converter is ADS7844, and the model of FPGA central processing unit module is EP3C25Q240C8.

The pipe ends that pressure transducer is installed in, pressure transducer gathers the instantaneous pressure data of pipeline, exported to the input end of signal condition module by output terminal, the output terminal of signal condition module connects the input end of A/D modular converter, the output terminal of A/D modular converter connects the input end of FPGA central processing unit module, an output terminal of FPGA central processing module connects alarm unit, and another output terminal connects the A/D modular converter.

Pressure sensor module is converted to voltage signal with pressure signal, the force value that collects is as research object because this pipeline leakage testing device is, so the sensitivity of pressure transducer is important, but the noise that high-precision sensitivity also can't filtering pressure signal itself again, therefore as long as select suitable sensor, need not too pursue high sensitivity.

The circuit theory diagrams of signal condition module as shown in Figure 2, this module realizes filtering and the amplification of signal, at first after filtration wave circuit filtering of the output of pressure transducer, then be connected to the inverting input of operational amplifier through the resistance R 2 of a 10K, in-phase input end connects the reference voltage of 2.5V, the end of the output terminal contact resistance R3 of operational amplifier, one end of resistance R 1 and an end of capacitor C 2, the other end of resistance R 3 connects the input end of A/D conversion chip as the output terminal of signal condition module, the other end of resistance R 1 connected the inverting input of calculating amplifier, the other end ground connection of capacitor C 2.The model of operational amplifier is AD824 in the present embodiment.

The interface circuit figure of A/D conversion chip and FPGA as shown in Figure 3, the A/D conversion chip is converted to digital signal with voltage signal, 6 different output terminals of A/D conversion chip connect respectively the self-defined I/O mouth of the happy time-sequence control module of FPG, the DCLK end that is the A/D conversion chip connects the happy I/O.23 end of FPG, the CS end of A/D conversion chip connects the I/O.24 end, the Din end of A/D conversion chip connects the I/O.25 end, the Busy end of A/D conversion chip connects the I/O.26 end, the Dout end of A/D conversion chip connects the I/O.27 end, wherein, the model of A/D conversion chip is ADS7844, and the model of FPGA is EP3C25Q240C8.

In the FPGA central processing unit module Leak testtion modular structure block diagram as shown in Figure 4, the Leak testtion module is divided into following four modules: history data store module 2, real time data cache module 1, neural metwork training and abnormal signal testing module 3 and two Coupled Chaotic model training and faint Leak testtion module 4.History data store module 2 is made of ROM in the sheet; Real time data cache module 1 is made of the outer SDRAM of sheet.

Neural metwork training and abnormal signal testing module 3 comprise such as lower unit: logic control element 301, arithmetic element 302, weights regulon 304 and weight storage unit 303.Two Coupled Chaotic model trainings and faint Leak testtion module 4 comprise such as lower unit: logic control element 401, arithmetic element 402, weights regulon 404, weight storage unit 403 and coupling factor regulon 405.The workflow of neural metwork training and abnormal signal testing module 3 and two Coupled Chaotic model trainings and faint Leak testtion module 4 is basic identical: the logic control element output signal comprises CLK clock, address signal, control signal.When control signal is 0, carry out the training of model, training is finished, control signal becomes 1, carry out the real-time detection of signal, signal is through arithmetic element prediction of output value, predicted value and actual value relatively after, by weights regulon real-time update weights, the weights after the renewal carry out the storage of new weights under the control of CLK clock and address signal; Unique difference is, two Coupled Chaotic model trainings and faint Leak testtion module have increased the adjusting of coupling factor when weights are regulated, and the coupling factor after the renewal is stored under the control of CLK clock and address signal.Wherein, the model of the outer happy M of SDR of sheet is HY57V641620.The interface circuit figure of the outer SDRAM of sheet and FPGA as scheme attached shown in.The CLK end of the outer SDRAM of sheet connects the CLK0 end of FPGA, the BA0 end of the outer SDRAM of sheet connects the BA0 end of FPGA, the BA1 end of the outer SDRAM of sheet connects the BA1 of FPGA, the nCS end of the outer SDRAM of sheet connects the nCS end of FPGA, the nWE end of the outer SDRAM of sheet connects the nWE end of FPGA, the nCAS end of the outer SDRAM of sheet connects the nCAS end of FPGA, the nRAS end of the outer SDRAM of sheet connects the nCAS end of FPGA, the A11-A0 end of the outer SDRAM of sheet connects the ADDR11-ADDA0 end of FPGA, the DQ15-DQ0 end of the outer SDRAM of sheet connects the DATA15-DATA0 end of FPGA, and the LDQM/UDQM end of the outer SDRAM of sheet connects the LDQM/UDQM end of FPGA.

The faint leak detecting device working procedure of pipeline is as follows in the present embodiment: pressure sensor module gathers the pressure change signal of pipeline head and end, and the electrical signal of conversion is exported to the signal condition module.Electrical signal is sent into the A/D modular converter after the signal condition module is carried out filtering, amplification; Under the control of the time-sequence control module in the happy central processing unit module of FPG, produce the work schedule of A/D modular converter, control A/D modular converter is realized mould/number (A/D) conversion, and the Leak testtion module that the digital signal after will changing is sent in the FPGA central processing unit module is carried out faint leakage signal detection.The Leak testtion module utilizes pipeline pressure time series (mixed signal that comprises faint leakage signal, Chaos Variable and other random noises) to have the characteristics of chaotic characteristic, at first, become fuzzy time-delay hyperbolic chaotic model when setting up neural network model and two coupling, and utilize the pipeline pressure historical data to train; Secondly, suitable threshold value is set, the pipeline pressure time series is carried out detecting the mark abnormal signal based on the abnormal signal of neuron network; The faint leakage signal that becomes fuzzy time-delay hyperbolic chaotic model when at last, carrying out based on two coupling to the abnormal signal that is labeled detects.

Adopt the faint leak detecting device realization of pipeline based on the detecting method of fuzzy hyperbolic chaotic model, its flow process may further comprise the steps as shown in Figure 6:

Step 1: the instantaneous pressure signal that gathers the pressure transducer that is installed in pipe ends;

Step 2: choose certain time period, certain pipeline section, when normally moving, the pipeline pressure data of utilizing the faint leak detecting device of pipeline to collect consist of historical data, and the number of historical data is N;

Step 3: choose a n in the historical data from step 2 ₁Data, n ₁＜N uses this n ₁The neural network model that individual data off-line training detects abnormal signal.

Step 4: choose n in the historical data from step 2 ₂Individual data, n ₂For removing n among the N ₁Historical data is afterwards calculated this n ₂The best of individual data embeds dimension d and delay time T, and it is different retard time that the best of different pieces of information collection embeds peacekeeping.

Step 5: to the n in the step 4 ₂Individual data are carried out phase space reconfiguration, n originally ₂Individual data are with n ₂* 1 vector form represents, phase space reconfiguration is exactly to convert original vector form to matrix form, the best of trying to achieve in the ranks number of this matrix and the step 4 embed dimension and retard time relevant, the line number of matrix equals n ₂-(d-1) τ, matrix column number equal the best dimension d that embeds;

Step 6: become fuzzy time-delay hyperbolic chaotic model when utilizing two coupling that the matrix off-line training that obtains in the step 5 detects faint leakage signal, formula is as follows:

$\left\{\begin{array}{c}{X}_n=F\left(X_{n-1}\right)+\mathrm{\ϵ}(Y_n-\mathrm{hF}\left(X_{n-1}\right))\\ Y_n=F\left(Y_{n-1}\right)+\mathrm{\ϵ}(X_n-\mathrm{hF}\left(Y_{n-1}\right))\end{array}\right.$

Wherein, X _n, Y _nBe system variable, h is power gain, and ε is coupling factor, and when ε=0, the coupled relation of two systems disappears; When ε ≠ 0, follow according to Chaotic Synchronization Theory, select to make Y _n→ X _nThe ε that sets up is as coupling factor; When being, top two F () function becomes the function expression of fuzzy time-delay hyperbolic model, two system variable X _n, Y _nCan under the impact of coupling, be tending towards rapidly synchronous.

The expression of F () function is as follows:

$\stackrel{\·}{x}=A\tanh \left(\mathrm{Kx}\right)+\mathrm{Bu}$

Wherein, A, B are normal matrix, and K is diagonal matrix.

As shown in Figure 7, become in the time of two when fuzzy time-delay hyperbolic chaotic model consists of two coupling and become fuzzy time-delay hyperbolic chaotic model, the input of two models is respectively X _N-1, Y _N-1, output is respectively F (X _N-1), F (Y _N-1), F (X _N-1) and F (Y _N-1) pass through separately after the h amplification doubly, regulate whether reach synchronous by coupling factor ε, the output after two models are coupled is respectively X _n, Y _n

Step 7: the pipeline pressure real time data is sent into neural network model abnormal signal is detected in real time, the variable weight of neural network model will constantly be adjusted renewal according to the variation of real time data, the predicted value u of neural network model ₁With with u ₁Corresponding actual value x ₁Between difference E ₁With threshold value T ₁Compare, if difference E ₁Threshold value T ₁, then detect abnormal signal, abnormal signal is carried out mark, otherwise then repeated execution of steps 7;

Step 8: will when the signal of step 7 mark is sent into based on two coupling, become fuzzy time-delay hyperbolic chaotic model faint leakage is detected in real time, the variable weight and the coupling factor that become fuzzy time-delay hyperbolic chaotic model during based on two coupling will constantly be adjusted renewal according to the variation that passes the marking signal of coming in the step 7, become the predicted value u of fuzzy time-delay hyperbolic chaotic model during two coupling ₂With with u ₂Corresponding actual value x ₂Between difference E ₂With threshold value T ₂Compare, if E ₂Threshold value T ₂, then detect faint leakage signal.Detailed process is as follows:

(1) selected threshold value T ₂: make the leakage signal that comprises in the data of error less than threshold value more few better, the threshold value of this moment is best threshold value, the false alarm reduction rate;

(2) pass the marking signal x that comes in the step 7 _iBecome the input of fuzzy time-delay hyperbolic chaotic model during as two coupling, obtain next step prediction output

(3) utilize in the step 7 and to become fuzzy time-delay hyperbolic chaotic model when passing in real time the in real time two coupling of training of marking signal of coming, the weights that obtain become the weights of fuzzy time-delay hyperbolic chaotic model when being used for two coupling that next step prediction of real-time update uses; Also to adjust coupling factor ε in the time of the weights adjustment, because the change of weights causes X _n, Y _nThe change of function expression can make both no longer synchronous, so will adjust coupling factor, makes X _n→ Y _n

(4) the actual value x in next moment (i+1 constantly) _I+1With its predicted value

Predicated error

As the foundation of faut detection, if E ₂＜T ₂, faint leakage does not then occur; Otherwise, faint leakage then occurs.

(5) repeat (3)-(5), finish the detection to flag data.

Below in conjunction with concrete example above-mentioned detecting step is done explanation:

The first step, 500 data when selected pipeline normally moves form historical data, suppose that the representation of this historical data is (a ₁, a ₂, a ₃..., a ₅₀₀), these data all are pressure datas, unit is MPa;

Second step selects front 100 data as (a ₁, a ₂, a ₃..., a ₁₀₀) off-line training neural network model that abnormal signal is detected;

In the 3rd step, the best of calculating rear 400 data embeds dimension d and delay time T, and it is attached adopting pseudo-nearest neighbour method to try to achieve d, and adopting the autocorrelation analysis method to try to achieve τ is 4.Will be with vector form (a ₁₀₁, a ₁₀₂, a ₁₀₃..., a ₅₀₀) data reconstruction of expression becomes the phase space that represents with matrix form τ, makes (b ₁, b ₂, b ₃..., b ₄₀₀)=(a ₁₀₁, a ₁₀₂, a ₁₀₃..., a ₅₀₀), then matrix form is; V=(X ₁, X ₂..., X _t) ^T, t=484, this matrix are 484 * 5 matrixes.Wherein, X _i=(b _i, b _I+4, b _I+8, b _I+12, b _I+16), i=1,2,3 .., t.And, become fuzzy time-delay hyperbolic chaotic model when utilizing the two coupling of this matrix V off-line training;

The 4th step is with real time data x _iSend into neural network model abnormal signal is detected in real time, find out abnormal signal y _j, with y _jBecome fuzzy time-delay hyperbolic chaotic model when sending into two coupling and carry out leakage signal and detect, if the leakage signal of being judged as then report to the police, otherwise, continue to detect the new real time data of input,

Although more than described the specific embodiment of the present invention, the those skilled in the art in related domain should be appreciated that these only illustrate, and can make various changes or modifications to these mode of executions, and not deviate from principle of the present invention and essence.Scope of the present invention is only limited by appended claims.

Claims (5)

1. faint leak detecting device of pipeline is characterized in that: comprising:

Pressure sensor module: be used for to gather the pressure change signal of pipeline head and end, and export to the signal condition module after pressure signal is converted to electrical signal;

Signal condition module: be used for that pressure transducer is passed the electrical signal of coming and carry out filtering, amplification, and the electrical signal that will nurse one's health is exported to the A/D modular converter;

A/D modular converter: be used for electrical signal is carried out analog-to-digital conversion, and the digital signal that produces is exported to Leak testtion module in the FPGA central processing unit module;

FPGA central processing unit module: be used for pipeline leakage signal is detected; FPGA central processing unit module further comprises:

Time-sequence control module: for generation of the work schedule of A/D modular converter;

The Leak testtion module: adopt abnormal signal based on neuron network to detect and the hierarchical detection pattern that becomes fuzzy time-delay hyperbolic chaotic model during based on two coupling detects faint leakage signal, the Leak testtion module further comprises:

Abnormal signal detection sub-module based on neuron network: be used for that real time data is carried out abnormal signal and detect, and with the abnormal signal mark;

Become the faint Leak testtion submodule of fuzzy time-delay hyperbolic chaotic model based on two whens coupling: for detection of the abnormal signal of mark, judge whether leakage signal of these abnormal signals.

2. the faint leak detecting device of pipeline according to claim 1 is characterized in that: described pressure sensor module connects FPGA through signal condition module, A/D modular converter.

3. adopt the faint leak detecting device realization of pipeline claimed in claim 1 based on the detecting method of fuzzy hyperbolic chaotic model, it is characterized in that: may further comprise the steps:

Step 1: the instantaneous pressure signal that gathers the pressure transducer that is installed in pipe ends;

Step 2: choose certain time period, certain pipeline section, when normally moving, the pipeline pressure data of utilizing the faint leak detecting device of pipeline to collect consist of historical data;

Step 3: choose a part of data in the historical data from step 2, train the neural network model that abnormal signal is detected with these off-line datas;

Step 4: choose another part data from historical data, the best of calculating these data embeds dimension d and delay time T again;

Step 5: the historical data of selecting in the step 4 is carried out phase space reconfiguration, will be converted to matrix form with the data that vector form represents;

Step 6: become fuzzy time-delay hyperbolic chaotic model when utilizing two coupling that the matrix off-line training that obtains in the step 5 detects faint leakage signal;

Step 7: the pipeline pressure real time data is sent into neural network model abnormal signal is detected in real time, the variable weight of neural network model will constantly be adjusted renewal according to the variation of real time data, the predicted value u of neural network model ₁With with u ₁Corresponding actual value x ₁Between difference E ₁With threshold value T ₁Compare, if difference E ₁Threshold value T ₁, then detect abnormal signal, abnormal signal is carried out mark, execution in step 8, otherwise then repeated execution of steps 7;

Step 8: will when the signal of step 7 mark is sent into based on two coupling, become fuzzy time-delay hyperbolic chaotic model faint leakage is detected in real time, the variable weight and the coupling factor that become fuzzy time-delay hyperbolic chaotic model during based on two coupling will constantly be adjusted renewal according to the variation that passes the marking signal of coming in the step 7, become the predicted value u of fuzzy time-delay hyperbolic chaotic model during two coupling ₂With with u ₂Corresponding actual value x ₂Between difference E ₂With threshold value T ₂Compare, if E ₂Threshold value T ₂, then detect faint leakage signal.

4. the detecting method based on the fuzzy hyperbolic chaotic model according to claim 3 is characterized in that: the representation that becomes fuzzy time-delay hyperbolic chaotic model during the described two coupling of step 6 into:

Wherein, X _n, Y _nBe system variable, h is power gain, and ε is coupling factor, and when ε=0, the coupled relation of two systems disappears; When ε ≠ 0, follow according to Chaotic Synchronization Theory, select to make Y _n→ X _nThe ε that sets up is as coupling factor; When being, top two F () function becomes the function expression of fuzzy time-delay hyperbolic model, two system variable X _n, Y _nCan under the impact of coupling, be tending towards rapidly synchronous;

The expression of F () function is as follows:

Wherein, A, B are normal matrix, and K is diagonal matrix.

5. the detecting method based on the fuzzy hyperbolic chaotic model according to claim 1 is characterized in that: step 8 is described, and to become the detecting step of faint Leak testtion module of fuzzy time-delay hyperbolic chaotic model during based on two coupling as follows:

(1) selected threshold value T ₂: make the leakage signal that comprises in the data of error less than threshold value more few better, the threshold value of this moment is best threshold value, the false alarm reduction rate;

(2) pass the marking signal x that comes in the step 7 _iBecome the input of fuzzy time-delay hyperbolic chaotic model during as two coupling, obtain next step prediction output

(3) utilize in the step 7 and to become fuzzy time-delay hyperbolic chaotic model when passing in real time the in real time two coupling of training of marking signal of coming, the weights that obtain become the weights of fuzzy time-delay hyperbolic chaotic model when being used for two coupling that next step prediction of real-time update uses; Also to adjust coupling factor ε in the time of the weights adjustment, because the change of weights causes X _n, Y _nThe change of function expression can make both no longer synchronous, so will adjust coupling factor, makes X _n→ Y _n

(4) next actual value x constantly _I+1With its predicted value

Predicated error

As the foundation of faut detection, if E ₂＜T ₂, faint leakage does not then occur; Otherwise, faint leakage then occurs.

(5) repeat (3)-(5), finish the detection to flag data.

Images