CN106678552A - Novel leakage early warning method - Google Patents

Abstract

The invention discloses a novel leakage early warning method. The method comprises the steps that noise samples are collected according to predetermined time intervals; the noise sample collected each time is subjected to fast fourier transform (FFT), noise frequency domain information is obtained, and the magnitude spectrum is calculated; environment noise influence on a frequency domain is eliminated through a self-adaption frequency spectrum noise-reduction algorithm, and relatively stable true leakage noise information is reserved; noise-eliminated data is subjected to frequency spectrum correlation coefficient operation through a frequency spectrum correlation analysis algorithm, and a noise correlation coefficient value is obtained; and a relevant result is subjected to comprehensive statistical analyzing through an early warning analysis algorithm, and accordingly whether leakage occurs or not is determined. By adoption of the novel leakage early warning method, the technical problems that an existing leakage early warning algorithm is poor in anti-disturbance performance, and can only be used at night, 24-hour monitoring is not available, and the false alarm rate and the missing report rate are generally high can be solved.

Description

A kind of new seepage method for early warning

Technical field

The present invention relates to Signal and Information Processing technical field, in particular it relates to a kind of seepage method for early warning.

Background technology

During pipe leakage, fluid media (medium) passes through at a high speed seepage space, due to vibrations, friction, slows down, expansion, clashes into etc., stream Body produces eddy stress or shearing force, forms seepage noise.Seepage sound wave is with the plane wave of non-frequency dispersion and the high-order acoustic mode of frequency dispersion State form is propagated in pipeline fluid, is affected by fluid and pipeline attenuation characteristic, and as propagation distance increases, noise intensity is rapid Weaken.Using this feature of generation noise during pipe leakage, various pipeline leakage detection/monitoring devices are had now been developed both at home and abroad (for example：Leakage measuring instrument by sonic, correlator, seepage early warning etc.), noise detecting method is also the current main using method of water supply network.

Because seepage noise is extremely faint, easily by ambient noise interference, how seepage noise is correctly recognized, needed to seepage Noise and environment noise each the characteristics of be analyzed.Theoretical and actual test is proved：Pipeline environment noise belongs to Gauss white noise Sound, spectral power distribution meets Gauss distribution；Versus environmental noise, any point in seepage noise transmission approach, seepage noise With more stable noise intensity and spectrum distribution feature.Wherein noise intensity feature decision method is also that current foreign countries' main flow is oozed Leakage early warning product is (for example：Britain is bold and generous) method that generally adopts.

Using noise intensity feature decision method, night 2 is typically chosen in:00~4:00 period acquisition noise data, it is main Syllabus is, in order to reduce ambient noise interference, to reduce wrong report and miss probability.During collection, according to the fixed interval T second (general 5 ~10 seconds, different manufacturers were slightly different) acquisition noise data and noise intensity is calculated, gather total degree n times (General N>=1000 It is secondary), for the ease of analysis, the noise intensity discrete serieses that n times are collected are defined as F (t), while defining strength range threshold Value Vth (general to arrange 5db or so, different manufacturers are slightly different) and excision value Vcut are (general to arrange 20db or so, different manufacturers It is slightly different), seepage decision method：

1st, as MIN (F (t)>=Vcut and MAX (F (t))-MIN (F (t))<During=Vth, there is seepage；

2nd, as MIN (F (t)<Vcut or MAX (F (t))-MIN (F (t))>During Vth, ne-leakage occurs；

Wherein MAX () takes sequence maximum, and MIN () takes sequence minima.This method Main Basiss seepage noise intensity It is relatively stable, and the stronger feature of environment noise randomness judges whether seepage occurs.

Although seepage strength characteristic distinguished number has, algorithm is simple, the low advantage of hardware cost, and interference free performance is poor, Can only use at night, there is rate of false alarm and rate of failing to report is generally higher etc. not enough.

The content of the invention

It is an object of the invention to provide a kind of new seepage method for early warning, to solve current seepage warning algorithm anti-interference Can be poor, can only use at night, it is impossible to monitor within 24 hours, there is the rate of false alarm and rate of failing to report generally technical problem such as higher.

In order to realize foregoing invention purpose, the technical solution adopted in the present invention is as follows：

A kind of new seepage method for early warning, including：Adaptive spectrum de-noising, three steps of frequency spectrum related algorithm and early warning；

(1) adaptive spectrum de-noising,

The seepage noise sample of pipeline weak is picked up by the sensor of collecting unit, defining double acquisition interval is TS；A length of TL when defining single acquisition；Once complete gatherer process, i.e. acquisition interval+collection duration is defined, is a collection week Phase TP, then TP=TS+TL；

Continuous m collection period is defined as into de-noising cycle T F；Define regulation coefficient array G [n], wherein n=2*m + 1, make G [n]=- 1/2, -1/4 ... ..-1/2m, 0,1/2m ... ..1/4,1/2；

A Variable Learning g, and 0≤g are defined for each frequency<N, Variable Learning g are used to indicate current frequency adjustment system Several positions, corresponding to the subscript of G [n] array, and Variable Learning g is initialized as m；FFT is performed to each gathered data to become Change, and it is Vi to obtain input range spectrum；Definition study amplitude spectrum Vb；In each collection period to input range spectrum Vi and study width One by one frequency is compared degree spectrum Vb, and is adjusted as follows:

First by frequency regularized learning algorithm variable:

Then, by frequency regularized learning algorithm amplitude spectrum:Vb (n)=Vb (n) * (1+G [g (n)])；

(2) frequency spectrum related algorithm,

Newest amplitude spectrum and previous de-noising periodic amplitude spectrum to receiving performs related operation, obtains frequency spectrum correlation coefficient Value；

Preservation connects the amplitude spectrum for newly receiving to built-in variable；

(3) early warning,

Relevance degree to receiving carries out statistical analysiss, when in continuous certain hour, during relevance degree >=0.8, judges Seepage occurs.

Advantages of the present invention and effect are as follows：

The present invention eliminates the impact of environment noise using adaptive spectrum denoising algorithm；Comprehensively commented using frequency spectrum related algorithm Noise intensity and spectrum distribution feature are estimated, so inventive algorithm has preferable interference free performance, extremely low wrong report and fails to report Rate, it is adapted to continue to monitor for 24 hours, the features such as meet seepage early warning requirement of real-time.

Description of the drawings

Fig. 1 is inventive algorithm schematic flow sheet；

Fig. 2 is the time diagrams such as collection period, the de-noising cycle of the present invention；

Fig. 3 is the frequency spectrum correlation timing schematic diagram of the present invention；

Fig. 4 is adaptive spectrum denoising algorithm schematic diagram of the present invention；

Fig. 5 is the example adaptive spectrum denoising algorithm flow chart of embodiments of the present invention；

Fig. 6 is the example frequency spectrum related algorithm flow chart of embodiments of the present invention；

Fig. 7 is the example early warning analysis algorithm flow chart of embodiments of the present invention.

Specific embodiment

The specific embodiment of the present invention is described in detail below in conjunction with accompanying drawing.It should be appreciated that this place is retouched The specific embodiment stated is merely to illustrate and explains the present invention, is not limited to the present invention.

Affected by aqueous fluid and pipeline attenuation characteristic, seepage noise will be far above low in communication process high frequency attenuation speed The frequency rate of decay, therefore seepage noise frequency can be gathered be mainly distributed in the range of 50~2Khz, according to sampling thheorem, sampling frequency Rate should be higher than that 4Khz, and 8Khz sample frequencys are typically taken in practical application.

According to seepage noise behavior：With metastable noise intensity and spectrum distribution feature.This algorithm synthesis foundation Noise intensity and spectrum distribution feature, are processed noise data from frequency domain and are analyzed, and algorithm process flow process is as shown in Figure 1；

Acquisition noise data Jing FFT (fast Fourier) first conversion every time, obtains noise frequency domain information；By " adaptive Answer frequency spectrum de-noising " algorithm elimination Environmental Noise Influence, retain metastable seepage noise information；Data are by " frequency spectrum after de-noising Correlation analysiss " algorithm is carrying out frequency spectrum correlation operation according to the de-noising cycle, obtains degree of relevancy on noise temporal；Correlation knot Fruit carries out statistical analysiss by " early warning analysis " module, is confirmed whether that seepage occurs.

For convenience of description the algorithm course of work, defines double acquisition interval for TS, defines a length of during single acquisition TL, once complete gatherer process (acquisition interval+collection duration) is a collection period TP for definition, then TP=TS+TL.Will be continuous M collection period is defined as de-noising cycle T F, such as Fig. 2 (collection period, the de-noising cycle time diagram of the present invention, Note：M=5 in figure) shown in:

Each collection period TP gathers a noise data, gathers duration TL, and gathered data Jing FFT computings first are obtained Input range is composed, then by " adaptive spectrum de-noising " algorithm process eliminating environmental disturbances noise；Often through a de-noising week Phase TF, by current de-noising amplitude spectrum and previous de-noising periodic amplitude spectrum " frequency spectrum correlation analysiss " computing is performed, and obtains noise phase Relation number r (i).Such as Fig. 3 (the frequency spectrum correlation timing schematic diagram of the present invention, note：M=5 in figure) shown in.

The correlation coefficient r (i) for calculating carries out statistical disposition by " early warning analysis " module, judges whether seepage occurs. (note：Actual test shows：When TS >=5 second；TL >=1 second；During m >=5, performance is optimal.)

In order to better illustrate the present invention, various pieces are described in detail below：

Adaptive spectrum de-noising part：

Theory and practice is proved:Pipeline environment noise belongs to white Gaussian noise, and spectral power distribution meets Gauss distribution；And Seepage noise has more stable noise intensity and more stable spectrum distribution.For above-mentioned environment noise and the spy of seepage noise Point, designs the self-adapted noise elimination model as shown in Fig. 4 (adaptive spectrum denoising algorithm schematic diagram).

In order to illustrate the operation principle of self-adapted noise elimination, regulation coefficient array G [n], wherein n=2*m+1, order are defined first G [n]=- 1/2, -1/4 ... ..-1/2m, 0,1/2m ... ..1/4,1/2 }；If setting m=5, i.e., 5 collection period definition For a filtering cycle, then regulation coefficient array G (n)={ -1/2, -1/4, -1/8, -1/16, -1/32,0,1/32,1/16,1/ 8,1/4,1/2 }.A Variable Learning g, and 0≤g are defined for each frequency<N, Variable Learning g are used to indicate current frequency adjustment The position of coefficient, corresponding to the subscript of G [n] array, during initial acquisition, Variable Learning g is set to m.Each gathered data is held Row FFT, and it is Vi to obtain input range spectrum；Definition study amplitude spectrum Vb, study amplitude spectrum Vb are used for tracking study input Noise changes, and according to the strong feature of environment noise time randomness, eliminates Environmental Noise Influence, finally gives making an uproar after de-noising Amplitude sound spectrum.

During work, gathered data Jing FFT (fast Fourier) is converted and is calculated acquisition input range spectrum Vi, by input range One by one frequency is compared spectrum Vi and study amplitude spectrum Vb, and computing in the following manner：

First by frequency regularized learning algorithm variable:

Then, by frequency regularized learning algorithm amplitude spectrum:Vb (n)=Vb (n) * (1+G [g (n)])；

Fig. 5 is example adaptive spectrum denoising algorithm flow chart according to the embodiment of the present invention, first to gathering number According to execution FFT, and calculate acquisition input range spectrum Vi；Then input range spectrum Vi and study amplitude spectrum are compared by frequency Vb, obtains spectral change state；According to the spectral change state for obtaining, regularized learning algorithm variable g and study amplitude spectrum Vb；When completing In one de-noising cycle, study amplitude spectrum Vb is passed to into next stage and is processed.

Frequency spectrum correlation analysiss part：

We select simple correlation coefficient algorithm in frequency spectrum correlation analysiss, also known as Pearson's correlation coefficient or " Pearson came product Square correlation coefficient ", it describes the tightness degree contacted between two spacing variables, is a kind of linearly dependent coefficient.Pearson's phase Relation number is the statistic for reflecting two linear variable displacement degrees of correlation, and correlation coefficient is represented with r.

After a complete de-noising cycle, through the amplitude modal data of adaptive spectrum de-noising, with previous de-noising week The amplitude modal data of phase, according to Pearson's correlation coefficient formula computing is launched, and obtains the correlation degree of different time seepage noise. Operational formula is as follows：

This kind of method, by input data of the amplitude modal data directly as correlation analysiss computing, correlated results synthesis are launched Embody seepage noise intensity and spectrum distribution feature.Noise intensity method is relied on relative to simple, with more preferable accuracy And objectivity, advantageously reduce seepage misinformation probability.

The frequency spectrum correlation coefficient r of acquisition, span be [- 1,1], r>0 represents positive correlation, r<0 represents negatively correlated, | r | Illustrate the height of degree of correlation between variable.Distinguishingly, r=1 is referred to as perfect positive correlation, and r=-1 is referred to as perfect negative correlation, r =0 is referred to as uncorrelated.Theory and practice confirmation, when | r | is more than or equal to 0.8, had both thought that two variables had very strong linear phase Guan Xing.

Fig. 6 is example frequency spectrum related algorithm flow chart according to the embodiment of the present invention, first to current study amplitude The spectrum Vb and amplitude spectrum VPREV in previous de-noising cycle does correlation analysiss, obtains correlation coefficient r value；Afterwards Vb is saved in into VPREV In；Finally the correlation coefficient r value of calculating is passed to into early warning analysis to be processed.

Early warning analysis principle：

Jing frequency spectrum correlation analysiss, the correlation coefficient r value of acquisition, the signal for completing certain hour by early warning analysis algorithm is related Degree statistics is processed with early warning, is shown according to the practical application and actual test of current seepage early warning：When continuous time >=60 minute During r value >=0.8 of all acquisitions, pipe leakage occurs；When the conditions set forth above are not met, pipeline ne-leakage.Using this early warning Judgment mode relative to tradition only night collection and alarm mode, can accomplish 24 hours monitor, possess more preferable real-time and Practicality.

Fig. 7 is example early warning analysis algorithm flow chart according to the embodiment of the present invention, the phase obtained to correlation analysiss Relation number r values are counted, if r >=0.8, serial correlation degree variables TC=TC+1；Otherwise TC=0；Judge continuous phase Close whether degree variables TC exceedes early warning number of times (such as 60 times, early warning number of times=pre-warning time/TF), if it exceeds early warning time Number, then early warning output.

The preferred embodiment of the present invention is described in detail above in association with accompanying drawing, but, the present invention is not limited to above-mentioned reality The detail in mode is applied, in the range of the technology design of the present invention, various letters can be carried out to technical scheme Monotropic type, these simple variants belong to protection scope of the present invention.

It is further to note that each particular technique feature described in above-mentioned specific embodiment, in not lance In the case of shield, can be combined by any suitable means.

Additionally, combination in any can also be carried out between a variety of embodiments of the present invention, as long as it is without prejudice to this The thought of invention, it should equally be considered as content disclosed in this invention.

Claims (1)

1. a kind of new seepage method for early warning, including：Adaptive spectrum de-noising, three steps of frequency spectrum related algorithm and early warning；

(1) adaptive spectrum de-noising,

The seepage noise sample of pipeline weak is picked up by the sensor of collecting unit, it is TS to define double acquisition interval； A length of TL when defining single acquisition；Once complete gatherer process, i.e. acquisition interval+collection duration is defined, is a collection period TP, then TP=TS+TL；

Continuous m collection period is defined as into de-noising cycle T F；Definition regulation coefficient array G [n], wherein n=2*m+1, Make G [n]=- 1/2, -1/4 ... ..-1/2m, 0,1/2m ... ..1/4,1/2；

A Variable Learning g, and 0≤g are defined for each frequency<N, Variable Learning g are used to indicate current frequency regulation coefficient Position, corresponding to the subscript of G [n] array, and Variable Learning g is initialized as m；

FFT is performed to each gathered data, and it is Vi to obtain input range spectrum；Definition study amplitude spectrum Vb；Adopt at each The collection cycle, one by one frequency was compared to input range spectrum Vi and study amplitude spectrum Vb, and carried out study adjustment as follows:

First by frequency regularized learning algorithm variable:

Then, by frequency regularized learning algorithm amplitude spectrum:Vb (n)=Vb (n) * (1+G [g (n)])；

(2) frequency spectrum related algorithm,

Newest amplitude spectrum and previous de-noising periodic amplitude spectrum to receiving performs related operation, obtains frequency spectrum correlation coefficient value；

Preservation connects the amplitude spectrum for newly receiving to built-in variable；

(3) early warning,

Relevance degree to receiving carries out statistical analysiss, when in continuous certain hour, during relevance degree >=0.8, judging seepage Occur.