CN102914763A - Method for positioning damage sound emission source of hydraulic concrete - Google Patents

Abstract

The invention discloses a method for positioning a damage sound emission source of hydraulic concrete. The method for positioning the damage sound emission source of the hydraulic concrete comprises the following steps of simulating a sound emission signal on the surface of the hydraulic concrete and acquiring characteristic parameters of a damage sound emission signal of a test piece; building a damage sound emission source model based on a least-squares support vector machine; confirming an optimized combination mode of the characteristic parameters of the sound emission signal input by the model to obtain a preferable positioning model; and utilizing the model to input measured values of the characteristic parameters of the sound emission signal so as to achieve quick positioning of the damage sound emission source. The method for positioning the damage sound emission source of the hydraulic concrete does not need time difference and wave velocity of paths of a sound emission system, and accordingly overcomes the defect that a time difference positioning method relies on practical experiences of project technicians and utilizes a threshold crossing technology to decide arrival time and the like.

Description

A kind of concrete for hydraulic structure damage acoustic emission source locating method

Technical field

The present invention relates to a kind of concrete for hydraulic structure damage acoustic emission source locating method.

Background technology

For the Hydro-concrete Structures of large volume, how to detect the position of its damage existence and the development and change of degree of impairment, most important to the safety that ensures engineering.Acoustic emission is as a kind of passive Dynamic Non-Destruction Measurement, because acoustic emission signal from material internal, by collecting, analyzing sound emission signal characteristic, can be used to monitor security and the overload alarm of damage defect for a long time continuously.Wherein, the accurate location of acoustic emission source is to utilize acoustic emission to carry out an important indicator of concrete for hydraulic structure damage check and evaluation.At present, for different acoustic emission signal types, more existing localization methods as for the typical burst acoustic emission signal that is produced by crackle, adopt time difference positioning method and area positioning method usually.Time-of-arrival loaction is the mistiming that arrives different sensors according to the acoustic emission signal that same acoustic emission source sends, and determines class methods of sound source position through velocity of wave, the isoparametric measurement of transducer spacing and algorithm computing; Regional mapping method is to judge the residing zone of acoustic emission source according to number of probes and relative time error sequential.A large amount of low amplitude signals easily appears in the acoustic emission source time difference positioning method, and its bearing accuracy is subject to the impact of the many factors such as velocity of wave, decay, waveform; Area positioning method than faster, easy, but the location is rough, is mainly used in that compound substance etc. is too high owing to Frequency of acoustic emission, propagation attenuation is excessive or the limited occasion that adopts the time difference to locate of being difficult to of sense channel number.In addition, these two class methods also depend on engineering staff's practical experience to a great extent to the accurate location of acoustic emission source.

From mathematics, the location, source is in fact relational model between a structure sound emission source signal feature and sound source position, and the process that defines and seek least error (residual error) point by error can realize by the method for pattern-recognition.But classic method is many based on the empirical risk minimization criterion, and in the finite sample situation, empirical risk minimization might not mean that expected risk is minimum, can not guarantee that the model of setting up has good popularization and generalization ability.

Summary of the invention

Goal of the invention: the object of the invention is to for the deficiencies in the prior art, the concrete for hydraulic structure damage that a kind of modeling is simple and bearing accuracy is high acoustic emission source locating method is provided.

Technical scheme: the invention provides a kind of concrete for hydraulic structure damage acoustic emission source locating method, it is characterized in that may further comprise the steps:

(1) adopts disconnected lead stamp to intend burst acoustic emission signal on the concrete for hydraulic structure surface, carry out the acoustic emission artificial excitation source location test of concrete for hydraulic structure damage, the detection data of acquisition and processing test specimen damage FEATURE PARAMETERS OF ACOUSTIC EMISSION and sound source position;

(2) set up damage acoustic emission source location model based on least square method supporting vector machine, be input as described FEATURE PARAMETERS OF ACOUSTIC EMISSION, be output as the acoustic emission source position; Its nonlinear relationship expression formula is:

In the formula: y is the acoustic emission source position, and x is the characteristics of Acoustic Emission parameter,

Be the Nonlinear Mapping function of the input space to high-dimensional feature space, w is weight vector, and b is amount of bias, all is undetermined parameter.

Setting up the acoustic emission source location model with least square method supporting vector machine is summed up as and finds the solution following optimization problem:

In the formula: ξ=[ξ ₁, ξ ₂..., ξ _l] ^T, ξ _iBe slack variable; C is the punishment parameter.

For finding the solution above-mentioned optimization problem, structure Lagrange function:

In the formula: a=[a ₁, a ₂..., a _l] ^T, a _iBe the Lagrange multiplier.

Thereby, can by

The solving-optimizing problem.Behind cancellation w, the ξ, following formula can be expressed as following system of linear equations:

$\left[\begin{array}{cc}0& 1_v^T\\ 1_v& \mathrm{\Ω}+\frac{1}{C}I\end{array}\right]\left[\begin{array}{c}b\\ a\end{array}\right]=\left[\begin{array}{c}0\\ y\end{array}\right]$

In the formula: y=[y ₁, y ₂..., y _l] ^T1 _v=[1,1 ..., 1] ^TA=[a ₁, a ₂..., a _l] ^TI is l rank unit matrix; I, j=1,2 ..., l, K (x _i, x _j) be kernel function.

By above Solving Linear a, b, and will

Substitution formula (1) gets regression equation, also is the acoustic emission source location model:

(3) algorithm of described location model is quoted radial basis kernel function K (x _i, x _j) support least square method supporting vector machine to process nonlinear problem, with the inner product operation of high-dimensional feature space

A kernel function K (x with the former input space _i, x _j) equivalence, avoided in higher dimensional space, directly calculating the dimension disaster problem that causes.Because radial basis (RBF) kernel function has preferably statistic property, parameter is few, good classification effect, so the present invention adopts RBF nuclear:

$K(x,x_i)=\exp (-\frac{{\left|\right|x-x_i\left|\right|}^2}{2{\mathrm{\σ}}^2}),$ σ＞0

Above-mentioned location model parameters C and the parameter in the kernel function have determined study and the generalization ability of algorithm to a great extent, and it is a key issue of algorithm of support vector machine.The present invention adopts blanket k-folding cross validation method to determine parameter in the vector machine and the parameter in the radial basis kernel function, and it provides best half-way house between calculation cost and reliable parameter estimation, and it can prevent the study phenomenon.K-folding cross validation method specifically comprises the steps:

(3-1) training sample set is divided into randomly k mutually disjoint subset, the size of every subset about equally;

(3-2) utilize k-1 training subset, one group of given parameter is set up regression model, utilize the performance of the prediction square error RMSE evaluate parameter of last remaining subset;

(3-3) repeat k time according to above process, therefore every subset all has an opportunity to test, mean value according to the prediction square error that obtains after k the calculating estimates to expect extensive error, selects at last the parameter of one group of optimum according to the extensive error of expectation of estimating.

(4) model be input as following FEATURE PARAMETERS OF ACOUSTIC EMISSION: rise time, Ring-down count, amplitude, duration and signal intensity; Model is output as the coordinate figure of acoustic emission source position.

For precision and the efficient that improves location model, these characteristics of Acoustic Emission parameters are made up, by the learning efficiency comparison of different models, determine more excellent location model.This invention compares research to following 5 kinds of assembled schemes: 1) amplitude, rise time, duration and Ring-down count; 2) amplitude, rise time, duration and signal intensity; 3) amplitude, duration, Ring-down count and signal intensity; 4) rise time, duration, Ring-down count and signal intensity; 5) amplitude, rise time, duration, Ring-down count and signal intensity.

Model Output rusults to each array mode is done linear regression with actual disconnected plumbous position, obtains both related coefficients, distinguishes the quality of institute's established model with this.Experimental studies have found that, adopt the second array mode, the learning efficiency of model is optimum, and for this reason, the location model that can select array mode 2 to make up carries out the location prediction of structural damage acoustic emission source.Determine the optimal combination mode of the FEATURE PARAMETERS OF ACOUSTIC EMISSION of mode input, obtain more excellent location model;

(5) utilize the above-mentioned location model of having succeeded in school, input the measured value of FEATURE PARAMETERS OF ACOUSTIC EMISSION, can realize damaging the quick location of acoustic emission source.

Beneficial effect: 1, the concrete for hydraulic structure damage acoustic emission source locating method based on least square method supporting vector machine of the present invention's announcement, the mistiming and the velocity of wave that do not need each passage of acoustic emission system, thereby overcome the practical experience that time difference positioning method relies on engineering technical personnel, utilized the threshold crossing technology to determine the deficiency of time of arrival etc.; 2, the positioning result of model of the present invention is substantially identical with actual disconnected plumbous position, thereby has overcome the rough shortcoming of area positioning method positioning result, provides a kind of modeling simple, calculating simplicity, and convergence is good, the new technology that bearing accuracy is high.

Description of drawings

The least square method supporting vector machine model schematic diagram of Fig. 1 the inventive method.

Embodiment

The below is elaborated to technical solution of the present invention, but protection scope of the present invention is not limited to described embodiment.

Embodiment: the present invention tests used mix proportion of concrete and is the concrete mix that adopts certain dam recommend adoption, and is as shown in table 1.

Table 1 mix proportion of concrete

Wherein cement is selected the 52.5# moderate heat cement, stone is pressed: Da Shi: middle stone: gravelstone=4:3:3, diameter range is respectively 4～8cm, 2～4cm and 0.5～2cm, after finishing, mixing concrete screens out the major diameter aggregate with 4cm square hole sieve, obtain the wet screening concrete, then water the right cylinder concrete for hydraulic structure test specimen that builds up 1 Ф 160mm * 150mm.

Selected instrument is mainly the SAMOS acoustic emission Acquisition Instrument that U.S. PAC company produces, and sensor is R6a type resonant mode calibrate AE sensor.

According to the pencil-lead of the ASTM specifications recommend method that fractures, disconnected lead comes the simulated sound emissive source on the concrete for hydraulic structure surface.In process of the test, guarantee that as far as possible the surperficial angle of lead for retractable pencil and concrete sample is about 30 °, apply the approximately equalised power lead for retractable pencil that fractures at every turn.Test is disconnected plumbous at 80 points altogether, every 3 times, 240 times altogether.When fractureing lead for retractable pencil, by six sensor connecting acoustic emission Acquisition Instruments reception signals of fixed position on the concrete surface at every turn.

Utilize least square method supporting vector machine to set up the acoustic emission source location model based on described acoustic emission signal, model structure as shown in Figure 1.The input x of model is the characteristics of Acoustic Emission parameter, and output y is the acoustic emission source position.Its nonlinear relationship expression formula is:

$y\left(x\right)=\underset{i=1}{\overset{l}{\mathrm{\Σ}}}{a}_{i}K(x,x_i)+b$

X is the input characteristics of Acoustic Emission parameter of model in the formula; Output y is the acoustic emission source position; a _iBe the Lagrange multiplier; B is amount of bias; Space kernel function K (x _i, x _j) select radial basis (RBF) to examine:

$K(x,x_i)=\exp (-\frac{{\left|\right|x-x_i\left|\right|}^2}{2{\mathrm{\σ}}^2}),$ σ＞0

Adopt k-folding cross validation method to determine parameter, it is at calculation cost and provide best half-way house between the parameter estimation reliably, prevents the study phenomenon.

Because this test is taked to resolve lead and come the simulated sound method for transmitting signal, so the signal that produces is burst acoustic emission signal, following acoustic emission parameters is selected in the input of model: rise time, Ring-down count, amplitude, duration and signal intensity.The coordinate figure of acoustic emission source position makes up the training sample set of model as the output of model.

For precision and the efficient that improves location model, these characteristics of Acoustic Emission parameters are made up, by the learning efficiency comparison of different models, determine more excellent location model.This invention compares research to following 5 kinds of assembled schemes: 1, amplitude, rise time, duration and Ring-down count; 2, amplitude, rise time, duration and signal intensity; 3, amplitude, duration, Ring-down count and signal intensity; 4, rise time, duration, Ring-down count and signal intensity; 5, amplitude, rise time, duration, Ring-down count and signal intensity.

Model Output rusults to each array mode is done linear regression with actual disconnected plumbous position, obtains both related coefficients, distinguishes the quality of institute's established model with this.Find by comparative analysis, adopt the second array mode, the learning efficiency of model is optimum.Combination 1 and combination 3 fitting effect are more or less the same, and the fitting effect of model is also relatively good, and make up 4 poor with combination 5 fitting effect.For this reason, the location model of Selection and Constitute mode 2 structures carries out the location prediction of structural damage acoustic emission source.

Utilize the above-mentioned good model of having learnt, the measured value of input FEATURE PARAMETERS OF ACOUSTIC EMISSION gets final product the quick location that implementation structure damages acoustic emission source.Find according to test simulation, utilize the least square method supporting vector machine technology to carry out the precision of acoustic emission location high, the positioning result of model is substantially identical with actual disconnected plumbous position.

As mentioned above, although represented and explained the present invention with reference to specific preferred embodiment, it shall not be construed as the restriction to the present invention self.Under the spirit and scope of the present invention prerequisite that does not break away from the claims definition, can make in the form and details various variations to it.

Claims (4)

1. a concrete for hydraulic structure damages acoustic emission source locating method, it is characterized in that may further comprise the steps:

(1) adopts disconnected lead stamp to intend burst acoustic emission signal on the concrete for hydraulic structure surface, carry out the acoustic emission artificial excitation source location test of concrete for hydraulic structure damage, the detection data of acquisition and processing test specimen damage FEATURE PARAMETERS OF ACOUSTIC EMISSION and sound source position;

(2) set up damage acoustic emission source location model based on least square method supporting vector machine, be input as described FEATURE PARAMETERS OF ACOUSTIC EMISSION, be output as the acoustic emission source position;

(3) algorithm of described location model is quoted the radial basis kernel function and is supported least square method supporting vector machine to process nonlinear problem, and adopts k-folding cross validation method to determine parameter in the vector machine and the parameter in the radial basis kernel function;

(4) determine the optimal combination mode of the FEATURE PARAMETERS OF ACOUSTIC EMISSION of mode input, obtain more excellent location model;

(5) utilize the described location model of succeeding in school, input the measured value of FEATURE PARAMETERS OF ACOUSTIC EMISSION, can realize damaging the quick location of acoustic emission source.

2. concrete for hydraulic structure according to claim 1 damages acoustic emission source locating method, it is characterized in that: the described k-folding of step (3) cross validation method specifically comprises the steps:

(3-1) training sample set is divided into randomly k mutually disjoint subset, the equal and opposite in direction of every subset;

(3-2) utilize k-1 training subset, one group of given parameter is set up regression model, utilize the performance of the prediction square error RMSE evaluate parameter of last remaining subset;

(3-3) repeat k time according to above process, therefore every subset all has an opportunity to test, mean value according to the prediction square error that obtains after k the calculating estimates to expect extensive error, selects at last the parameter of one group of optimum according to the extensive error of expectation of estimating.

3. concrete for hydraulic structure according to claim 1 damages acoustic emission source locating method, and it is characterized in that: described FEATURE PARAMETERS OF ACOUSTIC EMISSION comprises rise time, Ring-down count, amplitude, duration and signal intensity.

4. concrete for hydraulic structure according to claim 1 damages acoustic emission source locating method, it is characterized in that: the optimal combination mode of determining the FEATURE PARAMETERS OF ACOUSTIC EMISSION of mode input in the step (4), that the model Output rusults of each array mode is done linear regression with actual disconnected plumbous position, obtain both related coefficients, distinguish the quality of institute's established model with this.

Images