CN101458259A - Sensor setting method for supporting failure prediction - Google Patents

Abstract

The invention relates to a sensor setup method capable of supporting failure prediction. The method is technically characterized by comprising two steps: sensor number optimization and sensor location optimization: firstly, performing the sensor number optimization step: obtaining a minimum acceleration peak value of a signal according to parameters of an instrument, obtaining an incident wave propagation limit distance threshold according to a wave amplitude attenuation formula of incident wave and obtaining the minimum number n of the required sensors; and then obtaining the sensor arrangement location by a finite element model, and adjusting the number and the location of the sensors according to the incident wave propagation limit distance threshold. The method can help realize that parameter information which reflects real structure state is maximally obtained with the minimum sensor number. The method adapts to setup complexity of the sensors for structural health monitoring and damage detection, and can better meet the needs of the actual engineering application. The invention can provide a sensor setup standard for the structural damage detection and the structural health monitoring; and help configure a detection system or a monitoring system more economically and reasonably.

Description

Support the sensor setting method of failure prediction

Technical field

The present invention relates to a kind of sensor setting method of supporting failure prediction, belong to great product and great installation forecasting technique in life span field.

Background technology

For the damage check and the health monitoring of structure adopt sensor to obtain the parameter information of reflect structure time of day, with damage and the healthy failure prediction that reaches structure, this is at present in the common method that adopts of great product and great installation forecasting technique in life span field.For this reason, study a kind of sensor and set up standard, make the detection or the monitoring system that are disposed reasonable more economically, adapt, and to satisfy the engineering practical application request be necessary with the complicacy of sensor setting in monitoring structural health conditions and the damage check.

Over more than 20 year, people have carried out extensive studies to the optimization Layout Problem of sensor, many measuring point layout optimization algorithms have been proposed, but aspect number of sensors optimization, consider less, in general, the sensor of laying in the structure is many more, and the information of the structure that is collected is just detailed more, and the precision of kinetic parameter identification is just good more.Yet the quantity of sensor often is subjected to the restriction of aspects such as economic factors and structure operation state, can not be on all degree of freedom of total placement sensor, this also is unpractical.Therefore need make the standard that meets engineering reality to the usage quantity problem of sensor.

Position optimization be provided with aspect representative algorithm be effective independent method and kinergety method.The shortcoming of effective independent method is unusual for fear of matrix, and the number of target mode will equal the number of sensor, and sensor not necessarily is distributed in the bigger position of modal strain energy, thereby may cause losing of information.The basic thought of kinergety method is that the response that has on the degree of freedom of big modal strain energy is also bigger, the step of placement sensor and effective independent method are similar, main difference point is: the objective function of kinergety method is to ask the modal strain energy maximum, and the objective function of effective independent method then is to ask target modal vector maximum linear irrelevant.Kinergety method height depends on the division of finite element grid, if divide slightlyer, then sensor also will distribute far.That divides is thinner, and it is near excessively that sensor will distribute, thereby also can lose the parameter information of important reflect structure time of day.

Summary of the invention

The technical matters that solves

For fear of the deficiencies in the prior art part, the present invention proposes a kind of sensor setting method of supporting failure prediction, has solved the sensor setting method problem in active service bridge failure prediction and the life appraisal.

Technical scheme

Thought of the present invention is: at first utilize the elastic wave propagation principle to determine number of sensors; Effective independence-drive point residual error method of utilizing this case to propose is then carried out sensing station optimization setting, the final sensor optimization plan of establishment that forms a cover support failure prediction, the parameter information of reflect structure time of day is farthest obtained in realization with minimum number of sensors.

Technical characterictic of the present invention comprises two steps: number of sensors optimization and sensing station optimization; Elder generation's number of executions optimization step:

(1) according to the range A of instrument, the Y-PSNR threshold values p of precision η and surveying instrument obtains signal minimum acceleration peak value | a _Min| 〉=A * η * 10 ^0.05P

(2) utilize the wave amplitude decay formula of incident wave $S=\underset{i}{\overset{j}{\mathrm{\Π}}}\left(S_{i,j}P\right)$ And S ^D _a=| a _Min| obtain incident wave and propagate critical distance threshold values D; A represents the incident wave wave amplitude; P=-e wherein ^-ikD, D is an incident wave propagation critical distance threshold values,

K is the structural section shearing factor, and i, j represent the finite element model node serial number, the wave amplitude pad value of the incident wave between the S representative structure full-size two-end-point;

(3) obtain the minimum number n of required sensor, $n=\left[\frac{L}{D}\right];$ L is a biggest size of element for the treatment of geodesic structure;

Continue the position optimization step according to The above results then:

(4) utilize the minimum number n of finite element model and sensor, calculate E _DPR=Φ [Φ ^TΦ] ^-1Φ ^TC _DPR, obtain the transducer arrangements position; Wherein Φ is a modal matrix for the treatment of the n*N rank of geodesic structure, and n is the minimum number of sensor, and N is total number of degrees of freedom, of structural finite element model to be measured, C _DPRBe effective independence-drive point residual error coefficient matrix, $c_{{\mathrm{DPR}}_i}=\underset{j=1}{\overset{N}{\mathrm{\&Sigma;}}}\frac{{\mathrm{\&Phi;}}_{\mathrm{<figure-callout\; id="2"\; label="ij"\; filenames="A200710199216E00142.png,A200710199216E00151.png"\; state="\{\{state\}\}">ij</figure-callout>}}^2}{{{\mathrm{\&omega;}}^2}_j},$ φ _IjBe the element among the Φ, ω _jIt is the model frequency for the treatment of geodesic structure;

(5) check maximum transducer spacing,, then between this maximum sensor spacing, increase by 1 sensor, calculate E once more if propagate critical distance threshold values D greater than incident wave _DPR=Φ [Φ ^TΦ] ^-1Φ ^TC _DPRObtain the new position of sensor;

(6) repeat previous step when all the sensors spacing is propagated critical distance threshold values D less than incident wave, number of sensors and sensing station are determined.

Beneficial effect

Sensor optimization method to set up proposed by the invention can realize farthest obtaining with minimum number of sensors the parameter information of reflect structure time of day.The complicacy of sensor setting adapts in this method and monitoring structural health conditions and the damage check, more can satisfy the engineering practical application request.The present invention can provide a kind of sensor to set up standard for the damage check and the health monitoring of structure, and makes the detection or the monitoring system that are disposed reasonable more economically.

Description of drawings

Fig. 1: method flow diagram

Fig. 2: embodiment bridge structure Benchmark model;

A: vertical view; B: side view

Fig. 3: Benchmark model shearing wave acceleration peak value die-away curve

Fig. 4: Benchmark model sensor arrangenent diagram

Fig. 5: the variable damage model of high precision truss

Fig. 6: the finite element model of the variable damage model of high precision truss

Fig. 7: the sensor optimization arrangement plan of three kinds of algorithms

Fig. 8: vibration shape fitted figure

A a: first order mode (Z direction); B: two first order modes (directions X);

C: three first order modes (Z direction); D: the quadravalence vibration shape (directions X)

Fig. 9: the mean square deviation of three kinds of sensor configuration relatively

Figure 10: Fisher information battle array determinant figure

The a:X direction; The b:Z direction

Figure 11: mode guarantees the comparison of criterion

Embodiment

Now in conjunction with the accompanying drawings the present invention is further described:

The number of sensors optimization method

If the range of instrument is A, precision is η, and then the absolute error limit value is A * η.The computing formula of Y-PSNR threshold values P is: $P={10\log }_{10}(a_{\mathrm{<figure-callout\; id="2"\; label="Sig"\; filenames="A200710199216E00142.png,A200710199216E00151.png"\; state="\{\{state\}\}">Sig</figure-callout>}}^2/a_{\mathrm{Noi}}^2),$ P can be obtained by the accuracy of instrument statistics during practical application.

Try to achieve signal minimum acceleration peak value: | a _Min| 〉=A * η * 10 ^0.05P

According to revised Timoshenko beam theory, the transverse movement of beam can be expressed as

$\mathrm{\&kappa;AG}\frac{{\&PartialD;}^2{v}_s}{\&PartialD;x^2}=\mathrm{\&rho;A}\frac{{\&PartialD;}^2(v_b+v_s)}{\&PartialD;{\mathrm{<figure-callout\; id="2"\; label="t"\; filenames="A200710199216E00142.png,A200710199216E00151.png"\; state="\{\{state\}\}">t</figure-callout>}}^2}$

$\mathrm{\&kappa;AG}\frac{{\&PartialD;v}_s}{\&PartialD;x}=-\mathrm{EI}\frac{{\&PartialD;}^2{v}_b}{\&PartialD;x^3}+\mathrm{\&rho;I}\frac{{\&PartialD;}^3{v}_b}{\&PartialD;x{\&PartialD;\mathrm{<figure-callout\; id="2"\; label="t"\; filenames="A200710199216E00142.png,A200710199216E00151.png"\; state="\{\{state\}\}">t</figure-callout>}}^2}+\mathrm{\&rho;I}\frac{{\&PartialD;}^3{v}_s}{\&PartialD;x{\&PartialD;\mathrm{<figure-callout\; id="2"\; label="t"\; filenames="A200710199216E00142.png,A200710199216E00151.png"\; state="\{\{state\}\}">t</figure-callout>}}^2}$

κ is the cross section shearing factor in the formula, and A is the beam section area, and G is the material modulus of shearing, and ρ is a density of material, and v (x, t)=v _s(x, t)+v _b(x t) is lateral shift, v _sAnd v _bBe respectively and shear and the crooked lateral shift that causes.Last two formulas are carried out finding the solution behind the Fourier transform

$\left\{\begin{array}{c}{\hat{v}}_s(x,\mathrm{\&omega;})={\mathrm{\&alpha;}}_{1}a\left(\mathrm{\&omega;}\right)e^{-i{k}_{1}x}+{\mathrm{\&alpha;}}_{1}d\left(\mathrm{\&omega;}\right)e^{j{k}_{1}x}\\ {\hat{v}}_b(x,\mathrm{\&omega;})=a\left(\mathrm{\&omega;}\right)e^{-i{k}_{1}x}+d\left(\mathrm{\&omega;}\right)e^{j{k}_{1}x}\end{array}\right.$

^ represents the amount in the frequency domain in the formula, ${\mathrm{<figure-callout\; id="1"\; label="k"\; filenames="A200710199216E00151.png,A200710199216E00172.png"\; state="\{\{state\}\}">k</figure-callout>}}_1=\frac{\mathrm{\&omega;}}{\sqrt{E/\mathrm{\&rho;}}}\sqrt{\frac{1}{2}[(1+\frac{E}{\mathrm{\&kappa;G}})+\sqrt{{(1+\frac{E}{\mathrm{\&kappa;G}})}^2+\frac{4\mathrm{EA}}{\mathrm{I\&rho;}{\mathrm{\&omega;}}^2}}]};$ A (ω) expression incident wave wave amplitude; D (ω) expresses the ejected wave wave amplitude.With these amplitudes is fundamental unknown variables, and shearing force can be expressed as

$\hat{V}=\mathrm{i\&kappa;}{\mathrm{<figure-callout\; id="1"\; label="GAk"\; filenames="A200710199216E00151.png,A200710199216E00172.png"\; state="\{\{state\}\}">GAk</figure-callout>}}_1{\mathrm{\&alpha;}}_1[a\left(\mathrm{\&omega;}\right)e^{-i{k}_{1}x}-d\left(\mathrm{\&omega;}\right)e^{i{k}_{1}x}]$

Following formula is expressed as matrix form

[d(w)]＝[e ^-2ikI][a(w)]+[i(κGAka) ^-1e ^-ikI][V ^J]

Order

$S=\left[e^{-2{\mathrm{ik}}_{1}I}\right],$ $\hat{s}=\left[i{\left(\mathrm{kGA}{k}_1{a}_1\right)}^{-1}{e}^{-{\mathrm{ik}}_{1}I}\right]\left[{\hat{V}}^J\right]$

Then matrix form can be simplified shown as

$\hat{d}=S\hat{a}+\hat{s}$

In the formula: S,

Be called local scattering matrix, localized source matrix.

The bridge Benchmark model that provides with Fig. 2 is an example.Select the east INV306U-5160 of institute intelligent signal collection Treatment Analysis instrument (amplitude error 1%) acquired signal for use, adopt the auto-power spectrum method to detect " adding " periodical high-frequency wave component of acquired signal.Scattering relation when the derivation elastic wave is propagated in structure.

Get No. 6 node analysiss, outgoing wave d ^6q(q=4.5.8), incident wave a ^6q(p=4.5.8) relation of wave amplitude can be expressed as

$\left\{\begin{array}{c}{\hat{d}}^{64}\\ {\hat{d}}^{65}\\ {\hat{d}}^{68}\end{array}\right\}=\left[\begin{array}{ccc}{S}^{464}& S^{465}& S^{468}\\ S^{564}& S^{565}& S^{568}\\ S^{864}& S^{865}& S^{868}\end{array}\right]\left\{\begin{array}{c}{\hat{a}}^{64}\\ {\hat{a}}^{65}\\ {\hat{a}}^{68}\end{array}\right\}+\left\{\begin{array}{c}{\hat{s}}^{64}\\ {\hat{s}}^{65}\\ {\hat{s}}^{68}\end{array}\right\}$

S ^Q6pThe expression ripple arrives the scattering matrix that q is ordered from the p point through node 6.Notice the outgoing wave of Liang Yiduan, also can be considered the incident wave of the other end, the relation between them can be expressed as

a ^JK＝Pd ^KJ

In the formula: J, K represent the two ends of beam, P=-e respectively ^-ikIObservation is sent to a train wave of node 14 from node 6, and is the shortest along path 6-8-10-12-14 propagation distance, can be expressed as along the scattering matrix of the ripple of this propagated

$S_{6-14}={\mathrm{\&Pi;}}_{l=3}^5\left(S_{2l,2l+2}P\right)$

In the formula, l represents the wave propagation distance.As long as the ripple signal peak along this propagated is not less than | a _Min|, signal processing system just can correctly be discerned this signal.Because the shear stress of node is proportional to the transverse acceleration of node, so the attenuation degree that can directly determine acceleration peak value with scattering matrix S is worked as Sa=|a _Min| the time, the l that calculates is called ripple and propagates critical distance threshold values D.

Suppose that sensor evenly arranges, if transducer arrangements limit spacing is D, the physical dimension maximal value for the treatment of geodesic structure is L, but then the required minimum number of sensors of preresearch estimates is $n=\left[\frac{L}{D}\right]$ ([] expression forward rounds in the formula).

The sensing station optimization method:

Support the output information of the sensor of failure prediction can do following expression:

$y_s=\mathrm{\&phi;q}+w=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{q}_i{\mathrm{\&phi;}}_i+w$

In the formula, Φ by detection mode by the modal matrix of the n*N after the candidate's measuring point reduction, n is a sensor position candidate number, q is a generalized coordinate, w is a variances sigma ²Static white Gaussian noise, φ _iI column vector of the Φ that is, q _iBe that the vibration shape participates in coefficient.

If the optimum estimate process of real generalized coordinate q does not estimate that for having partially effectively then the covariance matrix J of q estimated bias is:

$J=E\left[(q-\stackrel{\&OverBar;}{q}){(q-\stackrel{\&OverBar;}{q})}^T\right]={\left[\frac{1}{{\mathrm{\&sigma;}}^2}{\mathrm{\&Phi;}}^T\mathrm{\&Phi;}\right]}^{-1}=Q^{-1}$

Q represents the Fisher information matrix in the formula.

Fisher information battle array determinant maximization method is that a certain position candidate of sensor is removed, and calculates the determinant of Q then, when the Q determinant that obtains is maximum, shows that this position candidate is minimum to the linear independence influence of target mode.

By calculating the effective independent allocation vector E of expression candidate point to the linear independence contribution of modal matrix _DRealize

E _D＝[Φψ] ²λ ^-1I _k

In the formula, ψ is the proper vector of Q, and λ is corresponding eigenvalue matrix, I _kIt is the summation of capable all coefficients of k.Under the prerequisite that guarantees Q determinant maximum after each iteration, get rid of E _DThe candidate point position of coefficient minimum, thus the position of the sensor of given number obtained.

The step of kinergety method placement sensor and effective independent method are similar, and the main difference point is that the kinergety method is can maximize rather than Fisher information matrix determinant maximum is sought the optimization position of sensor by structure motion.

Define effective drive point residual error coefficient

Be the mode motion energy of element stiffness, promptly

$c_{{\mathrm{DPR}}_i}=\underset{j=1}{\overset{N}{\mathrm{\&Sigma;}}}\frac{{\mathrm{\&Phi;}}_{\mathrm{<figure-callout\; id="2"\; label="ij"\; filenames="A200710199216E00142.png,A200710199216E00151.png"\; state="\{\{state\}\}">ij</figure-callout>}}^2}{{{\mathrm{\&omega;}}^2}_j}$

Use C _DPRThe effective independent allocation matrix of weighting has

E _DPR＝Φ[Φ ¹Φ] ^-1Φ ¹C _DPR

Effectively independence-drive point residual error method is with effective independent allocation matrix E _DPRIterative, each iteration deletion E _DPRClinodiagonal element minimum value corresponding sensor position keeps its off diagonal element higher value corresponding sensor position, up to the sensor number of setting.Whole iterative process utilizes MATLAB language compilation program to calculate, and program itself is the process of a continuous iteration optimizing, in iterative process optimizing and to the end convergence result.Because it is that target is optimized that this algorithm keeps the maximum linear independence with the modal strain energy maximum and the vibration shape simultaneously, so both overcome the shortcoming that effective independent method institute cloth Sensor section is positioned at low-yield district, also overcome the shortcoming that kinergety method height relies on the division of finite element grid.Moreover the finite element modeling error is smaller for the influence of the vibration shape, particularly low order mode, and this method only need be utilized the low order mode data, and this just greatly reduces the influence of modeling error.

Determine the final sensor optimization plan of establishment:

With the constraint condition of transducer arrangements limit spacing, the sensing station that provides is carried out location test as effective independence-drive point residual error method (EFI-DPR) or sensitivity analysis method.If the spacing of any two sensors spacing that oversteps the extreme limit is arranged, then on the determined number of sensors of this computing basis, increase a sensor, use sensor optimization to arrange that algorithm rearranges sensor and verification scheme again, if need not to increase number of sensors, then this arrangement is that number of sensors is a position final optimization pass scheme.

It is Fig. 1 that sensor network optimization is provided with software workflow:

Import data (instrument range, precision, structured material, size, finite element model parameter and basic modal analysis result data) → determine that tentatively sensor minimum number → selections preferred arrangement method carries out position optimization → spacing and check → (as needs, supplemental amount is also carried out position optimization again) → determine final sensor optimization plan of establishment.

The foregoing description operation workflow: the composite signal that the sinusoidal wave Y-PSNR that produces with white noise stack back of single frequency that adopts the Matlab software emulation is N carries out statistical study to the anti-interference maximum peak value signal to noise ratio (S/N ratio) of auto-power spectrum method threshold values P.Repeat 10,000 times auto-power spectrum method anti-interference analysis, with spectral line peak value among each result more than one as once reporting by mistake, add up its peak value wrong report number of times.Statistics shows when P=-5db, and the wrong report number of times is less than 10 times, and rate of false alarm is less than 1 ‰, thus with P=-5db as auto-power spectrum method analysis result reliability decision threshold.When choosing range be ± 10m/s ²The time, substitution formula (5.1-2), try to achieve desired signal minimum acceleration peak value | a _Min| 〉=0.056m/s ²Consider the structural damping phenomenon, according to the hysteresis Damping Theory, when resonance takes place in structure, the hysteresis ratio of damping

Can utilize itself and viscous damping to calculate than the relation of ξ:

The material constant of Benchmark model sees Table 1.

Table 1 structured material constant

Choosing a row frequency is 2000Hz, and acceleration is 0.13m/s ²Incident wave, can try to achieve by aforementioned computing formula, decay to 0.45 times of (0.06m/s of former acceleration when the crest value acceleration ²) time, its propagation distance is about 5.7 meters, is 5.7 meters so ripple is propagated the critical distance threshold values.For satisfying the needs of structural damage detection, get transducer arrangements limit space D=5.7 meters.

Determine the final arrangement of sensor:

For selected detection system, according to evenly arranging principle, get transducer arrangements limit space D=5.7 meters, then above-mentioned bridge structure Benchmark model being carried out vibration-testing needs at least $n=\left[\frac{(18+6)\&times;0.3048m}{5.7m}\right]=2,$ I.e. 2 sensors.

It is 20,126 that the quantity substitution EFI-DPR method of tentatively determining is tried to achieve the transducer arrangements node number.Sensing station is carried out the check of limit spacing, satisfy condition, promptly this scheme is the final arrangement of sensor of Benchmark model vibration-testing.The transducer arrangements position is indicated on Benchmark model node diagram, as shown in Figure 4:

The variable damage model verification experimental verification of high precision truss:

In order to verify the superiority of the set scheme of the present invention, the variable damage model of one high precision truss has been carried out sensor optimization experiment has been set, and be least-mean-square-error criterion with three kinds of sensor optimization comparison criterions respectively, best criterion of noise robustness and mode guarantee that criterion compares three kinds of methods of distributing rationally.

First kind of comparison criterion---least-mean-square-error criterion (Mean square error, MSE), mean square deviation between the vibration shape by calculating finite element model and the vibration shape of three spline interpolation matches is assessed the ability of the sensor arresting structure dynamic response that each optimization method arranges, mean square deviation is more little, and the ability of expression arresting structure dynamic response is strong more.Being calculated as follows shown in the formula of mean square deviation:

${\mathrm{\&sigma;}}_{\mathrm{TMSE}}=\underset{i=1}{\overset{\mathrm{<figure-callout\; id="1"\; label="N"\; filenames="A200710199216E00151.png,A200710199216E00172.png"\; state="\{\{state\}\}">N</figure-callout>}}{\mathrm{\&Sigma;}}}\frac{\frac{1}{{\mathrm{\&sigma;}}_i}\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{({\mathrm{\&Phi;}}_{\mathrm{ij}}^{\mathrm{CS}}-{\mathrm{\&Phi;}}_{\mathrm{ij}}^{\mathrm{FE}})}^2}{n}\&CenterDot;$

σ wherein _iIt is the standard deviation of i first order mode.The comparative result of MSE criterion such as table 2 and Fig. 9.

Fig. 8 is with effective drive point residual error method, the structure vibration shape and the finite element software ANSYS that obtain through sensor optimization configuration back match analyze the vibration shape that obtains, both vibration shapes are very approaching as can be seen from Figure, illustrate that the modal parameter of the sensor measurement of effective drive point residual error method configuration can reflect by the real conditions of geodesic structure.Table 2 and Fig. 9 illustrate that the error mean square difference of effective independence-drive point residual error method all hangs down one, two order of magnitude than other two kinds of methods, and from the result, EFI-DPR is better than other two kinds of methods.

Second kind of comparison criterion---the best criterion of noise robustness, this criterion is used for estimating noise to measuring the influence of modal parameter, it is based on the relevant good more principle of the high more noiseproof feature of signal intensity of characteristics of mode, number percent with Fisher information battle array determinant and original determinant behind the deletion candidate sensing station is assessed result such as Figure 10.

As can be seen from Figure 10, along with the increase of sensor deletion number, EFI-DPR shows higher quantity of information than other two kinds of methods, and promptly to catch the ability of parameter information stronger.

The third comparison criterion---mode guarantees criterion, and Kind of Modal Confidence Factor MAC submatrix is a fine instrument of estimating the modal vector space angle of cut, and its computing formula is as follows

${\mathrm{MAC}}_{\mathrm{ij}}=\frac{{\left[{\mathrm{\&Phi;}}^{\left(i\right)T}{\mathrm{\&Phi;}}^{\left(j\right)}\right]}^2}{{\mathrm{\&Phi;}}^{\left(i\right)T}{\mathrm{\&Phi;}}^{\left(i\right)}{\mathrm{\&Phi;}}^{\left(j\right)T}{\mathrm{\&Phi;}}^{\left(j\right)}}$

Φ in the formula ⁽ⁱ⁾And Φ ^(j)Be respectively i rank and j rank modal vector.The nondiagonal element of MAC is big more, and the correlativity of vibration shape vector is strong more.The root mean square that this algorithm adopts nondiagonal element is estimated the quality of each sensor arrangement method as standard, and comparative result sees Table 3 and Figure 11.As can be seen from Table 3, the MAC off diagonal element of EFI-DPR is than low one, two order of magnitude of MAC off diagonal element of other two kinds of methods, and low equally two orders of magnitude nearly of mean square deviation, and this proves absolutely that EFI-DPR is better than other two kinds of methods.

Table 3 mode guarantees that criterion relatively

Above-mentioned numerical value and test findings show that all the present invention implements easily, and the sensor optimization method to set up that is proposed can be determined needed number of sensors and best position simultaneously, more can satisfy the engineering practical application request.The present invention can provide a kind of sensor to set up standard for the damage check and the health monitoring of structure, and makes the detection or the monitoring system that are disposed reasonable more economically.

Claims (1)

1. a sensor setting method of supporting failure prediction is characterized in that comprising number of sensors optimization step and biography

Sensor position optimization step; Elder generation's number of executions optimization step:

(1) according to the range A of instrument, the Y-PSNR threshold values p of precision η and surveying instrument obtains signal minimum acceleration peak value | a _Min| 〉=A * η * 10 ^0.05P

(2) utilize the wave amplitude decay formula of incident wave $S=\underset{i}{\overset{j}{\mathrm{\&Pi;}}}\left(S_{i,j}P\right)$ And S ^DA=|a _Min| obtain incident wave and propagate critical distance threshold values D; A represents the incident wave wave amplitude; P=-e wherein ^-ikD, D is an incident wave propagation critical distance threshold values,

K is the structural section shearing factor, and i, j represent the finite element model node serial number, the wave amplitude pad value of the incident wave between the S representative structure full-size two-end-point;

(3) obtain the minimum number n of required sensor, $n=\left[\frac{L}{D}\right];$ L is a biggest size of element for the treatment of geodesic structure; Continue the position optimization step according to The above results then:

(4) utilize the minimum number n of finite element model and sensor, calculate E _DPR=Φ [Φ ^TΦ] ^-1Φ ^TC _DPR, obtain the transducer arrangements position; Wherein Φ is a modal matrix for the treatment of the n*N rank of geodesic structure, and n is the minimum number of sensor, and N is total number of degrees of freedom, of structural finite element model to be measured, C _DPRBe effective independence-drive point residual error coefficient matrix, $C_{{\mathrm{DPR}}_i}=\underset{j=1}{\overset{N}{\mathrm{\&Sigma;}}}\frac{{\mathrm{\&Phi;}}_{\mathrm{ij}}^2}{{{\mathrm{\&omega;}}^2}_j},$ φ _IjBe the element among the Φ, ω _jIt is the model frequency for the treatment of geodesic structure;

(5) check maximum transducer spacing,, then between this maximum sensor spacing, increase by 1 sensor, calculate E once more if propagate critical distance threshold values D greater than incident wave _DPR=Φ [Φ ^TΦ] ^-1Φ ^TC _DPRObtain the new position of sensor;

(6) repeat previous step when all the sensors spacing is propagated critical distance threshold values D less than incident wave, number of sensors and sensing station are determined.

Images