CN101865789A - Fault detecting device of near field acoustic holography sound image mode identification and detecting method thereof - Google Patents

Abstract

The invention discloses a fault detecting device of near field acoustic holography sound image mode identification and a detecting method thereof in the field of industrial detection. The device comprises a microphone array, a scanner frame, a reference source, a clamped rib plate, an excitation source and a data acquiring system. The invention adopts sound pressure amplitude and distribution changes under mechanical normal and fault states, borrows ideas from applications of an image diagnosis technology in other fields and utilizes image processing, feature extracting and mode identifying technologies to process a near field acoustic holography image. Experiments show that the method is effective, vibration source positioning and identifying functions coordinated with the holographic imaging technology is proven, the image feather extracting and mode identifying technologies are combined with the acoustic imaging technology, thus broadening the application range of the acoustic imaging technology and bringing a set of effective acoustic fault diagnosis technologies and feasibility of being widely used in engineering thereof.

Description

Near field acoustic holography sound image pattern-recognition failure detector and detection method thereof

Technical field

What the present invention relates to is a kind of device and method of industrial detection technical field, specifically is a kind of near field acoustic holography sound image pattern-recognition failure detector and detection method thereof.

Background technology

Fault diagnosis technology based on vibration signal has limitation under some occasion, and mechanical noise is containing abundant machine state information, has the advantage of non-contact measurement, can partly substitute the means of vibration signal as fault diagnosis.The technology of utilizing noise signal to carry out fault diagnosis is referred to as the acoustics diagnose technology.Perfect acoustics diagnose technical requirement can either the effective recognition sound source position, can carry out spectrum analysis simultaneously.Traditional identification of sound source method mainly contains subjective estimate method, branch's running method, surface strength method, near field measurement method etc., but is subjected to the restriction of physical construction and working environment easily.Acoustical holography is rapid as a kind of technology development in recent years of identification of sound source, mainly is by measuring the acoustic pressure on the two-dimensional surface (holographic facet), using restructing algorithm to come the three-dimensional sound field (comprising acoustic pressure, the sound intensity and normal direction vibration velocity) on reconstruct sound source surface.The acoustical holography method has been utilized the strength information and the phase information of sound, and visual result can be easy to noise source is positioned, quantizes, and the route of transmission of energy display noise, and acoustical holography is carried out at frequency domain simultaneously, has also inherited the characteristics of frequency-domain analysis.Chinese patent application numbers 200910039395.5, denomination of invention is " the statistically optimal near-field acoustical holography used method and the method for operating thereof of the visual identification of air-conditioner noise sources ", discloses the application of a kind of acoustical holography in the identification of air-conditioning noise source; Chinese patent application number 03129405.7, denomination of invention is " method that adopts near field acoustic holography technology identification non-stationary sound source ", further expanded the scope that near field acoustic holography is used, by adopting the theoretical alternative traditional Fourier change technique of cyclo-stationary, select the physical quantity of spectral density function, proposed the near field acoustic holography technology of cyclo-stationary sound field as sound field rebuilding.

As a kind of identification of sound source algorithm, near field acoustic holography is with the obvious advantage, but the bibliographical information that is directly used in fault diagnosis seldom.Can obtain the distribution pattern of mechanical surface vibration velocity, acoustic pressure and the sound intensity behind the array acoustical signal process frequency domain near field acoustic holography, from figure, can obtain the position of vibration source under a certain frequency.

In image processing field, use very extensive based on the diagnostic techniques of image.Find through retrieval prior art, Chinese patent application numbers 200710069113.7, put down in writing a kind of " differentiating the method and apparatus of different variety green tea based on the texture analysis of multispectral image ", this technology obtains the image of different variety green tea by the 3CCD camera, calculate various green tea image texture characteristics and put into sorter and train classification, and then unknown green tea is carried out diagnostic classification.Reference forms a kind of method for diagnosing faults based on the near field acoustic holography sound image pattern-recognition based on the diagnostic techniques of image, can expand the application of acoustical holography and the development of promotion acoustics diagnose technology.

Summary of the invention

The present invention is directed to the prior art above shortcomings, a kind of near field acoustic holography sound image pattern-recognition failure detector and detection method thereof are provided, acoustic pressure distribution or vibration velocity distribution based on machinery a certain frequency under normal and malfunction can change, thereby in acoustic image, produce corresponding the variation, by acoustic image in noise source position, location, use the Feature Extraction Technology extraction sensitive features of Flame Image Process to characterize normally and the acoustic image under the failure condition, by the support vector machine in the pattern-recognition these proper vectors are trained classification and then carried out fault diagnosis and on-line monitoring then.

The present invention is achieved by the following technical solutions:

The present invention relates to a kind of failure detector of near field acoustic holography sound image pattern-recognition, comprise: scanning support, microphone array, reference source and data acquisition system (DAS), wherein: reference source is close to machinery, be arranged in the possible noise source position of machinery, scanning support is positioned at sound source one side, microphone array is fixed on the scanning support according to survey sound source sound field position by different way, can carry out side and top scanning survey to sound source, reference source microphone and scanning array microphone are connected with data acquisition system (DAS) simultaneously, storage of collected to time-domain signal and calculate testing result.

The present invention relates to the detection method of the failure detector of above-mentioned near field acoustic holography sound image pattern-recognition, may further comprise the steps:

The first step, the machinery with a plurality of noise sources is carried out normal divide and setting with the fail operation state, the situation ratio of vibration source number and position generation significant change is easier to judge from hologram during for fault;

Second the step, according to sound source face size (lx, ly) determine holographic measurement face size (Lx, Ly), and estimate sensitive frequency F scope determine holographic facet and source face apart from the sampling interval between Z and the holographic measurement face microphone (Δ x, Δ y);

Described definite holographic measurement face size is meant: Lx 〉=1.2*lx﹠amp; Ly 〉=1.2*ly

Be raising holographic reconstruction precision, Δ x≤λ/6, Δ y≤λ/6, Z≤λ/6 and Z 〉=max (Δ x, Δ y), wherein λ is the wavelength of being determined by F.

The 3rd step, the sound pressure signal P (t) in measurement mechanical under the condition of ground unrest unanimity under the normal and malfunction, and establishment sample storehouse S, carry out spectrum analysis and find sensitive frequency f (f≤F), each normal sound field sample, fault sound field sample are adopted near field acoustic holography.

The sound pressure signal of any point p can be expressed as the form of Helmholtz integral equation in the described sound field:

$p\left({\stackrel{\→}{r}}_p\right)=\underset{S}{\∫\∫}[p\left({\stackrel{\→}{r}}_Q\right)\frac{\∂G({\stackrel{\→}{r}}_p,{\stackrel{\→}{r}}_Q)}{\∂n}+\mathrm{j\ρck}{u}_n\left({\stackrel{\→}{r}}_Q\right)G({\stackrel{\→}{r}}_p,{\stackrel{\→}{r}}_Q)]\mathrm{dS}\left({\stackrel{\→}{r}}_Q\right)---\left(1\right)$

Wherein: S represents the sound source surface, and Q represents the border, U _nRepresentation is to vibration velocity, and G represents Green function.According to the suitable holographic algorithm of sound source structure choice: the acoustic pressure on the acoustical holography utilization measurement face of quadrature conformal structure is the convolution of source face acoustic pressure and Green function, fast two-dimensional fourier transformation is used for the Helmholtz equation, has realized rebuilding the distribution of acoustic pressure, vibration velocity and the sound intensity on the face of source by the acoustic pressure of holographic measurement face; The near field acoustic holography of arbitrary shape structure has BEM-based NAH and the equivalent source method based on numerical solution Helmholtz integral equation.

Calculate frequency sound source surface vibration velocity distribution matrix A when being F by near field acoustic holography, A is a complex matrix.Choose appropriate gray shade level n, matrix A is visualized as image B, and image B is the vibration velocity distribution pattern, position that therefrom can the main vibration source of identification.

The 4th step, the near field acoustic holography result is carried out feature extraction, structural attitude vector sample storehouse is convenient to sorter and is trained identification;

Described feature extraction is meant: adopt in the following dual mode any one:

A) directly the reconstruct matrix A is carried out svd and obtains singular value features:

$A=U{\mathrm{\ΛV}}^T=\underset{i=1}{\overset{k}{\mathrm{\Σ}}}{\mathrm{\α}}_i{u}_i{v}_i^T=\underset{i=1}{\overset{k}{\mathrm{\Σ}}}{\mathrm{\α}}_i{A}_i---\left(2\right)$

Wherein: k is the order of matrix A, α _iBe i singular value of matrix A, make α ₁〉=a ₂〉=... 〉=α _n, m singular value constitutive characteristic vector before getting is used for the training classification.

B) be to utilize image processing techniques to extract the texture statistics feature of image B:

The known image gray level is n, the second order joint probability density function P by asking for image (d θ) obtains gray level co-occurrence matrixes for i, j:

M(d，θ)＝[P(i，j，d，θ)] (3)

Wherein: function P has described image distance on the θ direction has the appearance of gray scale i and j respectively for a pair of pixel of d probability, d=1, θ gets 0 °, 45 °, 90 ° and 135 ° respectively, obtain the gray level co-occurrence matrixes of four direction, and ask for 12 kinds of textural characteristics coefficients based on gray level co-occurrence matrixes: angle second moment value, contrast value, correlation, entropy, variance yields, contrary gap value and mean value and variance yields and entropy, difference mean value, difference variance and difference entropy, and be used for pattern-recognition.

The 5th the step, obtain proper vector sample storehouse after, put into support vector machine and train classification, contrast the discrimination under the various conditions, obtain best classifier parameters: punishment parameters C and kernel function coefficient gamma, and feature extraction mode, the array signal that collects is in real time carried out feature extraction and adopts the sorter after optimizing to discern, realize fault diagnosis.

The present invention utilize machinery normal with malfunction under the variation of sound pressure amplitude and distribution, use for reference of the application of diagnostic imaging technology at other field, employing Flame Image Process, feature extraction and mode identification technology are handled the near field acoustic holography image.Experiment shows the validity of method, simultaneous verification cooperate vibration source location, the recognition function of holographic imaging technology, image treatment features extraction and mode identification technology are combined with the acoustic imaging technology, widen acoustic imaging The Application of Technology scope, form the feasibility of an effective Acoustic Based Diagnosis technology of cover and widespread use in engineering.

Description of drawings

Fig. 1 is a structural representation of the present invention.

Fig. 2 is exciting point position and floor synoptic diagram.

Fig. 3 is a schematic flow sheet of the present invention.

Fig. 4 is that embodiment tests the acoustic image comparison diagram.

Embodiment

Below embodiments of the invention are elaborated, present embodiment is being to implement under the prerequisite with the technical solution of the present invention, provided detailed embodiment and concrete operating process, but protection scope of the present invention is not limited to following embodiment.

As shown in Figure 1, present embodiment comprises: scanning support 1, microphone array 2, prop up floor 3, reference source 4, exciting source 5, data acquisition system (DAS) 6 and computer 7 admittedly, wherein: microphone array is arranged in the side of floor sound source, apart from sound source surface Z; Because floor exciting source excited frequency is identical, three vibration sources are relevant, adopt a reference source; Position by scanning support control microphone array forms the holographic measurement face; Gather the sound field information of floor sound source by data acquisition system (DAS) in each scanning step.

As shown in Figure 2, described exciting point position and floor: embodiment carries out 3 excitings to floor, and each exciting point distance is greater than the resolution of near-field holography; Floor props up on experimental stand admittedly.

As shown in Figure 3, said apparatus carries out work in the following manner:

Experiment is carried out in semianechoic room, adopts multi-point exciting to prop up the overriding noise source that floor is simulated machinery admittedly, as shown in Figure 2.The thickness of floor and rib all is 4mm; Floor area 640mm * 440mm; Rib 1 area 360mm * 40mm; Rib 2 area 560mm * 60m.The floor center is a true origin, exciting point 1 (180mm, 87mm), exciting point 2 (20mm, 87mm), exciting point 3 (120mm ,-87mm).

The first step, the machinery with a plurality of noise sources is carried out normal dividing and setting with the fail operation state:

Connect by force transducer between vibrator and the floor, simulate the normal and malfunction of each vibration source by measuring and adjust the exciting force size, as: vibration source 1 exciting force is in normal operating condition when changing between 6N～8.8N, be malfunction greater than the situation of 8.8N; Vibration source 2 exciting forces are in normal operating condition when changing between 4N～5.8N, be malfunction greater than the situation of 5.8N; Vibration source 3 exciting forces are in normal operating condition when changing between 5N～7.4N, be malfunction greater than the situation of 7.4N.Set 30 in normal condition sample altogether, 35 of various failure operation state sample.

Second the step, estimate sensitive frequency scope F, be lower than the first natural frequency 473Hz that props up floor admittedly.By sensitive frequency F determine holographic facet and source face apart from the sampling interval between Z and the holographic measurement face microphone (Δ x, Δ y), according to sound source face size determine holographic measurement face size (Lx, Ly);

As shown in Figure 1: because sound source is regular, embodiment adopts the FFT-based NAH based on the quadrature conformal structure; Owing to the restriction of acquisition instrument equipment, the method that embodiment adopts linear array scanning to measure is gathered acoustical signal simultaneously; Arrange the microphone linear array at distance sound source face Z=0.1m place, linear array is made of 9 microphones, and adjacent microphone is at a distance of Δ x=Δ y=0.1m; The sound source face is 0.64m * 0.44mm, scanning holography face 0.9m * 0.8m; Sound field is concerned with, and places a microphone as the reference source near vibration source 2; The acoustic field signal of each 10 passages of synchronous acquisition, and the volume coordinate position of record microphone.

The 3rd step, measurement mechanical under the condition of ground unrest unanimity normal with malfunction under sound pressure signal P (t), and set up sample storehouse S, set sensitive frequency f=300Hz, to each normal sound field sample, fault sound field sample employing near field acoustic holography.

The acoustic pressure of any point p can be expressed as the form of Helmholtz integral equation in the sound field:

$p\left({\stackrel{\→}{r}}_p\right)=\underset{S}{\∫\∫}[p\left({\stackrel{\→}{r}}_Q\right)\frac{\∂G({\stackrel{\→}{r}}_p,{\stackrel{\→}{r}}_Q)}{\∂n}+\mathrm{j\ρck}{u}_n\left({\stackrel{\→}{r}}_Q\right)G({\stackrel{\→}{r}}_p,{\stackrel{\→}{r}}_Q)]\mathrm{dS}\left({\stackrel{\→}{r}}_Q\right)---\left(1\right)$

Wherein: S represents the sound source surface, and Q represents the border, U _nRepresentation is to vibration velocity, and G represents Green function.According to the suitable holographic algorithm of sound source structure choice: the acoustic pressure on the acoustical holography utilization measurement face of quadrature conformal structure is the convolution of source face acoustic pressure and Green function, fast two-dimensional fourier transformation is used for the Helmholtz equation, has realized rebuilding the distribution of acoustic pressure, vibration velocity and the sound intensity on the face of source by the acoustic pressure of holographic measurement face; The near field acoustic holography of arbitrary shape structure has BEM-based NAH and the equivalent source method based on numerical solution Helmholtz integral equation.

Calculate frequency sound source surface vibration velocity distribution matrix A when being f by near field acoustic holography, A is a complex matrix.Choose appropriate gray shade level n, matrix A is visualized as image B, and image B is the vibration velocity distribution pattern, position that therefrom can the main vibration source of identification.

65 sound field samples are carried out FFT-based NAH calculate source image restructuring matrix and acoustic image, Fig. 4 a is the acoustic image under normal and the malfunction.Wherein can be referring to patent 03129405.7 based on the near field acoustic holography theory of algorithm of scanning method, patent name is " method that adopts near field acoustic holography technology identification non-stationary sound source ", steadily the cycle frequency of sound field is 0.The vibration velocity restructuring matrix is the complex matrix that comprises phase information, and gray level n get respectively 8,16,32,64 and 128 o'clock the visual acoustic image that changes into of matrix.Observe the vibration source position, with actual exciting position versus, because floor single order Mode Shape and influence of measurement error, vibration source and actual exciting position do not overlap, but can accurately locate the general location of main vibration source from the vibration velocity distributed image.

The 4th step, restructuring matrix is directly extracted singular value features, calculates the textural characteristics based on gray level co-occurrence matrixes to determining acoustic image after the gray level, form proper vector sample storehouse.

The 5th the step, obtain proper vector sample storehouse after, put into support vector machine and train classification, contrast the discrimination under the various conditions, obtain best classifier parameters: punishment parameters C and kernel function coefficient gamma, and feature extraction mode, the array signal that collects is in real time carried out feature extraction and adopts the sorter after optimizing to discern, realize fault diagnosis.

Use Libsvm to sample training classification, wherein as shown in table 1 based on the classification results of textural characteristics, as shown in table 2 based on the classification results of singular value features:

Table 1 is based on discrimination contrast (%) under each gray level of statistical nature

Can see that by table 1 discrimination is the highest when gray level is 32,45 ° of directions, reach 90.7%, punish parameters C=2048, kernel function coefficient gamma=0.0078125 this moment.

Table 2 is based on the discrimination contrast (%) of singular value

By table 2, the singular value number is that 8 o'clock recognition effects preferably reach 92.3%, punishment parameters C=32768, kernel function coefficient gamma=0.0078125.

Under the present embodiment condition, it is best to choose the recognition effect that preceding 8 big singular value constitutive characteristic vectors train classification to obtain, and has reached 92.3%, the limitations affect of sample size recognition effect, but surpass the validity that 90% discrimination has illustrated method.

Claims (5)

1. the failure detector of a near field acoustic holography sound image pattern-recognition, comprise: scanning support, microphone array, reference source and data acquisition system (DAS), it is characterized in that: reference source is close to machinery, be arranged in the possible noise source position of machinery, scanning support is positioned at sound source one side, microphone array is fixed on the scanning support according to survey sound source sound field position by different way, can carry out side and top scanning survey to sound source, reference source microphone and scanning array microphone are connected with data acquisition system (DAS) simultaneously, storage of collected to time-domain signal and calculate testing result.

2. the detection method of the failure detector of a near field acoustic holography sound image pattern-recognition according to claim 1 is characterized in that, may further comprise the steps:

The first step, the machinery with a plurality of noise sources is carried out normal divide and setting with the fail operation state, the situation ratio of vibration source number and position generation significant change is easier to judge from hologram during for fault;

Second step, estimate sensitive frequency scope F determine holographic facet and source face apart from the sampling interval between Z and the holographic measurement face microphone (Δ x, Δ y), according to the definite holographic measurement face size of sound source face size (Lx, Ly);

The 3rd step, measurement mechanical under the condition of ground unrest unanimity normal with malfunction under sound pressure signal P (t), and set up sample storehouse S, carry out spectrum analysis and find sensitive frequency f, to each normal sound field sample, fault sound field sample employing near field acoustic holography;

The 4th step, the near field acoustic holography result is carried out feature extraction, structural attitude vector sample storehouse is convenient to sorter and is trained identification;

The 5th the step, obtain proper vector sample storehouse after, put into support vector machine and train classification, contrast the discrimination under the various conditions, obtain best classifier parameters: punishment parameters C and kernel function coefficient gamma, and feature extraction mode, the array signal that collects is in real time carried out feature extraction and adopts the sorter after optimizing to discern, realize fault diagnosis.

3. the detection method of the failure detector of near field acoustic holography sound image pattern-recognition according to claim 2 is characterized in that, described definite holographic measurement face size is meant: Lx 〉=1.2*lx﹠amp; Ly 〉=1.2*ly, wherein: Δ x≤λ/6, Δ y≤λ/6, Z≤λ/6 and Z 〉=max (Δ x, Δ y), λ are the wavelength of being determined by F.

4. the detection method of the failure detector of near field acoustic holography sound image pattern-recognition according to claim 2 is characterized in that, described sound pressure signal is:

$p\left({\stackrel{\→}{r}}_p\right)=\underset{S}{\∫\∫}[p\left({\stackrel{\→}{r}}_Q\right)\frac{\∂G({\stackrel{\→}{r}}_p,{\stackrel{\→}{r}}_Q)}{\∂n}+\mathrm{j\ρ}{\mathrm{cku}}_n\left({\stackrel{\→}{r}}_Q\right)G({\stackrel{\→}{r}}_p,{\stackrel{\→}{r}}_Q)]\mathrm{dS}\left({\stackrel{\→}{r}}_Q\right),$

Wherein: S represents the sound source surface, Q represents the border, the Un representation is to vibration velocity, G represents Green function, according to the suitable holographic algorithm of sound source structure choice: the acoustic pressure on the acoustical holography utilization measurement face of quadrature conformal structure is the convolution of source face acoustic pressure and Green function, fast two-dimensional fourier transformation is used for the Helmholtz equation, has realized rebuilding the distribution of acoustic pressure, vibration velocity and the sound intensity on the face of source by the acoustic pressure of holographic measurement face; The near field acoustic holography of arbitrary shape structure has BEM-based NAH and the equivalent source method based on numerical solution Helmholtz integral equation.

5. the detection method of the failure detector of near field acoustic holography sound image pattern-recognition according to claim 2 is characterized in that, described feature extraction is meant: adopt in the following dual mode any one:

A) directly the reconstruct matrix A is carried out svd and obtains singular value features:

$A=\mathrm{U\Λ}{V}^T=\underset{i=1}{\overset{k}{\mathrm{\Σ}}}{\mathrm{\α}}_i{u}_i{v}_i^T=\underset{i=1}{\overset{k}{\mathrm{\Σ}}}{\mathrm{\α}}_i{A}_i,$

Wherein: k is the order of matrix A, α _iBe i singular value of matrix A, make α ₁〉=α ₂〉=... 〉=α _n, m singular value constitutive characteristic vector before getting is used for the training classification;

B) be to utilize image processing techniques to extract the texture statistics feature of image B:

The known image gray level is n, the second order joint probability density function P by asking for image (d θ) obtains gray level co-occurrence matrixes for i, j:

M(d，θ)＝[P(i，i，j，d，θ)]，

Wherein: function P has described image distance on the θ direction has the appearance of gray scale i and j respectively for a pair of pixel of d probability, d=1, θ gets 0 °, 45 °, 90 ° and 135 ° respectively, obtain the gray level co-occurrence matrixes of four direction, and ask for 12 kinds of textural characteristics coefficients based on gray level co-occurrence matrixes: angle second moment value, contrast value, correlation, entropy, variance yields, contrary gap value and mean value and variance yields and entropy, difference mean value, difference variance and difference entropy, and be used for pattern-recognition.

Images