Array signal denoising method of optical fiber ultrasonic array sensor
Technical Field
The invention relates to an array signal denoising method of an optical fiber ultrasonic sensor for detecting partial discharge of electrical equipment, and belongs to the technical field of detection.
Background
The effective detection of partial discharges in electrical equipment is of great importance in the safe operation of the entire electrical power system. The local discharge ultrasonic array detection adopts an array sensor to receive ultrasonic signals, and the position of a local discharge source is determined according to characteristic information such as phase difference in the array signals. However, in actual operation, the receiving efficiency of the sensor is low, the sensitivity is low, the received ultrasonic signal is extremely weak, and the operation environment of the electrical equipment is very complex, so that the ultrasonic signal is often annihilated in strong noise, which leads to the reduction of the detection precision and even the failure of detection. Therefore, how to improve the sensitivity and the anti-interference capability of the array sensor is a problem for technicians.
Disclosure of Invention
The invention aims to provide an array signal denoising method of an optical fiber ultrasonic array sensor aiming at the defects of the prior art so as to improve the sensitivity and the anti-jamming capability of the array sensor.
The problems of the invention are solved by the following technical scheme:
an array signal denoising method of a fiber ultrasonic array sensor, the method comprising the steps of:
a. performing EMD (empirical mode decomposition) decomposition on signals collected by each array element in the fiber ultrasonic array sensor to obtain decomposed signals:
where x (t) represents the signal received by an element, c i (t) represents the i-th order IMF component (eigenmode component); r is n (t) represents the remainder;
b. reducing EMD decomposed signal
The number of array elements of the optical fiber ultrasonic array sensor is n, and the dimension of an IMF component obtained after EMD decomposition of a signal acquired by the ith array element is D i After EMD decomposition is carried out on signals acquired by all array elements, the dimension D of the obtained IMF component is = [ D ] 1 ,D 2 ,…,D n ]Keeping the first M-order IMF component of each array element signal, and discarding the rest signals, wherein M = min (D) 1 ,D 2 ,…,D n );
c. Performing cross-correlation operation on the corresponding IMF components of adjacent array element signals to obtain the time delay value, wherein the time delay error caused by all IMF components of the l order is as follows:
where the subscript numbers represent the array element designations,the time delay value from the local discharge source to the ith array element and the jth array element signal in the ith order is represented;
the vector operation error matrix delta formed by the delay errors of M-order IMF components is as follows:
δ=[δ 1 ,δ 2 ,…δ M ];
d. calculating the comprehensive support degree of each array element signal
Setting the measured data of the ith array element and the jth array element respectivelyIs x i And x j Calculating x i And x j Relative distance of (a):
d ij =abs|x i -x j |ij=1,2,…,n
calculating x i Is x by j Degree of support of (c):
r ij =e -dij
obtaining a support matrix:
defining the comprehensive support degree of the xi measured by other array elements as follows:
thereby obtaining the comprehensive support degree lambda = [ lambda ] 1 ,λ 2 ,···λ n ];
e. Calculating weights for IMF components of respective orders
The threshold on each IMF scale is defined as θ:
the weight of each order IMF component is:
the weight matrix corresponding to the IMF components is: ω = [ ω ] 1 ,ω 2 ,…ω M ];
f. Reconstruction of array sensor signals
The estimated signal after the ith array element is denoised is:
in the formula IMF i,l Is the I-th layer IMF component of the i-th array element.
The array signal denoising method of the optical fiber ultrasonic array sensor has the following specific processing process of performing EMD decomposition on signals acquired by an array element:
a. for any given array element signal x (t), firstly determining all extreme points on the x (t), connecting all the extreme points by a cubic spline curve to form an upper envelope curve, forming a lower envelope curve by the same method, and forming data x (t) and a mean value m of the upper envelope curve and the lower envelope curve by the same method 1 Is recorded as h 1 Namely:
h 1 =x(t)-m 1
h is to be 1 Regarding as new x (t), repeating the above steps until the ith h i Until IMF condition is satisfied, then h i Becomes the first order IMF component, denoted c, filtered from the original signal 1 (t);
b. C is to 1 (t) separating from x (t) to obtain a difference signal r from which the high frequency component is removed 1 (t), namely:
r 1 (t)=x(t)-c 1 (t)
handle r 1 (t) as a new signal, repeating step a until the difference signal of nth order:
r n (t)=r n-1 (t)-c n (t)
becomes a monotonic function and no longer can sieve out the IMF component, r n (t) is the remainder;
c. arranging IMF components of each order from high to low according to frequency, carrying out threshold processing on the IMF components of each order to obtain a new IMF signal, wherein the signal after EMD decomposition is represented as:
the array signal denoising method of the optical fiber ultrasonic array sensor is used for receiving the partial discharge ultrasonic signal and comprises an assembly body and a plurality of array elements fixed on the assembly body, wherein the assembly body is attached to a shell of detected electrical equipment, each array element comprises an optical fiber, an insulating support, a silicon sleeve and a silicon film, the silicon sleeve is fixed in a through hole in the assembly body, the silicon film is plugged at one end of the silicon sleeve and is opposite to the shell of the electrical equipment, the insulating support is plugged at the other end of the silicon sleeve, one end of the optical fiber is connected with a detection instrument, the other end of the optical fiber penetrates through the insulating support to enter the silicon sleeve and corresponds to the silicon film, and a gap is reserved between the optical fiber and the silicon film.
According to the array signal denoising method of the optical fiber ultrasonic array sensor, the side, facing the electrical equipment, of the silicon film is provided with the acoustic matching layer.
According to the array signal denoising method of the optical fiber ultrasonic array sensor, the magnet is embedded in the surface, attached to the shell of the detected electrical equipment, of the assembly body.
According to the array signal denoising method of the optical fiber ultrasonic array sensor, 16 array elements are arranged and are uniformly arranged along the same circumference.
According to the array signal denoising method of the optical fiber ultrasonic array sensor, the distance between adjacent array elements of the optical fiber ultrasonic array sensor is smaller than the half wavelength of the partial discharge ultrasonic signal.
The invention distinguishes the partial discharge signal and the noise signal according to the array signal closed vector time delay criterion on the basis of EMD decomposition of the partial discharge ultrasonic array signal, thereby more simply solving the denoising problem of the array signal, greatly improving the sensitivity and the anti-interference capability of the array sensor and providing a new path for processing the ultrasonic array signal.
Drawings
FIG. 1 is a schematic diagram of single array element light interference;
FIG. 2 is a schematic diagram of an array element structure;
FIG. 3 is a schematic structural diagram of a fiber-optic ultrasonic array sensor;
FIG. 4 isbase:Sub>A cross-sectional view A-A of FIG. 3;
FIG. 5 is a cross-sectional view B-B of FIG. 3;
FIG. 6 is a schematic diagram of a test of data received by a fiber optic ultrasonic array transducer;
FIG. 7 is a schematic diagram of a delay vector closing principle;
FIG. 8 is an EMD decomposition flow diagram;
FIG. 9 is a flow chart of array signal denoising.
The reference numbers in the figures are: 1. optical fiber, 2, an insulating support, 3, a silicon sleeve, 4, a silicon film, 5, an acoustic matching layer, 6, an array element, 7, an assembly body, 8 and a magnet.
The symbols in the text are: x (t) represents a signal received by an array element, c i (t) represents an i-th order IMF component; r is a radical of hydrogen n (t) represents the remainder, D i The dimension of IMF component obtained after EMD decomposition of the signal collected for the ith array element,the time delay value delta from the partial discharge source to the ith array element and the jth array element in the ith order l Delay error due to all IMF components of order I, d ij Is x i And x j Relative distance of (a), r ij Is x i Is x by j Degree of support of (D), λ i Is the integrated support of xi by the measured data of other array elements, theta is the threshold value on each IMF scale, omega l Is the weight, x, of each order IMF component i ' (t) is the estimated signal after the i-th array element is denoised.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings.
The design and detection principle of the array sensor is as follows:
laser light emitted from a laser light source is transmitted into the array element through an optical fiber filter along the optical fiber, wherein part of the light is reflected back to the optical fiber, and the rest of the light is refracted into a cavity formed between the end of the optical fiber and the surface of the silicon film, and the part of the laser light is on the surface of the optical fiber and on the surface of the silicon filmMultiple refraction and reflection (as shown in FIG. 1) occur on the surface of the silicon film, and a series of reflected light beams are coherently superposed (I in FIG. 1) R ) In which the reflected light I after superposition R The strength is related to the cavity length d, and the ultrasonic wave deforms on the silicon film to reduce the distance d between the silicon film and the end face of the optical fiber, so that the detection of the ultrasonic wave signal can be realized through the detection of the reflected light.
The structure of a single array element is shown in fig. 2, the single array element mainly comprises an optical fiber, a silicon film, a silicon sleeve, an acoustic matching layer and an insulating support, wherein the optical fiber is used for transmitting optical signals, the silicon film can convert ultrasonic signals into deformation, the insulating support plays a role in fixing, the silicon sleeve connects the silicon film and the optical fiber into a whole, the acoustic matching layer mainly reduces the loss of sound waves and comprises two layers of materials, namely a quartz single crystal material first layer and a silicon rubber material second layer. The invention adopts a sixteen-array-element circular array sensor (as shown in figure 3), in order to inhibit the side lobe effect, the distance between the array elements is smaller than half wavelength, and an assembly body of the array sensor is mainly used for fixing a single array element. Be provided with magnet on the equipment body and can adsorb on electrical equipment surface, aviation aluminum product is chooseed for use to the assembly body material, has better mechanical properties and formability, and is difficult to corrode in abominable external environment. The assembly body has convenient to carry, easy operation, and engineering technology advantages such as stable performance still have following advantage simultaneously except having: (1) The assembly body can be used in a plug-and-play manner, and can be replaced in time when a single array element is damaged or the sensitivity is reduced, so that the experiment is ensured to be completed smoothly; (2) The assembly body can also guarantee the consistency and the integrity of the contact between the receiving surface of each sensor array element and the outer wall of the electrical equipment when fixing the interval between the sensor array elements.
The process of collecting partial discharge signals by the array sensor is shown in fig. 6, monochromatic light emitted by a laser light source is firstly transmitted to each array element of the array sensor through a filter and a coupler and then through an optical fiber, incident light is reflected and interfered in each array element, the interference light is related to a cavity length d, ultrasonic waves generate deformation on a silicon film to change the size of the cavity length d, so that the strength of the interference light is influenced, at the moment, a sound wave signal is converted into an optical signal, the changed interference optical signal is collected by a multi-channel data collector through the coupler and a photoelectric converter, and finally, collected data are processed by an array signal processing method.
Closed vector delay criterion of ultrasonic array signals:
assuming that each sensor array element of the ultrasonic array is i, j, k, l, and the position of the partial discharge source is O, the time vector from the partial discharge source to each array element isAccording to the vector subtraction principle, the time delay vectors from the partial discharge source to the sensor array element are respectivelyAccording to the vector closure principle, the delay vectors of the array elements can form a vector quadrangle in the counterclockwise direction, and the quadrangle represents the closure criterion of the delay vectors, namely:
in the process of partial discharge signal transmission, different array element time delay differences exist and meet vector closure, and environmental noise or measurement interference generally exists and generally does not meet the vector closure, so that the ultrasonic array signal vector time delay closure is a necessary condition for judging and rejecting effective components and noise interference of an original signal.
EMD decomposition
The EMD method is a self-adaptive signal processing method, and decomposes a complex signal into the sum of a plurality of finite Intrinsic Mode Functions (IMFs) and a remainder according to the self characteristic time scale of a nonlinear and non-stationary signal, namely:
where x (t) represents the signal received by an array element, c i (t) represents the ith IMF component; r is n (t) represents the remainder.
The specific treatment process is as follows:
(1) For any given signal x (t), firstly determining all extreme points on x (t), connecting all the extreme points by a cubic spline curve to form an upper envelope curve, forming a lower envelope curve by the same method, and forming data x (t) and a mean value m of the upper envelope curve and the lower envelope curve 1 Is recorded as h 1 Then:
h 1 =x(t)-m 1 (3)
h is to be 1 Regarding as new x (t), repeating the above steps until h i If IMF condition is satisfied, it becomes the first order IMF, denoted as c, selected from the original signal 1 (t), typically the first order IMF component c 1 (t) contains the highest frequency component of the signal.
(2) C is to 1 (t) separating from x (t) to obtain a difference signal r from which the high frequency component is removed 1 (t), namely:
r 1 (t)=x(t)-c 1 (t) (4)
handle r 1 (t) repeating the step of (1) as a new signal until the residual signal of the nth order is a monotonic function and the IMF component can no longer be screened.
r n (t)=r n-1 (t)-c n (t) (5)
(3) Arranging IMF components of each order from high to low according to frequency, wherein the high-frequency components correspond to noise signals, carrying out threshold processing on each IMF component to obtain new IMF signals, and expressing the denoised signals as follows:
denoising an array signal:
signals x received by individual array elements i And (t) different array element signal delay values in the same IMF scale after EMD decomposition theoretically meet the delay vector matching criterion. The number of array elements is n, all signals are subjected to EMD, and the dimension D of the obtained IMF component is = [ D ] 1 ,D 2 ,…,D n ]Taking the first M-order IMF component of each array element to perform noise reduction processing, and discarding the rest signals, wherein M = min (D) 1 ,D 2 ,…,D n )。
And performing cross-correlation operation on the corresponding IMF components of adjacent array elements to obtain the time delay value, wherein the time delay errors caused by all IMF components of the l order are as follows:
the subscript numbers in the above formula represent the array element labels.
The delay error of the M-order IMF component may form a vector operation error matrix δ, as shown in equation (8):
δ=[δ 1 ,δ 2 ,…δ M ] (8)
the real ultrasonic signals are transmitted to the array sensor, the signal delay meets the vector closure criterion, the noise exists in the environment all the time, the probability that the noise meets the vector closure rule is very low and is almost 0, and therefore the delta matrix reflects the degree of signal correlation of each array element on different scales after multi-scale decomposition. Before the experiment, the performance of the array elements is evaluated respectively, and the comprehensive support degree lambda = [ lambda ] of each array element is given 1 ,λ 2 ,···λ n ]The concept of integrated support is given below.
Setting n array elements to measure the same target, wherein the measured data of the ith array element and the jth array element is x i And x j To reflect x i And x j Introducing the concept of relative distance, and setting the following steps:
d ij =abs|x i -x j l ij =1,2, …, n where d ij Represents x i And x j The relative position of (A) can be seen from the above formula ij The larger, x i And x j The larger the difference is, x i And x j The lower the mutual support degree of (A); d ij The smaller, x i And x j The smaller the difference of (A), (B), x i And x j The greater the mutual support of (a).
Introduction of a sum d ij Function r of correlation ij To represent x i And x j Mutual support degree between r ij The following two conditions are satisfied:
(a)r ij need to be at a relative distance d ij In inverse proportion to
(b)r ij The value range should be limited to [0,1 ]]I.e. r ij ∈[0,1]
If r ij If =0, x is indicated i And x j They are not compatible with each other, or they do not support each other; if r ij If =1, x is indicated i And x j Good compatibility, or they are mutually supported, 0<r ij &1, then x is indicated i And x j They are fused or supported in part. If one sensor is not supported by other sensors or is supported by only a few sensors and the support degree is poor, the observation value of the sensor is considered to be invalid, and the observation value is rejected when data fusion is carried out.
Degree of support r ij The invention provides a new support function
r ij =e -dij
The support matrix is:
wherein r is i1 、r i2 …r iN Respectively represent x i Is x by 1 、x 2 、x n To reflect the comprehensive support degree of xi observed by other sensors, the support degree of xi is defined as follows:
λ=[λ 1 ,λ 2 ,···λ n ]to a comprehensive degree of support
Defining the threshold value on each IMF scale as theta, then:
the weight of each order of IMF is:
at this time, the weight matrix corresponding to the IMF component is: ω = [ ω ] 1 ,ω 2 ,…ω M ]Therefore, the signal meeting the time delay vector closure criterion is a partial discharge signal, the signal with the weight of zero is a noise signal, and the true estimated signal can be obtained by multiplying the decomposed multi-scale signal by the corresponding weight, and the specific calculation formula is as follows:
in the formula IMF i,l Is the I-th layer IMF component of the i-th array element.
The method introduces the EMD decomposition into the denoising research of the partial discharge ultrasonic array signal for the first time, solves the defect that the traditional EMD decomposition can only denoise a single sensor and a single-channel signal, and reduces the direction-finding positioning error caused by the interference of a field noise signal.