CN105353305A - Motor rotor eccentricity fault diagnosis method based on completely self-adaptive matrix pencil - Google Patents

Abstract

The invention provides a motor rotor eccentric fault diagnosis method based on a fully adaptive matrix beam, which solves the shortcomings of the traditional matrix beam algorithm in the motor rotor eccentric fault diagnosis, and has dual functions of adaptive threshold truncation and adaptive fundamental notch filter Effect. By calculating the maximum likelihood estimation of the singular value matrix, the singular value truncation threshold of the matrix bundle is adaptively adjusted; at the same time, the present invention utilizes the singular value matrix inside the matrix bundle to realize fully self-adaptive trapping of the fundamental wave current component. The time length of the current collected by the present invention is very short, and does not need to exceed 1 second (10kHz sampling frequency), so it is suitable for applications where the motor load changes slowly; the self-adaptive notch current fundamental wave component, and the characteristic accuracy of eccentric fault diagnosis are high , there is basically no possibility of misjudgment. According to the different rotation speeds and different loads of the motor, the threshold value of the singular value matrix truncation is adaptively set, and there is no possibility of failure of the diagnosis method.

Description

Motor Rotor Eccentric Fault Diagnosis Method Based on Fully Adaptive Matrix Bundle

technical field

The invention belongs to the field of fault diagnosis of motor rotors, and in particular relates to a method for diagnosing motor rotor eccentric faults with fully self-adaptive notch matrix beams.

Background technique

Motors exist in almost all industrial occasions, and most of the equipment is driven by motors. Electric motors, especially large ones, are the key to the normal functioning of society. Due to the extensive use of motors, failures will inevitably occur; especially in some occasions where the application environment is harsh and the load impact is large, the failure rate of the motor is high. In a complex production line, if a key motor fails to be found and maintained in time, the product quality of the entire production line will be affected; serious motor failures can cause industrial accidents and economic losses. In manned motor applications, the importance of motor status detection and fault diagnosis technology is self-evident.

Motor rotor eccentricity faults include static eccentricity, dynamic eccentricity and mixed eccentricity. The actual common eccentricity fault is mixed eccentricity fault. As a parameter identification method, matrix beam can be used to analyze the motor current signal containing noise, so as to detect the motor eccentricity fault. There are singular value decomposition and truncation process inside the traditional matrix beam, and its existence makes the matrix beam algorithm have a strong characteristic of eliminating noise. However, there are two problems with traditional matrix bundles in motor rotor fault diagnosis. First, the singular value truncation process needs to artificially set a suitable threshold, which is unknown in advance and can only be obtained through repeated trial and error verification. The threshold is closely related to the motor operating speed and load. Improper threshold setting will lead to failure of the matrix beam algorithm. Second, compared with the current fundamental component, the characteristic harmonics of rotor eccentricity in the current are weak characteristic signals, and the traditional and other improved matrix beam algorithms cannot eliminate the influence of the fundamental component on the identification accuracy of weak characteristic signals.

Contents of the invention

The purpose of the present invention is to provide a completely self-adaptive matrix beam method for diagnosing motor rotor eccentricity faults, which includes self-adaptive truncation threshold and self-adaptive notch. Its characteristic is that by calculating the maximum likelihood estimation of the singular value matrix, the singular value truncation threshold of the matrix bundle can be adaptively adjusted according to the different working conditions of the motor, so that the rotor eccentricity fault diagnosis method provided by the present invention will not Failure under any speed and load; at the same time, in view of the problem that the motor rotor eccentricity fault features are weak and the identification accuracy is difficult to guarantee, the present invention uses the singular value matrix inside the matrix bundle, and through several steps of operation, it can realize a fully adaptive notch base The wave current component does not need to specify conventional notch filter parameters such as notch frequency, and has an ideal adaptive notch function, thereby highlighting fault characteristics, improving the identification accuracy of rotor eccentric fault characteristics, and reducing the possibility of misjudgment.

The motor rotor eccentricity fault diagnosis method based on the fully self-adaptive notch matrix beam described in the present invention is suitable for the motor steady-state operation or progressive steady-state operation process, including the following steps:

1) Obtain any phase current of the motor through the current signal acquisition device (the duration does not need to exceed 1 second), and after preprocessing these continuous sampling points, obtain a data sequence of length N denoted as x[n];

2) Use the sequence x[n] to construct a (N-L)×(L+1)-dimensional Hankel matrix X;

3) Singular value decomposition is performed on the matrix X to obtain a singular value matrix Σ;

4) Calculate the likelihood estimation curve of the singular value matrix Σ, and seek its maximum likelihood estimation, and obtain the adaptive threshold value required for the truncation of the singular value matrix Σ;

5) Transform the matrix bundle correlation matrix to realize the effect of fundamental wave adaptive notch filter;

6) Complete the remaining steps of the traditional matrix beam algorithm, including singular value truncation, solving the generalized eigenvalues of the matrix beam, and solving the overdetermined equation; extracting the frequency and amplitude information of all harmonics in the identified low frequency band;

7) Taking the frequency as the horizontal axis and the amplitude as the vertical axis, after standardizing all the harmonic amplitudes, draw the rod diagram of all harmonics in the low frequency band, and diagnose whether there is an eccentric fault in the motor rotor according to the characteristics of the graph.

Further, the data preprocessing measures in step 1) are: to ensure that N needs to cover at least p fundamental current cycles, where p is the number of pole pairs of the motor; and the first sampling point must be a zero-crossing point.

Further, the Hankel matrix X form in the step 2) is:

where L is called the bundle parameter,

Further, the singular value decomposition method in step 3) is: X=UΣV ^T ; where the left singular matrix U=(u ₁ , u ₂ ... u _(L+1) ), the right singular matrix V=(v ₁ , v ₂ …v _(L+1) ), each column of U is the left singular per unitized eigenvector of matrix X, each column of V is the right singular per unitized eigenvector of matrix X; matrix Σ is the diagonal Matrix, the elements on the diagonal are the singular values of the matrix X, and they are arranged in descending order; the specific form of Σ is:

Further, the method for calculating the maximum likelihood estimation curve in said step 4) is:

Let {X _i |i=1～L+1} be samples from the population, σ ₁ ～σ _L+1 respectively correspond to the number of samples of these (L+1) samples, and the total number of samples is

${\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; a</span>}\mathrm{<span\; class="notranslate">\; l</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}$

The proportion of the number of each sample to the total number of samples is approximately equal to the probability p _i , that is

p _i =σ _i /σ _total

Then the likelihood function L of the first n samples is

$\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; +</span><span\; class="notranslate">\; ...</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; )</span>}{<span\; class="notranslate">\; \&Pi;</span>}}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}^{{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}{{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}^{{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; ...</span>{{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}}^{{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}}$

Taking the logarithm of the likelihood function, we get

$\mathrm{<span\; class="notranslate">\; l</span>}\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; l</span>}\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>$

When dln(L)/dn≈0, the maximum likelihood estimation is obtained, which indicates that the probability of the existence of the first n singular values is the largest. The minimum value of n that satisfies the establishment of dln(L)/dn≈0 is the number m of retained singular values. The logarithmic likelihood function is always negative, and the formula for the maximum likelihood estimation point is

(L(n+1)-L(n))/L(n)≈0

Further, the adaptive notch filtering method of the step 5) is:

Delete the first two columns of matrix U to get a new matrix U'; delete the first two columns of matrix V to get a new matrix V'; delete the first two rows and first two columns of matrix Σ to get a new matrix Σ'; decrement the value of m by 2 .

Further, the singular value truncation of the step 6), solving the generalized eigenvalue of the matrix bundle, and solving the overdetermined equation

The method is:

1) The new matrix Σ' is reduced to m dimensions, while only keeping the first m column vectors of the new matrix U' and the new matrix V';

2) Delete the first row of matrix V' to obtain matrix V ₁ '; delete the last row of matrix V' to obtain matrix V ₂ ';

3) Calculation matrix X ₁ = UΣV ₁ ′ ^T , calculation matrix X ₂ = UΣV ₂ ′ ^T ; at this time, there is no noise component in matrix X ₁ and X ₂ , expressed as

4) Construct a matrix bundle X ₂ -λX ₁ , all complex numbers λ satisfying det(X ₂ -λX ₁ )=0 are called the characteristic roots of the matrix bundle X ₂ -λX ₁ , and the set of characteristic roots of the matrix bundle {λ _i | i=1,2,...,m} are all Z-domain poles of the signal mode, which can be obtained by solving the eigenvalues of X ₁ ⁺ X ₂ , where X ₁ ⁺ is the pseudo-inverse of X ₁ ;

5) Solve the following overdetermined equations with the least squares method or its related algorithms to obtain the complex amplitude set {R _i |i=1,2,···,m}.

Further, the specific method of each harmonic frequency f _i and amplitude B _i calculation of said step 6) is:

f _i =(1/2π)Im(lnλ _i /T _s )

${\mathrm{<span\; class="notranslate">\; B</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; R</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; ,</span>& \begin{array}{cc}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; f</span>}& {\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&NotEqual;</span>\mathrm{<span\; class="notranslate">\; 0</span>}\end{array}\\ <span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; R</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; ,</span>& \begin{array}{cc}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; f</span>}& {\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}\end{array}\end{array}\right.$

Further, the method of diagnosing whether there is an eccentric fault in the motor rotor according to the characteristics of the map in step 7) is: according to the characteristic frequency f _ecc =f _s ±kf _r , k∈N ⁺ of the eccentricity of the motor rotor, as long as it is in the map, there will be The rotor eccentricity fault can be effectively judged if the fault side frequency band conforming to this formula appears; among them, for the asynchronous motor,

${\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}$

For permanent magnet motors,

${\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; p</span>}}{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}$

In the formula, p is the number of pole pairs of the motor, k is any positive integer, f _s is the power supply frequency, f _r is the frequency corresponding to the rotor speed of the motor.

The present invention has the following advantages:

(1) The time length of the current to be collected is very short and does not need to exceed 1 second (10kHz sampling frequency), so it is suitable for applications where the motor load changes slowly.

(2) The self-adaptive notch current fundamental wave component, the eccentric fault diagnosis feature has high precision, and there is basically no possibility of misjudgment.

(3) According to the different rotation speeds and different loads of the motor, the threshold value of the singular value matrix truncation is adaptively set, and there is no possibility of failure of the diagnosis method.

Description of drawings

Figure 1 Schematic diagram of vector control system for asynchronous motor,

Fig. 2 is a flow chart of rotor eccentric fault diagnosis method based on fully adaptive matrix beam.

Figure 3 The maximum likelihood estimation experiment results of the singular value truncation process

Figure 4 Diagram of test results of rotor eccentric fault diagnosis of asynchronous motor

detailed description

The present invention will be further elaborated below in conjunction with the accompanying drawings.

Figure 1 is an asynchronous motor frequency conversion speed regulation system. The three-phase AC power supply is uncontrolled rectified to obtain the DC bus voltage, which is supplied to the voltage source inverter. The weak current part adopts vector control mode, including voltage and current sensors, 3-phase/2-phase static coordinate transformation module, 2-phase static/2-phase synchronous speed coordinate transformation module, speed loop PI controller module, current loop PI controller module, 2-phase synchronous speed/2-phase static coordinate transformation module, voltage sinusoidal pulse width modulation module. These parts are the functional modules required by the vector control of the asynchronous motor, and the invention mainly relates to a fault diagnosis device for the eccentricity of the motor rotor.

In this embodiment, the motor used is a 15kW asynchronous motor, the number of pole pairs is p=3, the power supply frequency is 15Hz, the motor is unloaded, and the motor rotor has a mixed eccentric fault.

Figure 2 is the implementation steps of the eccentric fault diagnosis method, and the specific process of the entire method application is described below.

1. During the steady-state operation of the motor, select any phase from the three-phase current of the asynchronous motor measured by the current sensor as an input, and send it to the rotor eccentricity fault detection device. For example, intercept the A-phase current from 0.5s to 1s at a sampling frequency of 10kHz, and perform cycle integrity detection and zero-crossing detection. The number of cycles of the sampling current needs to cover an integer multiple of the number of motor pole pairs p, and the first sampling point must be the zero-crossing point.

2. Construct the Hankel matrix and perform singular value decomposition to obtain L+1 singular values.

3. Let {X _i |i=1～L+1} be samples from the population, σ ₁ ～σ _L+1 respectively correspond to the number of samples of these (L+1) samples, and the total number of samples is

${\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; a</span>}\mathrm{<span\; class="notranslate">\; l</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}$

When the sample is large, the proportion of the number of each sample to the total number of samples is approximately equal to the probability p _i , that is

p _i =σ _i /σ _total

Calculate the log-likelihood function of the first n points, namely

$\mathrm{<span\; class="notranslate">\; l</span>}\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; l</span>}\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>$

The maximum likelihood estimation is when dln(L)/dn≈0, which means that the probability of the existence of the first n singular values is the largest. The minimum value of n that satisfies the establishment of dln(L)/dn≈0 is the number m of singular values that need to be retained for singular value truncation. The logarithmic likelihood function is always negative, and the formula for the maximum likelihood estimation judgment threshold is

(L(n+1)-L(n))/L(n)≈0

Figure 3 is the change curve of the likelihood estimation curve with the value of n, which tends to be negatively saturated with the increase of n.

4. Temporarily do not perform the truncation operation, and instead perform the fundamental notch operation on the matrix. The operation process includes: subtracting 2 from the m value obtained in step 3; deleting the first two columns of the matrix U to obtain a new matrix U'; deleting the matrix V The first two columns of the new matrix V' are obtained; the first two rows and the first two columns of the matrix Σ are deleted to obtain the new matrix Σ'.

5. The new matrix U', V', Σ' replaced the old U, V, Σ; perform singular value truncation operations on the new U', V', Σ', solve the generalized characteristic root of the matrix bundle, and solve In the step of overdetermined equations, the generalized eigenvalues and complex amplitudes of the matrix bundle are finally obtained. The generalized eigenvalues are the poles of the signal and contain the frequency information of each harmonic of the signal. The complex magnitude contains the magnitude information of the individual harmonics. The conversion formula is as follows:

f _i =(1/2π)Im(lnλ _i /T _s )

${\mathrm{<span\; class="notranslate">\; B</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; R</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; ,</span>& \begin{array}{cc}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; f</span>}& {\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&NotEqual;</span>\mathrm{<span\; class="notranslate">\; 0</span>}\end{array}\\ <span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; R</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; ,</span>& \begin{array}{cc}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; f</span>}& {\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}\end{array}\end{array}\right.$

6. Taking the frequency as the horizontal axis and the amplitude as the vertical axis, after standardizing the amplitudes of all harmonics, draw a rod diagram of all harmonics in the low frequency band.

7. Figure 4 shows the rod-shaped spectrum obtained from the experiment. For the asynchronous motor used in the experiment in this embodiment, the number of pole pairs p=3, the given power supply frequency is 15Hz, and the motor slip rate s=0.001. It can be seen that the rotor frequency f _r for

${\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 4.995</span>}\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; z</span>}$

The black arrows in Fig. 4 point to the characteristic harmonic components caused by rotor eccentricity according to the present invention. It can be seen that at f _s -fr (10.005Hz), f _{s -2fr} ( _5.01Hz ), f At _s + f _r (19.995Hz), f _s + _2fr (24.99Hz), f _s + _3fr (29.985Hz), a rod-shaped mark appears. Therefore, it can be fully proved that the asynchronous motor in this embodiment has a rotor eccentric fault.

8. It can be seen from Figure 4 that due to the adaptive notch operation on the matrix, the rod-shaped feature corresponding to the power supply frequency of 15Hz does not appear, indicating that it is notch filtered.

9. Through this embodiment, the effectiveness of the present invention can be illustrated.

Claims (6)

1. a kind of motor rotor fault diagnosis method based on complete adaptive matrix bundle, is characterized in that, comprises the following steps: 1) Obtain any phase current of the motor through the current signal acquisition device, and the acquisition time does not need to exceed 1 second, and after preprocessing these continuous sampling points, a data sequence of length N denoted as x[n] is obtained; 2) Construct the Hankel matrix X of (N-L)×(L+1) dimension with the sequence x[n] in step 1); 3) performing singular value decomposition on the matrix X in step 2) to obtain a singular value matrix Σ; 4) Calculate the logarithmic likelihood estimation curve of the singular value matrix Σ in step 3), and implement the singular value matrix truncation threshold adaptive tuning based on maximum likelihood estimation; 5) Implement the adaptive fundamental wave notch filtering method; 6) Complete the remaining steps of the traditional matrix beam algorithm, including performing singular value truncation on the matrix in step 5), solving the generalized eigenvalues of the matrix beam, and solving the overdetermined equation; extracting the frequencies and Amplitude information; 7) Take the frequency obtained in step 6) as the horizontal axis, and the amplitude as the vertical axis. After standardizing the amplitude of all harmonics, draw a rod diagram of all harmonics in the low frequency band, and diagnose whether the motor rotor has eccentricity according to the characteristics of the graph. Fault. 2. The motor rotor fault diagnosis method based on fully adaptive matrix bundle according to claim 1, characterized in that: said method belongs to the motor rotor eccentric fault diagnosis method. 3. the motor rotor fault diagnosis method based on complete self-adaptive matrix beam according to claim 1, is characterized in that, the concrete method of sampling data preprocessing in the described step 1 is: The data length N needs to cover at least p fundamental current cycles, where p is the number of pole pairs of the motor; the first sampling point is the zero-crossing point. 4. the motor rotor fault diagnosis method based on complete adaptive matrix bundle according to claim 1, is characterized in that, in described step 4) the specific method based on the singular value matrix truncation threshold adaptive tuning of maximum likelihood estimation is : Let {X _i |i=1～L+1} be samples from the population, σ ₁ ～σ _L+1 respectively correspond to the number of samples of these (L+1) samples, and the total number of samples is ${\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; a</span>}\mathrm{<span\; class="notranslate">\; l</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}$ The proportion of the number of each sample to the total number of samples is approximately equal to the probability p _i , that is p _i =σ _i /σ _total Then the likelihood function L of the first n samples is $\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; +</span><span\; class="notranslate">\; ...</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; )</span>}{<span\; class="notranslate">\; \&Pi;</span>}}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}^{{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}{{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}^{{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; ...</span>{{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}}^{{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}}$ Taking the logarithm of the likelihood function, we get $\mathrm{<span\; class="notranslate">\; l</span>}\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; l</span>}\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>$ When dln(L)/dn≈0, the maximum likelihood estimation is obtained, which indicates that the probability of the existence of the first n singular values is the largest. The minimum value of n that satisfies the establishment of dln(L)/dn≈0 is the number of retained singular values m, and the formula for the maximum likelihood estimation point is (L(n+1)-L(n))/L(n)≈0 . 5. the motor rotor fault diagnosis method based on complete self-adaptive matrix bundle according to claim 1, is characterized in that, described step 5) in the concrete method of self-adaptive fundamental wave notch wave filtering is: Delete the first two columns of the left singular matrix U to obtain a new matrix U'; delete the first two columns of the right singular matrix V to obtain a new matrix V'; delete the first two rows and first two columns of the matrix Σ to obtain a new matrix Σ'; The m value is reduced by 2. 6. the motor rotor fault diagnosis method based on the complete self-adaptive matrix beam according to claim 1, is characterized in that, the method for diagnosing whether there is an eccentricity fault in the motor rotor according to the map features of the step 7) is: according to the eccentricity of the motor rotor eigenfrequency f _ecc =f _s ±kf _r ,k∈N ⁺ , as long as the fault sidebands conforming to this formula appear in the spectrum, the rotor eccentricity fault can be effectively judged; among them, for the asynchronous motor, ${\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}$ For permanent magnet motors, ${\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; p</span>}}{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}$ In the formula, p is the number of pole pairs of the motor, k is any positive integer, f _s is the power supply frequency, f _r is the frequency corresponding to the rotor speed of the motor.