CN115115864B - Dynamic balance testing method and system for pump precision rotor shaft - Google Patents
Dynamic balance testing method and system for pump precision rotor shaft Download PDFInfo
- Publication number
- CN115115864B CN115115864B CN202211028717.8A CN202211028717A CN115115864B CN 115115864 B CN115115864 B CN 115115864B CN 202211028717 A CN202211028717 A CN 202211028717A CN 115115864 B CN115115864 B CN 115115864B
- Authority
- CN
- China
- Prior art keywords
- spectrum
- value
- spectrogram
- distance
- rotor shaft
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M1/00—Testing static or dynamic balance of machines or structures
- G01M1/14—Determining unbalance
- G01M1/16—Determining unbalance by oscillating or rotating the body to be tested
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E30/00—Energy generation of nuclear origin
- Y02E30/30—Nuclear fission reactors
Abstract
The invention relates to the technical field of dynamic balance testing of rotor shafts, in particular to a method and a system for testing the dynamic balance of a precision rotor shaft of a pump, wherein the method comprises the steps of acquiring rotation data of the rotor shaft to be tested in the dynamic balance mode, and converting the rotation data into a first image and a second image to obtain a corresponding first frequency spectrogram and a corresponding second frequency spectrogram; for the first spectrogram and the second spectrogram, classifying the spectral data of the spectrograms to obtain a plurality of categories, obtaining an adaptive segmentation threshold value of each category, and denoising the spectrograms by using the adaptive segmentation threshold values to respectively obtain a denoised first spectrogram and a denoised second spectrogram; and carrying out inverse operation on the de-noised first frequency spectrogram and the de-noised second frequency spectrogram to obtain de-noised rotation data, and obtaining a dynamic balance test result of the rotor shaft by using the de-noised rotation data. Noise removal is carried out through the self-adaptive segmentation threshold values of multiple regions of the frequency spectrogram, and the dynamic balance testing precision and effectiveness of the rotor shaft are improved.
Description
Technical Field
The invention relates to the technical field of dynamic balance testing of a rotor shaft, in particular to a method and a system for testing the dynamic balance of a precision rotor shaft of a pump.
Background
The rotor shaft is important part in the pump, and be a precision part, play the power transmission effect, if the rotor shaft is when the dynamic balance test, receive the noise influence can't effectively accomplish the dynamic balance test, then the rotor shaft can produce unbalanced motion when transmitting power, lead to the power of being driven unstably, be unfavorable for the stable work of pump, and if the rotor shaft unbalanced degree of pump is too big, light then lead to the aggravation of rotor shaft wearing and tearing, heavy then can lead to the rotor shaft fracture, cause the incident, so when the rotor shaft drops into use, need carry out the dynamic balance test, there is the unbalanced condition in the rotor shaft of pump in order to prevent to drop into use.
The existing dynamic balance test method of the rotor shaft usually adopts a shaft center track for analysis to obtain a test result of the current dynamic balance of the rotor shaft, but when the dynamic balance test of the rotor shaft is carried out, an acquired signal often has noise, wherein the noise is generated because after the vibration amount of the rotor shaft during rotation reaches a certain amplitude, the whole rotor system can be subjected to electromagnetism with different strengths and noise generated by machinery, and then the noise occurs in a measurement result of a sensor.
At present, a sensor signal denoising method is to convert sensor signal data into two-dimensional image data, perform fourier transform on the two-dimensional image data to obtain a corresponding spectrogram, and remove noise in the sensor signal data according to a set fixed threshold of the spectrogram. However, if the set fixed threshold of the spectrogram is too large, data distortion is caused, and if the set fixed threshold of the spectrogram is too small, data denoising is unreasonable.
Disclosure of Invention
In order to solve the technical problems, the invention aims to provide a method and a system for testing the dynamic balance of a precise rotor shaft of a pump, and the adopted technical scheme is as follows:
in a first aspect, an embodiment of the present invention provides a method for testing dynamic balance of a precision rotor shaft of a pump, where the method includes the following steps:
the method comprises the steps of obtaining rotation data of a rotor shaft to be tested in dynamic balance, wherein the rotation data comprise a first moving distance sequence and a second moving distance sequence corresponding to a first moving distance and a second moving distance at two ends of the rotor shaft under a rotation period; correspondingly converting the first moving distance sequence and the second moving distance sequence of the plurality of rotating data into a first image and a second image respectively to obtain a first spectrogram of the first image and a second spectrogram of the second image;
for the first spectrogram and the second spectrogram, the spectrogram is divided into two parts based on symmetry, and any one part is taken as the spectral data to be analyzed of the corresponding spectrogram; respectively calculating the distance between each frequency spectrum point in the frequency spectrum data to be analyzed and the frequency spectrum point corresponding to the highest frequency spectrum value, calculating the clustering distance between any two frequency spectrum points according to the distance and the frequency spectrum value of each frequency spectrum point, and dividing the frequency spectrum points in the frequency spectrum data to be analyzed into N categories based on the clustering distance, wherein N is a positive integer; calculating a spectrum value variance value and an average distance value of each category, respectively obtaining an adaptive segmentation threshold corresponding to each category by using the spectrum value variance value and the average distance value, denoising the to-be-analyzed spectrum data by using the adaptive segmentation threshold, and symmetrically denoising the other part of the spectrogram by using the adaptive segmentation threshold to obtain a denoised spectrogram; the de-noised spectrogram comprises a de-noised first spectrogram corresponding to the first spectrogram and a de-noised second spectrogram of the second spectrogram;
and carrying out inverse operation on the de-noised first frequency spectrogram and the de-noised second frequency spectrogram to obtain corresponding de-noised rotation data, and obtaining a dynamic balance test result of the rotor shaft by using the de-noised rotation data.
Further, the method for acquiring the first image includes:
respectively carrying out normalization processing on the first moving distance sequence in each rotation period, and multiplying the normalized elements by 255 to obtain corresponding gray values;
and taking the gray value corresponding to the first moving distance sequence of n rotation periods as a first row of the first image, wherein n is a positive integer, taking the gray value corresponding to the first moving distance sequence of n +1 to 2n rotation periods as a second row of the first image, and repeating the steps in the same manner to obtain a first image with the size of m x n, wherein m is a positive integer.
Further, the method for acquiring the second image includes:
respectively carrying out normalization processing on the second moving distance sequence in each rotation period, and multiplying the normalized elements by 255 to obtain corresponding gray values;
and taking the gray value corresponding to the second movement distance sequence of n rotation periods as a first row of the second image, wherein n is a positive integer, taking the gray value corresponding to the second movement distance sequence of n +1 to 2n rotation periods as a second row of the second image, and repeating the steps in the same manner to obtain a second image with the size of m x n, wherein m is a positive integer.
Further, the method for calculating the clustering distance between any two spectrum points according to the distance and the spectrum value of each spectrum point comprises the following steps:
respectively calculating a distance difference absolute value and a spectrum difference absolute value between two spectrum points;
establishing a window with a set size by taking the two spectrum points as a central point, normalizing all spectrum values in the window to obtain normalized spectrum values of the two spectrum points, calculating a mean value of the normalized spectrum values of the two spectrum points, and taking the inverse number of the mean value as a first adjustment coefficient;
and acquiring a product of the absolute value of the spectrum value difference and the first adjusting coefficient, and taking an addition result of the product and the absolute value of the distance difference as the clustering distance between the two spectrum points.
Further, the method for obtaining the adaptive segmentation threshold corresponding to each category by using the variance value of the spectrum values and the average distance value includes:
acquiring the maximum distance in the spectral data to be analyzed according to the distance of each spectral point; calculating a spectrum value variance value according to the spectrum value of each spectrum point in the Nth category, and calculating an average distance value according to the distance of each spectrum point in the Nth category;
obtaining an initial segmentation threshold of the Nth category through Otsu method; and calculating the ratio between the average distance value and the maximum distance, acquiring the product between the reciprocal of the second adjustment coefficient and the ratio to obtain the addition result of the product and the variance value of the spectral values, and taking the product between the addition result and the initial segmentation threshold as the self-adaptive segmentation threshold in the Nth category.
Further, the method for denoising the spectral data to be analyzed by using the adaptive segmentation threshold includes:
comparing the spectrum value of each spectrum point in the spectrum data to be analyzed with the adaptive segmentation threshold of the category to which the spectrum value belongs, and setting the corresponding spectrum value to be 0 when the spectrum value is smaller than the corresponding adaptive segmentation threshold; when the spectral value is greater than or equal to the corresponding adaptive segmentation threshold, the corresponding spectral value is kept unchanged.
Further, the method for obtaining the dynamic balance test result of the rotor shaft by using the de-noised rotation data includes:
and acquiring a central track curve of the rotor shaft by utilizing the de-noised rotation data, and inputting the central track curve into the trained neural network to obtain a dynamic balance test result corresponding to the rotor shaft.
In a second aspect, an embodiment of the present invention further provides a system for testing dynamic balance of a pump precision rotor shaft, including a memory, a processor, and a computer program stored in the memory and running on the processor, where the processor implements the steps of any one of the methods when executing the computer program.
The embodiment of the invention at least has the following beneficial effects: the rotating data of the rotor shaft is converted into image data, a spectrogram corresponding to the image data is obtained through Fourier transformation, spectrum points of the spectrogram are clustered to obtain a plurality of categories, one category corresponds to one region, a self-adaptive segmentation threshold value of each region is obtained, spectral noise is removed from the spectrogram through the self-adaptive segmentation threshold value, the denoising effect is stabilized and enhanced, then the denoised spectrogram is subjected to inverse operation processing to obtain effective rotating data of the rotor shaft, namely the denoised rotating data, the effective rotating data is more beneficial to drawing a stable and effective shaft center track, and the dynamic balance testing accuracy and effectiveness of the rotor shaft of the pump are improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions and advantages of the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a flowchart illustrating steps of a method for testing dynamic balance of a precision rotor shaft of a pump according to an embodiment of the present invention.
Detailed Description
To further illustrate the technical means and effects of the present invention adopted to achieve the predetermined objects, the following detailed description of the method and system for testing the dynamic balance of the precision rotor shaft of a pump according to the present invention, with reference to the accompanying drawings and preferred embodiments, describes the specific implementation, structure, features and effects thereof. In the following description, the different references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The following describes a specific scheme of a method and a system for testing dynamic balance of a precision rotor shaft of a pump provided by the invention in detail by combining with the accompanying drawings.
The embodiment of the invention aims at the following specific scenes: and collecting the rotation data of the rotor shaft, drawing a shaft center track during the dynamic balance test of the pump precision rotor shaft, and completing the dynamic balance test of the pump precision rotor shaft according to the shaft center track.
Referring to fig. 1, a flow chart of steps of a method for testing dynamic balance of a pump precision rotor shaft according to an embodiment of the present invention is shown, the method including the following steps:
step S001, obtaining rotation data of a rotor shaft to be tested in dynamic balance, wherein the rotation data comprises a first movement distance sequence and a second movement distance sequence corresponding to a first movement distance and a second movement distance at two ends of the rotor shaft under a rotation period; and correspondingly converting the first moving distance sequence and the second moving distance sequence of the plurality of rotating data into a first image and a second image respectively to obtain a first spectrogram of the first image and a second spectrogram of the second image.
Specifically, two distance sensors are placed at two ends of a rotor shaft to be tested in dynamic balance, rotation data of the rotor shaft are obtained according to the distance sensors, namely a first moving distance and a second moving distance of the rotor shaft are obtained through the two distance sensors correspondingly, the first moving distance and the second moving distance of the rotor shaft at different positions in a rotation period are made to correspond to form a first moving distance sequence and a second moving distance sequence, and one sequence corresponds to one distance sensor.
Since the movement signal of the rotor shaft is small, the signal of the movement distance of the rotor shaft obtained by the distance sensor can be amplified.
After the first moving distance sequence and the second moving distance sequence are obtained, the first moving distance sequence and the second moving distance sequence can be converted into image data, denoising is carried out through a spectrogram, the denoised image data is restored into rotation data of the rotor shaft, and denoising of the rotation data of the rotor shaft is completed, so that the rotation data of the rotor shaft is converted into the image data, and the specific conversion method comprises the following steps: taking the rotation period of a rotor shaft to be tested in dynamic balance as a reference, firstly respectively carrying out normalization processing on a first moving distance sequence and a second moving distance sequence in one rotation period, and then multiplying by 255 to obtain corresponding gray values; taking the gray value corresponding to the first moving distance sequence of n rotation periods as a first line of the image data, wherein n is a positive integer, the gray value corresponding to the first moving distance sequence of n +1 to 2n rotation periods as a second line of the image data, and so on, so as to obtain a first image with the size of m × n corresponding to the first moving distance sequence, and m is a positive integer; and similarly, obtaining a second image with the size of m × n according to the gray value corresponding to the second moving distance sequence.
After the rotation data of the rotor shaft is converted into image data, a first spectrogram of the first image and a second spectrogram of the second image are respectively obtained by utilizing a Fourier transform algorithm, wherein the Fourier transform algorithm is a known technology and is not repeated in the scheme.
Step S002, for the first spectrogram and the second spectrogram, respectively dividing the spectrogram into two parts based on symmetry, and taking any one part as the spectral data to be analyzed of the corresponding spectrogram; respectively calculating the distance between each spectrum point in the spectrum data to be analyzed and the spectrum point corresponding to the highest spectrum value, and calculating the clustering distance between any two spectrum points according to the distance of each spectrum point and the spectrum value so as to divide the spectrum points in the spectrum data to be analyzed into N categories; and obtaining a self-adaptive segmentation threshold value of each category, denoising the spectral data to be analyzed by using the self-adaptive segmentation threshold value, and symmetrically denoising the other part of the spectrogram by using the self-adaptive segmentation threshold value to obtain a denoised spectrogram, wherein the denoised spectrogram comprises a denoised first spectrogram corresponding to the first spectrogram and a denoised second spectrogram of the second spectrogram.
Specifically, it is considered that setting a fixed threshold to the spectrogram can remove noise in the rotational data acquired by the distance sensor, but if the set fixed threshold is too large, data distortion may be caused, and if the set fixed threshold is too small, data denoising may be unreasonable. And noise in the rotation data acquired by the distance sensor appears as noise in the spectrogram, the noise belongs to high-frequency information in the spectrogram, and when the noise is more obvious, the energy of the high-frequency information in the spectrogram becomes higher, so that the threshold value of the high-frequency information should be increased during denoising.
Since the data in the spectrogram are symmetrical about the most central position of the spectrogram, the calculation amount can be reduced by acquiring half of the data in the spectrogram and analyzing the half of the data. Wherein, half of the data acquisition modes are as follows: and dividing the spectrogram into a left part and a right part along the center of the column number, and taking one part of the spectrogram as the spectral data to be analyzed.
Taking the first spectrogram as an example, the spectral data to be analyzed of the first spectrogram is obtained based on a half-data acquisition method. After the spectrum data to be analyzed is obtained, the spectrum points with the highest spectrum values in the spectrum data to be analyzed are obtained through traversing the spectrum data to be analyzed, the Euclidean distance is utilized to respectively calculate the distance L between each spectrum point in the spectrum data to be analyzed and the spectrum point with the highest spectrum value, one spectrum point corresponds to one distance, the larger the L value is, the higher the frequency of the periodic data represented by the current spectrum point is, and the larger the spectrum value of each spectrum point is, the more obvious the periodic response represented on the first spectrogram is.
Since the whole spectrum values in the first spectrogram have a gradual change trend, the larger the L value is, the lower the relative spectrum value is, however, a single segmentation threshold cannot well complete the threshold segmentation of the first spectrogram, so that different segmentation thresholds need to be selected according to different distances, and since high-frequency information in the spectrogram often shows a state of similar noise, when a region is selected and divided according to the L segmentation threshold, the proportion of the spectrum value to the distance result should be reduced, so that spectrum points with similar distances L and certain differences in the spectrum values can be classified into one class.
When dividing the spectrum points in the spectrum data to be analyzed, clustering by using a DBSCAN clustering method, wherein the clustering process not only considers the distances of the spectrum points, but also needs to consider the spectrum values for classification, so that the clustering distance between any two spectrum points is calculated based on the spectrum value and the distance of each spectrum point, and the calculation formula of the clustering distance is as follows:
wherein, the first and the second end of the pipe are connected with each other,is the clustering distance;is the distance of spectral point 1;is the distance of spectral point 2;is the spectral value of spectral point 1;is the spectral value of spectral point 2;is a first adjustment factor; | is an absolute value sign.
The smaller the value of (A) is, the more the two corresponding spectrum points are classified into the same class;the smaller the value of (A) is, the more similar the distance between two corresponding spectrum points is, the more the two spectrum points are classified into the same class;the smaller the value of (A) is, the more similar the spectrum value between two corresponding spectrum points is, the more the two spectrum points are classified into the same class.
It should be noted that the first adjustment coefficientIs negative, for reducingWeights in clustering process, wherein first adjustment coefficientThe acquisition method comprises the following steps: constructing a 5*5 window by taking the frequency spectrum point 1 as a central point, and acquiring the position in the 5*5 windowNormalizing the spectrum values of all the spectrum points in a 5*5 window to obtain a normalized spectrum value of a spectrum point 1, wherein the smaller the normalized spectrum value is, the larger the difference between the normalized spectrum value and a certain surrounding spectrum point is; similarly, a 5*5 window is established with the spectrum point 2 as a central point, spectrum values of all spectrum points in the 5*5 window are obtained, normalization is performed on the spectrum values of all spectrum points in the 5*5 window to obtain a normalized spectrum value of the spectrum point 2, an average value of the normalized spectrum values corresponding to the spectrum point 1 and the spectrum point 2 is further calculated, and the opposite number of the average value is used as a first adjustment coefficient。
Based on the clustering distance between any two spectrum points in the spectrum data to be analyzed, dividing all the spectrum points in the spectrum data to be analyzed into N categories by using a DBSCAN clustering method, wherein N is a positive integer.
If there should be different segmentation thresholds for different classes, taking the nth class as an example, the method for obtaining the adaptive segmentation threshold of the nth class includes: acquiring the maximum distance in the spectrum data to be analyzed according to the distance of each spectrum point, calculating a spectrum value variance value according to the spectrum value of each spectrum point in the Nth category, and calculating an average distance value according to the distance of each spectrum point in the Nth category; and obtaining an initial segmentation threshold of the Nth category by an Otsu method, and optimizing the initial segmentation threshold by using a spectrum value variance value, an average distance value and a maximum distance to obtain a self-adaptive segmentation threshold of the Nth category.
As an example, the adaptive segmentation threshold is calculated by the following formula:
wherein the content of the first and second substances,an adaptive segmentation threshold for the nth class;the variance value of the spectral value of the Nth category;average distance value of Nth category;is the maximum distance;an initial segmentation threshold for the nth class;is the second adjustment coefficient, and the empirical value in this scheme is 0.5.
The larger the variance value of the spectral value of a category is, the less periodicity the data in the spectrogram reflected by the spectral points in the corresponding category is, so that for the category with the larger variance value of the spectral value, the stronger the segmentation effect of the segmentation threshold value is, so as to filter out the data which does not belong to the periodicity in the spectrum or the data with poor periodicity, so that the variance value of the spectral value isThe larger the adaptive segmentation threshold representing the corresponding classThe larger the spectral values, i.e. the more spectral points that need to be filtered out.
The larger the value of (A), the more high frequency information is represented, and the noise of the rotation data of the rotor shaft is the high frequency information, so thatThe larger the value of (A) represents the adaptive segmentation threshold of the corresponding classThe larger the spectrum value, that is, the more the spectrum value of the spectrum point to be filtered;is the maximum distance of a spectral point in the spectral data to be analyzed,the larger the segmentation threshold adjustment, i.e. the adaptive segmentation thresholdThe larger.
The initial segmentation threshold of the nth class, i.e. the segmentation threshold obtained by the algorithm of the great amount of money, is used to distinguish between the effective spectrum points and the inefficient spectrum points.
And obtaining the self-adaptive segmentation threshold corresponding to each of the N categories of the frequency spectrum data to be analyzed by using a self-adaptive segmentation threshold obtaining method, wherein the frequency spectrum points of different categories are subjected to frequency spectrum value segmentation by adopting the corresponding self-adaptive segmentation threshold, namely, the frequency spectrum value of the frequency spectrum point is set to be 0 when the frequency spectrum value of the frequency spectrum point is smaller than the self-adaptive segmentation threshold of the category to which the frequency spectrum point belongs, otherwise, the frequency spectrum value is not changed, and then the frequency spectrum value of the frequency spectrum point belonging to the noise in the frequency spectrum data to be analyzed is removed, so that the denoised frequency spectrum data to be analyzed is obtained.
Because the spectral data to be analyzed is half spectral data in the first spectrogram, based on the symmetry of the first spectrogram, a plurality of adaptive segmentation thresholds obtained by the spectral data to be analyzed are used for symmetrically denoising another part of the spectral data of the first spectrogram, so that a complete denoised first spectrogram, namely the denoised first spectrogram, is obtained.
Similarly, a de-noised second spectrogram of the second spectrogram is obtained based on the method for obtaining the de-noised first spectrogram of the first spectrogram.
And S003, carrying out inverse operation on the de-noised first spectrogram and the de-noised second spectrogram to obtain corresponding de-noised rotation data, and obtaining a dynamic balance test result of the rotor shaft by using the de-noised rotation data.
Specifically, the denoised first spectrogram and the denoised second spectrogram are subjected to inverse operation to obtain corresponding denoised rotation data, and the specific inverse operation method is as follows: and obtaining the denoised image data by adopting inverse Fourier transform on the denoised spectrogram, converting the denoised image data into the sequence value of the image data according to the rotation data of the rotor shaft, and obtaining the rotation data of the rotor shaft through reverse operation, namely the denoised rotation data.
And when the de-noised rotation data of the rotor shaft is obtained, overlapping the data corresponding to the two distance sensors at the same moment to obtain a central track curve of the rotor shaft. And then, training a neural network by using the central track curve to obtain dynamic balance test results corresponding to different central track curves of the rotor shaft.
The training method of the neural network comprises the following steps: taking a data point of a central track curve of the rotor shaft as the input of a neural network, and outputting a dynamic balance test result of the corresponding rotor shaft, wherein the dynamic balance test result of the rotor shaft comprises a qualified result and an unqualified result; marking the data set as adopting a central track curve of the rotor shaft, and marking the data set as qualified data and unqualified data; the loss function is a cross entropy loss function; the neural network structure is an Encoder-FC network structure.
In summary, the embodiment of the present invention provides a method for testing dynamic balance of a precision rotor shaft of a pump, the method includes acquiring rotation data of a rotor shaft to be tested for dynamic balance, and converting the rotation data into a first image and a second image to obtain a corresponding first spectrogram and a corresponding second spectrogram; for the first spectrogram and the second spectrogram, classifying the spectral data of the spectrograms to obtain a plurality of categories, obtaining an adaptive segmentation threshold value of each category, and denoising the spectrograms by using the adaptive segmentation threshold values to respectively obtain a denoised first spectrogram and a denoised second spectrogram; and carrying out inverse operation on the de-noised first frequency spectrogram and the de-noised second frequency spectrogram to obtain de-noised rotation data, and obtaining a dynamic balance test result of the rotor shaft by using the de-noised rotation data. Noise removal is carried out through the self-adaptive segmentation threshold values of multiple regions of the frequency spectrogram, and the dynamic balance testing precision and effectiveness of the rotor shaft are improved.
Based on the same inventive concept as the method, the embodiment of the present invention further provides a system for testing the dynamic balance of the pump precision rotor shaft, which includes a memory, a processor, and a computer program stored in the memory and running on the processor, where the processor implements the steps of any one of the above methods when executing the computer program.
It should be noted that: the sequence of the above embodiments of the present invention is only for description, and does not represent the advantages or disadvantages of the embodiments. And specific embodiments thereof have been described above. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that are within the spirit of the present invention are intended to be included therein.
Claims (3)
1. A dynamic balance test method for a pump precision rotor shaft is characterized by comprising the following steps:
the method comprises the steps of obtaining rotation data of a rotor shaft to be tested in dynamic balance, wherein the rotation data comprise a first moving distance sequence and a second moving distance sequence corresponding to a first moving distance and a second moving distance at two ends of the rotor shaft under a rotation period; correspondingly converting the first moving distance sequence and the second moving distance sequence of the plurality of rotating data into a first image and a second image respectively to obtain a first frequency spectrogram of the first image and a second frequency spectrogram of the second image;
for the first spectrogram and the second spectrogram, the spectrogram is divided into two parts based on symmetry, and any one part is taken as the spectral data to be analyzed of the corresponding spectrogram; respectively calculating the distance between each frequency spectrum point in the frequency spectrum data to be analyzed and the frequency spectrum point corresponding to the highest frequency spectrum value, calculating the clustering distance between any two frequency spectrum points according to the distance and the frequency spectrum value of each frequency spectrum point, and dividing the frequency spectrum points in the frequency spectrum data to be analyzed into N categories based on the clustering distance, wherein N is a positive integer; calculating a spectrum value variance value and an average distance value of each category, respectively obtaining an adaptive segmentation threshold corresponding to each category by using the spectrum value variance value and the average distance value, denoising the to-be-analyzed spectrum data by using the adaptive segmentation threshold, and symmetrically denoising the other part of the spectrogram by using the adaptive segmentation threshold to obtain a denoised spectrogram; the de-noised spectrogram comprises a de-noised first spectrogram corresponding to the first spectrogram and a de-noised second spectrogram of the second spectrogram;
carrying out inverse operation on the de-noised first frequency spectrogram and the de-noised second frequency spectrogram to obtain corresponding de-noised rotation data, and obtaining a dynamic balance test result of the rotor shaft by using the de-noised rotation data;
the first image acquisition method comprises the following steps:
respectively carrying out normalization processing on the first moving distance sequence in each rotation period, and multiplying the normalized elements by 255 to obtain corresponding gray values;
taking the gray value corresponding to the first moving distance sequence of n rotating periods as a first row of the first image, wherein n is a positive integer, taking the gray value corresponding to the first moving distance sequence of n +1 to 2n rotating periods as a second row of the first image, and repeating the steps in the same manner to obtain a first image with the size of m x n, wherein m is a positive integer;
the second image acquisition method comprises the following steps:
respectively carrying out normalization processing on the second moving distance sequence under each rotation period, and multiplying the normalized elements by 255 to obtain corresponding gray values;
taking the gray values corresponding to the second moving distance sequences of n rotation periods as a first row of a second image, wherein n is a positive integer, taking the gray values corresponding to the second moving distance sequences of n +1 to 2n rotation periods as a second row of the second image, and repeating the steps in the same manner to obtain a second image with the size of m x n, wherein m is a positive integer;
the method for calculating the clustering distance between any two spectrum points according to the distance and the spectrum value of each spectrum point comprises the following steps:
respectively calculating the absolute value of the distance difference and the absolute value of the spectrum difference between two spectrum points;
respectively taking the two spectrum points as central points to construct a window with a set size, respectively normalizing all spectrum values in the window to obtain normalized spectrum values of the two spectrum points, calculating the mean value of the normalized spectrum values of the two spectrum points, and taking the inverse number of the mean value as a first adjustment coefficient;
obtaining a product of the absolute value of the spectrum value difference and the first adjusting coefficient, and taking an addition result of the product and the absolute value of the distance difference as a clustering distance between the two spectrum points;
the method for respectively obtaining the adaptive segmentation threshold corresponding to each category by using the variance value and the average distance value of the frequency spectrum values comprises the following steps:
acquiring the maximum distance in the spectrum data to be analyzed according to the distance of each spectrum point; calculating a spectrum value variance value according to the spectrum value of each spectrum point in the Nth category, and calculating an average distance value according to the distance of each spectrum point in the Nth category;
obtaining an initial segmentation threshold of the Nth category through Otsu method; calculating a ratio between the average distance value and the maximum distance, acquiring a product between the reciprocal of a set second adjustment coefficient and the ratio to obtain an addition result of the product and the spectrum value variance value, and taking the product between the addition result and the initial segmentation threshold as an adaptive segmentation threshold in the Nth category;
the method for obtaining the dynamic balance test result of the rotor shaft by using the de-noised rotation data comprises the following steps:
and acquiring a central track curve of the rotor shaft by using the de-noised rotation data, and inputting the central track curve into the trained neural network to obtain a dynamic balance test result corresponding to the rotor shaft.
2. The method for testing the dynamic balance of the pump precision rotor shaft according to claim 1, wherein the method for denoising the spectral data to be analyzed by using the adaptive segmentation threshold comprises the following steps:
comparing the spectrum value of each spectrum point in the spectrum data to be analyzed with the adaptive segmentation threshold of the category to which the spectrum value belongs, and setting the corresponding spectrum value to be 0 when the spectrum value is smaller than the corresponding adaptive segmentation threshold; when the spectral value is greater than or equal to the corresponding adaptive segmentation threshold, the corresponding spectral value is kept unchanged.
3. A pump precision rotor shaft dynamic balance testing system comprising a memory, a processor and a computer program stored in said memory and run on said processor, wherein said processor when executing said computer program implements the steps of the method of any one of claims 1-2.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211028717.8A CN115115864B (en) | 2022-08-26 | 2022-08-26 | Dynamic balance testing method and system for pump precision rotor shaft |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211028717.8A CN115115864B (en) | 2022-08-26 | 2022-08-26 | Dynamic balance testing method and system for pump precision rotor shaft |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115115864A CN115115864A (en) | 2022-09-27 |
CN115115864B true CN115115864B (en) | 2022-11-15 |
Family
ID=83336255
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211028717.8A Active CN115115864B (en) | 2022-08-26 | 2022-08-26 | Dynamic balance testing method and system for pump precision rotor shaft |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115115864B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116992297B (en) * | 2023-09-27 | 2023-12-15 | 广东石油化工学院 | Rotor monitoring method, device, terminal and medium based on segmented root mean square value ratio |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112687280A (en) * | 2020-12-25 | 2021-04-20 | 浙江弄潮儿智慧科技有限公司 | Biodiversity monitoring system with frequency spectrum-time space interface |
Family Cites Families (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR101908480B1 (en) * | 2017-08-14 | 2018-12-18 | 주식회사 디오티 | Device for extracting dynamic characteristics of structures |
CN108827539B (en) * | 2018-03-30 | 2020-02-21 | 江苏大学 | Online dynamic balance test device for multi-rotor assembly |
CN108801551B (en) * | 2018-08-28 | 2020-05-12 | 北京精密机电控制设备研究所 | Strong magnetic motor rotor dynamic balance testing arrangement |
CN110646138A (en) * | 2019-08-20 | 2020-01-03 | 北京化工大学 | Dynamic balance method and analysis device for rotary machine without key phase and trial weight |
CN112729152B (en) * | 2020-12-08 | 2022-03-25 | 广东工业大学 | Peak extraction method based on hyperbolic threshold denoising |
CN113009334B (en) * | 2021-02-18 | 2022-07-01 | 华中科技大学 | Motor fault detection method and system based on wavelet packet energy analysis |
AU2021104217A4 (en) * | 2021-07-16 | 2021-09-09 | Neeraj Chandnani | A system and method for optimum wavelet basis function selection for ecg arrhythmia denoising using artificial intelligence |
CN114323443B (en) * | 2021-11-30 | 2022-10-18 | 南京航空航天大学 | Reusable dynamic balance experiment platform and experiment method for engine rotor |
CN114112400A (en) * | 2021-12-01 | 2022-03-01 | 盐城工学院 | Mechanical bearing fault diagnosis method based on multi-angle information fusion |
CN113935998B (en) * | 2021-12-16 | 2022-03-22 | 武汉帕克橡塑制品有限公司 | Rubber and plastic part mottling detection method based on machine vision |
CN113989313B (en) * | 2021-12-23 | 2022-03-22 | 武汉智博通科技有限公司 | Edge detection method and system based on image multidimensional analysis |
CN114387272B (en) * | 2022-03-23 | 2022-05-24 | 武汉富隆电气有限公司 | Cable bridge defective product detection method based on image processing |
CN114627115B (en) * | 2022-05-13 | 2022-08-30 | 济宁安泰矿山设备制造有限公司 | Dynamic balance detection system and fluid pump impeller production line using same |
-
2022
- 2022-08-26 CN CN202211028717.8A patent/CN115115864B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112687280A (en) * | 2020-12-25 | 2021-04-20 | 浙江弄潮儿智慧科技有限公司 | Biodiversity monitoring system with frequency spectrum-time space interface |
Also Published As
Publication number | Publication date |
---|---|
CN115115864A (en) | 2022-09-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Pan et al. | An improved bearing fault diagnosis method using one-dimensional CNN and LSTM. | |
CN111238814B (en) | Rolling bearing fault diagnosis method based on short-time Hilbert transform | |
CN107657088B (en) | Rolling bearing fault diagnosis method based on MCKD algorithm and support vector machine | |
CN112084974A (en) | Multi-label rolling bearing fault diagnosis method based on meta-learning | |
CN110595780B (en) | Bearing fault identification method based on vibration gray level image and convolution neural network | |
CN110147760B (en) | Novel efficient electric energy quality disturbance image feature extraction and identification method | |
CN111562108A (en) | Rolling bearing intelligent fault diagnosis method based on CNN and FCMC | |
CN113405825B (en) | Belt conveyor fault diagnosis method based on sound signals | |
CN115115864B (en) | Dynamic balance testing method and system for pump precision rotor shaft | |
Du et al. | Fault diagnosis under variable working conditions based on STFT and transfer deep residual network | |
CN112418013A (en) | Complex working condition bearing fault diagnosis method based on meta-learning under small sample | |
CN110398370B (en) | Bearing fault diagnosis method based on HTS-CNN model | |
Zabin et al. | Hybrid deep transfer learning architecture for industrial fault diagnosis using Hilbert transform and DCNN–LSTM | |
CN111222442A (en) | Electromagnetic signal classification method and device | |
CN112906158A (en) | Mechanical fault diagnosis method based on multi-sensor multivariate data fusion | |
CN116030829A (en) | Heart sound classification method based on depth residual neural network | |
Pham et al. | Efficient fault diagnosis of rolling bearings using neural network architecture search and sharing weights | |
CN112364706A (en) | Small sample bearing fault diagnosis method based on class imbalance | |
CN114491823B (en) | Train bearing fault diagnosis method based on improved generation countermeasure network | |
Zhang et al. | Fault diagnosis using unsupervised transfer learning based on adversarial network | |
CN115908142A (en) | Contact net tiny part damage testing method based on visual recognition | |
CN114595728A (en) | Signal denoising method based on self-supervision learning | |
CN106295478A (en) | A kind of image characteristic extracting method and device | |
CN111611943A (en) | Gear classification method based on sound analysis | |
CN111289251A (en) | Rolling bearing fine-grained fault identification method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |