CN-121977832-A - Rotating part circumferential matrix characterization method and system based on rotational speed super-resolution correction

CN121977832ACN 121977832 ACN121977832 ACN 121977832ACN-121977832-A

Abstract

The invention discloses a rotating part circumferential matrix characterization method and a rotating part circumferential matrix characterization system based on rotating speed super-resolution correction, which relate to the field of rotating machinery fault diagnosis and comprise the steps of collecting original vibration signals of a rotating part, carrying out synchronous compression wavelet transformation, extracting a rotating frequency ridge line, and obtaining a preliminary estimated rotating frequency sequence; the method comprises the steps of carrying out equal-angle resampling on an original vibration signal through a preliminary estimated frequency conversion sequence by utilizing a keyless phase order tracking technology, converting an equal-angle interval signal into an equal-angle resampling signal, carrying out pulse ridge line extraction on a circumferential matrix formed after equal-angle resampling by utilizing a sparse pulse ridge point dynamic programming tracking algorithm, carrying out sub-pixel level super-resolution correction on an initial rotating speed to obtain a corrected rotating speed signal, carrying out secondary angular domain resampling on the corrected rotating speed signal, and generating a circumferential characterization map capable of representing fault space-time distribution characteristics. According to the method, the rotating speed is subjected to fine correction, so that the interpretation and identification degree of two-dimensional characterization can be improved.

Inventors

YANG JIANWEI
ZHU BIN
YAO DECHEN
Hu Zhongshuo
WANG JINHAI
ZENG HENG
WANG NING
Huo Jiyuan
LI XI
ZHANG SHENGLONG

Assignees

北京建筑大学

Dates

Publication Date: 20260505
Application Date: 20260123

Claims (10)

1. The method for correcting the circumferential matrix representation of the rotating component based on the super-resolution of the rotating speed is characterized by comprising the steps of collecting original vibration signals of the rotating component, carrying out synchronous compression wavelet transformation, extracting a rotating frequency ridge line and obtaining a preliminary estimated rotating frequency sequence; The original vibration signal is subjected to equal-angle resampling through a preliminary estimated frequency conversion sequence by using a keyless phase order tracking technology, so that an equal-angle interval signal is converted into an equal-angle resampling signal; The method comprises the steps of performing circumferential matrix arrangement on equal-angle resampling signals according to equal-angle sampling frequency, and performing pulse ridge line extraction on a circumferential matrix formed after equal-angle resampling by using a sparse pulse ridge point dynamic programming tracking algorithm; Sub-pixel level super-resolution correction of the initial rotating speed is carried out by quantifying the position offset of the pulse points in adjacent periods, so as to obtain a corrected rotating speed signal; and (3) carrying out secondary angular domain resampling by using the corrected rotating speed signal to generate a circumferential characterization map capable of characterizing the space-time distribution characteristics of faults.
2. The rotational speed super-resolution correction rotating part circumferential matrix characterization method based on claim 1, wherein the obtaining the preliminary estimated frequency conversion sequence comprises identifying a frequency conversion of the original equidistant sampling vibration signal; synchronous compression wavelet transform SWT slave signal Continuous wavelet transform of (a) Firstly, obtaining a time-frequency representation TFR matrix result; obtaining synchronous compression wavelet transformation characterization result, time-frequency characterization TFR matrix, extracting ridge line with maximum energy from TFR matrix by adopting punished forward-backward greedy algorithm, and minimizing in each time slice To locate the maximum time-frequency ridge line, wherein Is the absolute value of the TFR matrix, by penalty factors The hopping of the frequency is limited and, Proportional to the distance between candidate ridgeline points; extracting ridge point set from TFR matrix, and linear interpolation of ridge points according to original time sequence to obtain primary estimated rotating speed signal 。
3. The method for characterizing a circumferential matrix of a rotating component based on rotational speed super-resolution correction as recited in claim 2, wherein said performing an equal angular resampling of the raw signal by the preliminary estimated rotation sequence comprises sampling the vibration acceleration signal of the mechanical component at generally equal time intervals, providing a vibration signal Collecting with fixed sampling frequency, and primarily estimating rotation speed signal as Obtain the time of Vibrating the acceleration signal for a spaced time domain sequence; The order tracking algorithm performs equal-angle sampling on the original non-stationary vibration signal to obtain an angular domain stationary signal, and for the vibration signal From instantaneous preliminary estimated rotational speed signal Calculating a reference angle; According to the basic characteristics of the change rule of the rotation frequency, different rotation periods correspond to a specific rotation angle value, sampling points with the same number as the undetermined coefficients are selected, and corresponding polynomial coefficients are fitted; according to the obtained time point, interpolation processing is carried out on the vibration signal to obtain the amplitude of the vibration signal after equiangular sampling, and the waveform of the signal after equiangular sampling is set as Equal angle resampling to convert the non-stationary signal sampled at equal intervals into stationary signal at equal intervals, and obtaining continuous instantaneous rotation speed curve Setting target angle axis as equidistant sequence with accumulated angle, obtaining time point corresponding to the equal angle by interpolation, and obtaining original vibration signal Linear interpolation resampling is carried out to obtain an angular domain signal after resampling by estimating the rotation speed angle And the angular sampling frequency needs to meet the nyquist criterion.
4. The rotating speed super-resolution correction rotating part circumferential matrix characterization method based on claim 3, wherein the sparse pulse ridge point dynamic programming tracking algorithm comprises the steps of performing double constraint on path programming by introducing amplitude and position penalty factors, and performing ridge point extraction on the circumferential matrix in a jumping sparse sampling mode; the method comprises the steps of formalizing a pulse ridge point extraction problem into an energy minimization problem, replacing row-by-row ridge point extraction by adopting a sparse sampling strategy, and sampling rows at fixed intervals; on the sampling line, before selection The extreme points are used as candidate ridge points to form a candidate point set, the ridge line extraction problem is converted into an optimal path selection problem among multiple candidate points on the basis of sparse sampling, and the state transition equation formally defines that the optimal path selection problem reaches the first point The first sampling line Minimum cumulative cost for each candidate point.
5. The method for characterizing a rotating member circumferential matrix based on rotational speed super-resolution correction as defined in claim 4, wherein said sub-pixel level super-resolution correction of the initial rotational speed to obtain a corrected rotational speed signal comprises, based on a sequence of equiangular sampling signals And the equiangular sampling frequency, rearranged into a circumferential matrix with each row being a circle of vibration behavior response, and the number of points of each row is equal angular resampling number of points The number of lines is Obtaining the rearrangement weeks Xiang Juzhen ; Reconstructed circumferential matrix The goal of ridge extraction is to find a mapping function such that for each row Ridge point location Corresponding to the location in the row where the pulse feature is located.
6. The rotational speed super-resolution based modified rotating member circumferential matrix characterization method of claim 5, wherein the performing secondary angular domain resampling using the modified rotational speed signal comprises converting a ridge extraction problem into an optimal path selection problem between multiple candidate points based on sparse sampling; Defining a cost function representing the cost from a candidate point to the next candidate point, wherein the cost function is used for measuring the cost when a path is selected and comprises accumulated cost, distance penalty term and value difference penalty term for reaching the previous candidate point; Defining a state transition equation representation to arrive at The first sampling line Minimum cumulative cost for each candidate point; for each sampling line Detecting the front in the current line The extreme points are used as candidate ridge points, and the extreme points are selected according to the magnitude of matrix values; starting from the second sampling line, the dynamic programming process gradually calculates the minimum cost reaching each candidate point, and records the precursor node of the optimal path; in the initialization stage, all candidate point costs of the first sampling line are set to 0; In the state transfer process, for each candidate point, calculating the transfer cost between the candidate point and all candidate points of the previous sampling line, selecting the minimum cost path, and simultaneously recording the optimal precursor node; After the dynamic programming is completed, the optimal path is searched back from the last sampling line; Obtaining a ridge point position set on a sparse sampling line For non-sampling rows, filling the positions of the ridge points by adopting a spline interpolation method; According to the original matrix Linear interpolation is carried out on the sparse pulse ridge point position by the line number, and resampling is carried out to obtain ; Deducing the relation between the position difference of the pulse ridge point and the corrected rotating speed, setting the first The positions of the circle pulse points are First, the The positions of the circle pulse points are Then its pulse point location interpolates Obtaining the sampling angle between every two adjacent points according to the sampling points of each circle, calculating the first Circle and the first The time interval of the circle is used for obtaining the rotating speed correction quantity, the rotating speed correction quantity is subjected to linear interpolation to enable the length of the rotating speed correction quantity to be consistent with the length of the original vibration signal and the rotating speed signal, an interpolated rotating speed correction sequence is obtained, and the corrected rotating speed is calculated through time-frequency characterization estimation of the rotating speed.
7. The method for characterizing a circumferential matrix of a rotating member based on rotational speed super-resolution correction as recited in claim 6, wherein generating a circumferential characterization map capable of characterizing a space-time distribution characteristic of a fault comprises performing two-dimensional image conversion by using a speed-corrected angle subsampled signal, since each cycle is resampled The CM matrix after secondary resampling can be obtained by a plurality of points; Mapping the CM matrix after secondary resampling to a three-dimensional polar coordinate system, wherein the position of any point is determined by the length of a connecting line between the current point and a fixed origin and the included angle between the connecting line and a reference direction, and the conversion relation between the current point and a rectangular coordinate system is that , ; And (3) with Dividing the time sequence signal according to a rotation period, wherein each period corresponds to a complete circumference in a polar coordinate system, sampling points in each period are mapped to equally divided points on the circumference, and signal amplitudes of the sampling points are mapped to the polar diameters of the sampling points; The number of turns is used as the polar diameter, the amplitude is used as the unit8 triplet value of the pixel rgb, and the position of the surface damage of the rotary mechanical structure is positioned, and the time-space relationship of faults is represented.
8. The rotating part circumferential matrix characterization system based on the rotating speed super-resolution correction is based on the rotating speed super-resolution correction rotating part circumferential matrix characterization method according to any one of claims 1-7, and is characterized by comprising a rotating speed estimation module, a virtual speed sensing information, a first speed sensor module and a second speed sensor module, wherein the rotating speed estimation module performs time-frequency characterization on vibration signals of the rotating part by utilizing synchronous compression wavelet transformation SWT; The angular domain signal processing module is used for resampling the signal at equal angles and converting the non-stationary signal into a stationary signal; The ridge point extraction module is used for extracting ridge points from the circumferential matrix through a dynamic programming method based on a sparse pulse ridge point extraction algorithm, selecting an optimal path from a plurality of candidate ridge points and correcting a rotating speed signal; And the space-time feature mapping module is used for carrying out secondary angle resampling by adopting the corrected rotating speed information, mapping the signals to a three-dimensional polar coordinate system, and representing space-time fault features by polar coordinates to obtain a map capable of representing space-time distribution of faults.
9. The computer equipment comprises a memory and a processor, wherein the memory stores a computer program, and the computer program is characterized in that the processor realizes the steps of the rotating part circumferential matrix characterization method based on the rotational speed super-resolution correction according to any one of claims 1-7 when executing the computer program.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method for characterizing a rotational speed super-resolution based modified rotating member circumferential matrix according to any one of claims 1 to 7.

Description

Rotating part circumferential matrix characterization method and system based on rotational speed super-resolution correction Technical Field The invention relates to the technical field of fault diagnosis of rotary machinery, in particular to a method and a system for correcting a circumferential matrix of a rotary part based on rotational speed super-resolution. Background Gears, universal shafts and the like are important components of mechanical transmission systems, and are widely applied to traffic and transportation equipment such as automobiles, trains and the like. Its health directly affects the stability and safety of the mechanical system. Because of internal and external excitation such as component interference and variable load, various types of faults often occur on gears and universal shafts, so that the system is abnormal in vibration, and safety accidents are caused and personnel and property losses are caused when the system is serious. And therefore health status monitoring thereof is particularly important. However, under the condition of incomplete data, the rotating speed working condition is generally fuzzy and unknown, and the vibration signals collected under the variable speed working condition often show non-stable characteristics, so that the positioning of faults of key components is challenged. One of the current mainstream fault diagnosis methods is to map one-dimensional vibration signals into two-dimensional feature maps and input the two-dimensional feature maps into a deep learning model for diagnosis. The frequency spectrum difference is smaller under various faults of gears and universal shafts, and the prior method has insufficient attention on the aspects of interpretability, physical significance and the like of the representation result. Disclosure of Invention The present invention has been made in view of the above-described problems occurring in the prior art. Therefore, the invention provides a characterization method for correcting the circumferential matrix of the rotating component based on the rotating speed super-resolution, which solves the problems that the research rotating speed working condition has a certain interpretability and is not interfered by prior parameters, and provides high-quality input for an identification model. In order to solve the technical problems, the invention provides the following technical scheme: in a first aspect, the invention provides a rotating component circumferential matrix characterization method based on rotational speed super-resolution correction, which comprises the steps of collecting original vibration signals of a rotating component, performing synchronous compression wavelet transformation, extracting a frequency conversion ridge line, and obtaining a preliminary estimated frequency conversion sequence; The original vibration signal is subjected to equal-angle resampling through a preliminary estimated frequency conversion sequence by using a keyless phase order tracking technology, so that an equal-angle interval signal is converted into an equal-angle resampling signal; The method comprises the steps of performing circumferential matrix arrangement on equal-angle resampling signals according to equal-angle sampling frequency, and performing pulse ridge line extraction on a circumferential matrix formed after equal-angle resampling by using a sparse pulse ridge point dynamic programming tracking algorithm; Sub-pixel level super-resolution correction of the initial rotating speed is carried out by quantifying the position offset of the pulse points in adjacent periods, so as to obtain a corrected rotating speed signal; and (3) carrying out secondary angular domain resampling by using the corrected rotating speed signal to generate a circumferential characterization map capable of characterizing the space-time distribution characteristics of faults. The method for representing the circumferential matrix of the rotating component based on the rotational speed super-resolution correction comprises the steps of obtaining a preliminary estimated frequency conversion sequence, and identifying the frequency conversion of an original equal-time interval sampling vibration signal; synchronous compression wavelet transform SWT slave signal Continuous wavelet transform of (a)Firstly, obtaining a time-frequency representation TFR matrix result; obtaining synchronous compression wavelet transformation characterization result, time-frequency characterization TFR matrix, extracting ridge line with maximum energy from TFR matrix by adopting punished forward-backward greedy algorithm, and minimizing in each time slice To locate the maximum time-frequency ridge line, whereinIs the absolute value of the TFR matrix, by penalty factorsThe hopping of the frequency is limited and,Proportional to the distance between candidate ridgeline points; extracting ridge point set from TFR matrix, and linear interpolation of ridge points acco